Prestressing

Define prestressing tendons, compute losses, check transfer and service stresses, and verify ultimate capacity for prestressed and post-tensioned concrete sections.

Overview

Prestressing applies a compressive force to the concrete before service loads act, offsetting the tensile stresses that would otherwise cause cracking. ACS supports both pretensioned (bonded) and post-tensioned (bonded or unbonded) sections.

The prestressing workflow in ACS has three analysis stages:

1. Losses — compute how much of the initial jacking force is lost over time
2. Transfer and service stresses — verify concrete stresses remain within allowable limits at two critical stages
3. Ultimate capacity — check the moment capacity of the prestressed section at ULS

All three analyses run automatically once you define tendons and the section has valid geometry and materials.

When to use prestressing analysis

Use the prestressing features when designing:

• Precast pretensioned members (hollowcore slabs, bridge girders, prestressed piles)
• Post-tensioned beams or slabs with bonded or unbonded tendons
• Members where crack control under service loads is critical (water-retaining structures, long-span floors)

Defining tendons

Open the PT tab in the right panel to add and configure prestressing tendons.

Adding a tendon

Click Add Tendon to insert a new tendon. Each tendon requires:

:::caution[Jacking stress is required — August 2026 (#4882)] A tendon with no declared jacking stress $f_{pi}$ is now refused rather than analysed with an implicit or zero value (§874 fix, tracking #4818). If you saved a section with a tendon missing its jacking stress, open the PT panel, enter $f_{pi}$, and re-save before running the analysis. The refusal message names the tendon and states that the jacking stress is not set.

The jacking stress gates every loss computation — elastic shortening, friction, draw-in, creep, shrinkage, and relaxation all require it — so the engine has no honest way to continue without it. :::

The last two are per tendon, not per section, because friction accumulates along one duct from one anchorage. Two tendons on the same section genuinely differ whenever their drapes differ, their lengths differ, or they are jacked from opposite ends. They feed the friction loss described under Prestress losses below, together with the default anchorage distance and coefficients you set once in the PT panel.

Edge-relative tendon groups

When defining a tendon in edge-relative mode, you can place a group of evenly-spaced tendons along one edge in a single operation rather than adding them one by one.

The group is centred on the tangential position you enter: for a count of 3 and spacing of 200 mm, the group spans from −200 mm to +200 mm relative to the tangential position, so the centre tendon lands exactly there. The maximum group size is subject to an application cap — the dialog shows the limit and blocks Apply when it is exceeded.

Selecting an edge by clicking

You can select the edge for an edge-relative tendon in two ways: via the edge selector dropdown, or by clicking the edge directly in the live section preview beside the dialog. Clicking a straight edge in the preview sets it as the selected edge and resets the tangential position to the edge midpoint.

Clicking an arc edge is not supported — edge-relative placement is defined by perpendicular and tangential distances measured along the chord, which has no consistent meaning on a curved edge. The dropdown filters arc edges for the same reason.

Default position at the edge midpoint

When the dialog opens to create a new edge-relative tendon, the tangential distance defaults to the midpoint of the selected edge. This gives a sensible starting position away from the corners without requiring you to type a value before previewing the tendon. Changing the selected edge — via the dropdown or by clicking the preview — resets the tangential distance to the new edge’s midpoint.

An edit to an existing tendon always restores the persisted tangential distance.

Out-of-section warning

If any tendon in the group would fall outside the section boundary — because the tangential distance places it past the edge end, or because the group spacing projects members beyond the section outline — ACS shows an amber warning naming the affected tendon indices. The warning is non-blocking: you can still apply the pattern, and it is saved, but the engine will refuse to analyse a section whose tendons lie outside the concrete and will identify this pattern. Adjust the tangential distance or spacing to bring all tendons inside the section before running the analysis.

Custom tendon placement gate

For custom (non-edge-relative) tendons, the Apply button is disabled — not just warned — when the duct footprint falls outside the section outline. This is a hard gate: unlike the amber warning for edge-relative groups, it prevents placement rather than flagging it after the fact.

The footprint test uses the actual duct geometry:

• Round duct: the circular footprint centred on the placement point, using the stored duct diameter.
• Flat duct: the axis-aligned width × height bounding box centred on the placement point. Five points are tested — the centre and all four corners of the box — and every one must lie inside the section outline.

:::caution[Flat duct footprint corrected — September 2026 (#5641)] Prior to this release, the Apply gate for a flat duct probed a circle of radius max(ductWidth, ductHeight) / 2 rather than the actual axis-aligned bounding box. A wide, shallow duct — such as a 90 × 20 mm slab tendon — was tested as a circle of radius 45 mm, overstating the perpendicular half-extent by 25 mm. That caused placement to be refused even when the true duct box sat comfortably inside the section. Conversely, a duct whose corners genuinely lay outside the boundary could pass the circular check. The gate now tests the correct geometry. Wide-shallow flat ducts (typical in post-tensioned slabs) that were previously blocked near a section edge may now be accepted — and some placements that were previously allowed may now be correctly refused. Review any saved tendon position near the section boundary and re-run the analysis to confirm placement. :::

Duct shape

Each tendon duct can be specified as either a round (circular) or flat (stadium-profile) cross-section.

Select the duct shape in the tendon editor. When Standard sizing mode is active, ACS suggests dimensions from the flat-duct catalogue below; above its capacity, or for a strand diameter it does not cover, it suggests a round duct sized for groutability instead. Switch to Custom to enter exact dimensions.

Flat duct catalogue

Flat ducts are manufactured in three discrete widths, not to a continuous size range, so ACS suggests a catalogue size rather than computing one. Capacity depends on the strand diameter — the 90 mm duct takes six 12.7 mm strands but only five 15.2 mm.

Widths and capacities follow the SRG Global Prestressing Technology catalogue (rev. 2020-12); the 70 mm and 90 mm sizes and the strand capacities are common to every Australian supplier checked. Sources differ on the height — SRG states 19 mm, while the Post-Tensioning Institute of Australia’s Practical Prestress Detailing and two stockists state 20 mm. ACS uses 20 mm, the majority figure and the conservative one for the Cl. 8.2.1.5 duct deduction. These are outside dimensions, over a 0.4 mm galvanised wall.

The suggestion is a starting point, not a constraint: switch to Custom whenever your supplier’s product differs.

:::note[Tendons created before 2026-08-11] Earlier versions sized flat ducts from a formula rather than this catalogue, so a tendon created then carries the dimensions it was given — they are a saved input and are not rewritten. Existing results do not change. A tendon created now will differ from an otherwise identical older one; the duct dimensions are shown in the tendon editor, and you can set them to the catalogue size by hand. :::

:::caution[Shear deduction may be incorrect — flat duct tendons created before 2026-08-11] Flat duct tendons created before 2026-08-11 carry formula-computed dimensions that may differ substantially from the catalogue sizes above. As one example, the old formula suggested 60 × 18 mm for 3 strands of 12.7 mm strand; the catalogue gives 43 × 20 mm — a cross-sectional area 32 % larger. Because the duct width deducted from the effective web width $b_v$ under AS 3600 Cl. 8.2.1.5 comes directly from the stored duct dimensions, an over-sized formula input over-deducts from $b_v$ and under-states the shear capacity. Check the duct dimensions in the tendon editor and, where they differ from the catalogue, update them to the catalogue values and re-run the shear check. :::

The flat duct profile is used in post-tensioned construction. Flat duct tendons are always classified as post-tensioned to the loss engine — they attract wobble friction, curvature friction and anchorage draw-in losses; pretensioned tendons do not, because they are stressed against a fixed abutment rather than anchored in the hardened concrete. Multiple strands are grouped side-by-side in a shallow band, which is common in post-tensioned slabs. The width and height of the flat duct feed the AS 3600 Cl. 8.2.1.5 effective web width ($b_v$) deduction used in shear checks — the full flat-duct width is deducted from $b_v$ rather than a single circular diameter.

:::caution[Results change — flat duct tendons before 2026-08-09] Prior to PR #4488, flat duct tendons were incorrectly classified as pretensioned to the loss engine. Friction and anchorage draw-in losses were therefore not computed for them. If your section contains flat duct tendons and was last analysed before 2026-08-09, rerun the analysis to obtain correct loss results. :::

Bond type on saved sections

Bond type decides how a tendon’s stress is derived at ultimate, and it reaches four separate results: the tendon-stress law itself (a bonded strand follows the local fibre strain, an unbonded one carries the member-average stress capped by AS 3600 Cl. 8.1.8), whether the duct is deducted from the transformed section, whether the tendon counts toward the bonded crack-control detailing check, and the Cl. 8.2.1.5 duct factor $k_d$ in the shear deduction.

A tendon saved before September 2026 carries no bond declaration, because the bond selection was not stored on the section — it existed only on the older payload analysis endpoints. Those tendons are analysed as bonded (grouted), which is the reading every ACS result for them has always used, so no saved number changes. The assumption is now stated rather than taken silently: the analysis warnings say how many tendons it was applied to.

:::note[Where the notice appears] The undeclared-bond notice is a section-level analysis warning. In the editor it appears at the top of the Design Summary, above the governing checks, because it qualifies every verdict below it rather than any one of them. It is also printed in the design report.

Through the API and the MCP tools it is on the sectionWarnings field of every by-id derived read’s response envelope — flexure, properties, batch-design-check and the rest — not only analyze. The individual result panels in the editor do not repeat it; the Design Summary is the one place it is shown. :::

Bonded is the less conservative reading for flexure, so if a duct is genuinely ungrouted, declare it — an unbonded duct analysed as grouted overstates flexural capacity.

:::caution[Results change — tendon bond type now correctly persisted (#5564)] Prior to the September 2026 release, the tendon bondType ("bonded" / "unbonded") set via the API or MCP tools was not persisted to the section record. On every by-id read and in the generated PDF report, the engine read the bond type back as bonded regardless of the value sent at save time — so unbonded post-tensioned sections were being analysed with the strain-compatibility law rather than the AS 3600 Cl. 8.1.8 member-average formula. §874. The persistence path is now corrected.

If you have an unbonded PT section authored through the API or MCP tools before this release, re-run the analysis. The tendon stress at ultimate ($f_{ps}$) will now be derived from the correct unbonded formula, and $\phi M_{uo}$ may differ. Sections authored in the editor before that release are unaffected — no bond-type control existed there at the time, so those tendons carry no declaration and continue to be analysed as bonded. To declare one now, open the tendon and set Bond (see Declaring bond type in the editor).

:::

Declaring bond type in the editor

The tendon dialog carries a Bond control — a Bonded / Unbonded pair sitting beside Duct material, because the two together fix the AS 3600 Cl. 8.2.1.5 duct factor $k_d$. It defaults to Bonded, the common design case and the same reading the engine applies to an undeclared tendon, and an Apply stores the value shown: a tendon you save through the dialog carries an explicit declaration rather than relying on the fallback.

Editing an existing tendon seeds the control from its saved declaration, so changing a duct dimension will not quietly re-grout an unbonded duct. Declaring Unbonded also surfaces the member span-to-depth ratio input in the PT panel, which the Cl. 8.1.8 ultimate-stress formula needs.

Strand position within the duct

A strand does not sit on its duct’s centreline. At a sagging section the tendon is concave up, so the resultant of the tangential tension forces points up, toward the centre of curvature — equilibrium requires the duct to push back down, and only the duct’s top surface can do that. At a hogging section the curvature reverses and the strand bears on the bottom.

The difference is small but it is one-directional: at a sag the strand sits higher than the centreline, so modelling it on the centreline overstates the effective depth $d_p$ and therefore overstates the flexural capacity. On a 450 mm deep member with a 19 mm duct the error is around 0.6% of $\phi M_u$; it is proportionally larger in a shallow slab, where the duct is a bigger fraction of the depth.

Strand position in duct is where you declare it, per tendon. It offers the three positions a strand actually takes, and the offset each implies is derived from the duct you declared and the strand you selected, and shown on the option before you pick it:

Choose Manual to enter your duct supplier’s own figure — they publish it as the centroid distance, and RAPT calls it the same thing. Manual also takes a value the presets cannot derive: for a round duct carrying more than one strand the offset is a property of the whole strand bundle piled on the duct invert, not of one strand, so the derived figure comes from SRG Global’s published tendon eccentricities rather than the strand diameter alone. If your supplier tabulates a different one, use theirs.

For a flat duct the derivation is exact, because a flat duct holds a single row of strands: half the clear height, less half the strand diameter. A 15.2 mm strand in a 20 mm flat steel duct seats 2.0 mm off the centreline; a 12.7 mm strand in the same duct, 3.25 mm.

ACS cannot work the sign out for you — Top and Bottom are your call, not a default. It is a section tool: it sees one cross-section, not the tendon’s profile along the member, so it has no curvature to read a sag or a hog from, and the same section is checked against sagging and hogging load combinations alike. The control therefore starts at Not declared, which is treated as undeclared rather than as zero: ACS models the strand on the centreline and says so on the PT tab. Choosing Duct centreline is a different statement — it declares the strand centred, and raises no note.

If the duct you declared cannot seat its strands off the centreline at all, Top and Bottom are unavailable and the control says why rather than offering a number it cannot derive. An offset that would place the strand outside its duct is rejected rather than clamped, which is what catches a misplaced decimal point.

:::note[Strand position in duct extended to custom tendons — September 2026 (#5824)] The Strand position in duct selector is now available on both tendon kinds — edge-relative and custom. Before September 2026, the control appeared in the edge-relative tendon dialog only; a section built from absolute-coordinate custom tendons could not declare the strand seat, and ACS would disclose the undeclared basis on the PT tab with no way to resolve it. The control now appears in the custom tendon dialog with the same Top of duct / Duct centreline / Bottom of duct / Manual options. :::

:::caution[Results change — edge-relative tendons before 2026-08-11] Prior to PR #4499, when a tendon’s Y position was declared relative to an outline edge rather than from the section origin, the declared strand centroid offset was stored but not applied when resolving the tendon’s effective depth for capacity calculations — the engine used the duct centreline coordinate instead. For a 450 mm deep member with a 19 mm duct the error in $\phi M_u$ is approximately 0.6 %; it is proportionally larger in shallow slabs where the duct diameter is a greater fraction of the total depth. If any tendon in your section uses edge-relative Y positioning and declares a non-zero strand centroid offset, re-run the analysis. :::

Strand library

ACS provides standard 7-wire low-relaxation strand types, and one hot-rolled bar:

The yield stress $f_{py}$ is deliberately not a column here: it is not a property of the strand alone. It depends on the design code you are working to, and on what the product’s own certificate declares — see Strand yield stress below.

The 26 mm bar

Every other row above is 7-wire strand. The bar is the one AS/NZS 4672.1:2007 product in the library, and every value it declares is read from that standard’s Table 6.1 (hot-rolled bars), 26 mm row of the common bar sizes block: nominal diameter 26 mm, nominal cross-sectional area 562 mm², nominal tensile strength 1030 MPa, characteristic minimum breaking force 579 kN and a 0.1 % proof force of 469 kN. Its $f_{p0.1k}$ is that proof force over that area — 834.5 MPa — so the AS 3600 bar fallback below does not apply to it. Its $E_p$ of 200,000 MPa comes from Cl. 6.4.5, not from the 195,000 the strand products carry, and its $A_{gt}$ of 0.06 from Cl. 6.4.2.

It is tagged AS 3600 only, unlike the strand products. AS/NZS 4672.1 is the Australian product standard AS 3600 Cl. 3.3.1(b) defers to by name; nothing establishes that this particular bar is what an ACI 318 or EN 1992-1-1 job would specify, and offering it under those codes would be a claim ACS cannot support.

The bar is also the only library product that takes the 10 mm arm of the AS 3600 Cl. 5.3.3 fire axis-distance increase — every strand takes 15 mm. See the Section 5 fire assessment.

:::note[The EN 1992 crack-width check refuses for a bar tendon] EN 1992-1-1 writes its bonded-tendon bond machinery for strand. Cl. 6.8.2 states the equivalent tendon diameter $\phi_p$ only for single 7-wire and 3-wire strands, and Table 6.2’s bond ratio $\xi$ keys on the steel’s surface — 0.3 for smooth bars and wires against 0.7 for ribbed bars, bonded post-tensioned at $\leq$ C50/60. AS/NZS 4672.1 Cl. 6.3 permits a hot-rolled bar to be plain, ribbed or threaded, and the catalogue does not record which, so neither quantity is determined.

Both feed Cl. 7.3.2(3) Eq. (7.5) and Cl. 7.3.4, so a section whose crack control is provided by a bar tendon reports why the check cannot run rather than returning a width computed from a strand’s geometry. Every other EN 1992 check, and every AS 3600 check including fire, is unaffected. :::

Strand configuration

Every strand in the library above is 7-wire ordinary construction — one centre wire with six laid around it. ACS records that configuration on the strand product rather than assuming it, so the design checks that depend on the wire count read it from the product you selected. The 26 mm bar declares no strand configuration, because it is not a strand and there is no wire count to record.

This matters for one check in particular. The EN 1992-1-1 equivalent tendon diameter $\phi_p$ — used in crack control — is defined per wire count: $\phi_p = 1.75\,\phi_{wire}$ for 7-wire strand against $1.20\,\phi_{wire}$ for 3-wire (Cl. 6.8.2). It is a different quantity from the strand/wire/bar product form that drives the fire axis-distance increase of AS 3600 Cl. 5.3.3, even though both describe “what kind of tendon this is”.

Where a standard establishes no value for a product, ACS refuses the affected check rather than substituting a similar one. Compacted 7-wire strand is the case that arises: it is drawn through a die after stranding, which deforms the wires out of round, so Cl. 6.8.2’s “wire diameter” is undefined for it and crack control reports why instead of returning an approximate crack width.

Strand yield stress $f_{py}$

Both supported design codes define the tendon yield stress the same way: it is the characteristic 0.1% proof stress $f_{p0.1k}$, and the familiar fraction-of-$f_{pu}$ formula is only the fallback each code prescribes when that value has not been established.

• AS 3600:2018 Cl. 3.3.1(b) — $f_{py}$ “shall be taken either as the 0.1% proof stress as specified in AS/NZS 4672.1, or determined by test data”.
• EN 1992-1-1:2004 Cl. 3.3.6(1)P — analysis is performed on the characteristic values $f_{p0.1k}$, $f_{pk}$ and $\varepsilon_{uk}$, with the design value $f_{pd} = f_{p0.1k}/\gamma_S$ (Cl. 3.3.6(6)).

So where the strand product you select declares an $f_{p0.1k}$, ACS uses it directly under every design code.

The 1860 grades declare one

The three 1860 MPa strands in the library carry a declared $f_{p0.1k}$ of 1636.8 MPa. prEN 10138-3 (White Draft, May 2006) Table 4 note (c) specifies the characteristic 0.1% proof force as $F_{p0.1} = 0.88\,F_m$, and $f_{p0.1k} = F_{p0.1}/S_n$ (EN 1992-1-1 Cl. 3.3.1(5) Note), so with $F_m = R_m S_n$ the area cancels and $f_{p0.1k} = 0.88\,f_{pk} = 0.88 \times 1860$.

Because that is a declared value, $f_{py}$ for these products is the same under every design code — no fallback factor is reached. That is the point of the ordering: the proof stress is a property of the steel, not of the code you are checking against.

:::caution[This source is a draft, not a published standard] EN 10138 was never published. The series reached three successive drafts and was formally abandoned by CEN in April 2021, so the value above is cited to prEN 10138-3 White Draft May 2006 and nothing more. The April 2005 draft gives $0.86$ rather than $0.88$; the May 2006 draft is the later of the two. If your project requires the proof stress to be traceable to a published standard or to the product’s own certificate, treat 1636.8 MPa as a default to confirm rather than a certified value — and note that CEN’s restart renumbers the parts, so a future “EN 10138-3” will cover bars, with strand moving to Part 2.

Two of the three products this value is applied to are not prEN 10138-3 products. An August 2026 review of the catalogue found that only the 15.7 mm strand matches a row in that standard’s 7-wire table. The 15.2 mm strand is an ASTM A416 Grade 270 product (that standard’s Table 1 gives 261 kN over 140 mm², matching exactly, where prEN 10138-3 gives 139 mm² and AS/NZS 4672.1 gives 143 mm²), and ASTM A416 declares no 0.1 % proof stress at all — its yield is specified at 1.0 % extension, a different measurement. The 12.7 mm entry matches no product table: it carries AS/NZS 4672.1’s area with ASTM’s strength class. Their 1636.8 MPa is therefore a value taken from a standard that does not define them. It is retained unchanged so that no saved design silently shifts, and it is higher than the value AS 3600 would use in the absence of test data — $0.82 f_{pb} = 1525.2$ MPa, so 1636.8 MPa is 7.3% above it. If you are designing to AS 3600 with 12.7 mm or 15.2 mm strand, confirm $f_{p0.1k}$ against your supplier’s certificate before relying on the catalogue value. :::

Stress at 1 % extension ($\sigma_{p1.0}$) — a different measurement

The 15.2 mm strand additionally declares a stress at 1.0 % extension of 1674 MPa, from ASTM A416/A416M − 16 §6.3: “Yield strength in pounds [kN] shall be measured at 1.0 % extension under load. The minimum yield strength shall be 90 % of the breaking strength listed in Table 1.” So $\sigma_{p1.0} = 0.90 \times 1860$.

This is not a proof stress and is not interchangeable with $f_{p0.1k}$ above: $f_{p0.1k}$ is measured at a 0.1 % plastic offset, $\sigma_{p1.0}$ at 1 % total strain. They describe different points on the same curve. The other four strands declare no $\sigma_{p1.0}$ and ACS discloses the value it uses for them as a platform default rather than a standard’s figure.

The same row declares a total elongation under load of 3.5 % (§6.4), which AS/NZS 4672.1 Cl. 5.4.1.2 independently requires as $A_{gt} \geq 3.5\%$ for this class of product.

$\sigma_{p1.0}$ and $A_{gt}$ are the two parameters the tendon stress-strain curve is built from, so a strand that declares them is analysed on its own curve rather than the platform’s. The PDF report’s Prestressing Strand table states both, with the strain-hardening ratio $Q$, and marks each Declared or Platform default — the same basis wording the panel shows, so the two can be read side by side.

:::caution[Results differ for the 15.2 mm ASTM A416 strand before 2026-08-17] A strand’s declared $\sigma_{p1.0}$ and $A_{gt}$ reached the material panel but not the analysis: every tendon was evaluated on the platform-default curve ($0.875\,f_{pb}$ and $A_{gt} = 5\%$) whatever its product row declared, while the panel reported the declared values as Declared. The 15.2 mm ASTM A416 strand is the only library product affected — its $\sigma_{p1.0}$ was evaluated at 1627.5 MPa instead of the declared 1674 (2.9 % low) and its $A_{gt}$ at 5 % instead of the declared 3.5 %. Tendon stress at a working strain of 0.008 moves by about 2 %; the direction depends on the section. Corrected on 2026-08-17, and the report now states the curve so the two surfaces can be compared. If you have previously computed results for a section using the 15.2 mm ASTM A416 strand, re-run the analysis and compare. No other library product declares these values, so no other section changes. :::

When the product declares no proof stress

The two 1720 MPa strands carry no declared $f_{p0.1k}$ — that strength class appears neither in prEN 10138-3 nor in AS/NZS 4672.1 (reproduced as AS 3600:2018 Table 3.3.1, which lists 1850, 1870, 1790, 1830 and 1860 for 7-wire ordinary strand), so ACS has no source to take a value from and records the property as not established rather than inventing one. These products take their code’s own fallback. These differ between codes, and ACS applies the one belonging to the code you are designing to — it does not apply a single factor everywhere.

For 1720 MPa strand that is $f_{py} = 1410$ MPa under AS 3600 and $1548$ MPa under EN 1992-1-1.

Note how far apart those are — 9.8% on the same physical steel. It is worth reading the gap the other way as well: measured against prEN 10138-3’s $0.88$, AS 3600’s $0.82$ strand tier is 6.8% conservative, while EN 1992-1-1’s recommended $0.90$ is 2.3% unconservative. A fallback is a value the standard authorises in the absence of test data, not an estimate of the product — which is why a declared $f_{p0.1k}$ takes precedence over both.

AS 3600 gives two tiers for wire (as-drawn $0.80$, stress-relieved $0.83$) and two for bar (hot-rolled super grade $0.81$, hot-rolled ribbed $0.89$). ACS records the product form — strand, wire or bar — but not which of the two sub-types a product is, so it applies the lower tier of each pair. That is the conservative reading, because $f_{py}$ caps the tendon stress at ultimate. A product that declares its own $f_{p0.1k}$ bypasses the ambiguity entirely.

Effect on capacity

The yield stress caps the tendon stress at ultimate ($f_{ps} \leq f_{py}$), which is conservative for strain-hardening strand where $f_{py}$ lies below the actual 0.1% proof stress. It also sets the pure-tension end of the interaction diagram, and it is the quantity the analysis refuses on when a tendon’s $f_{py}$ reaches its own stress at 1% strain — so it reaches ultimate capacity, service stresses and the fire assessment alike. It does not set the knee of the tendon stress—strain curve: that is $\sigma_{p1.0}$, the stress at 1% total strain, which the platform resolves from the grade’s declared value or a disclosed default.

:::caution[Results may differ for non-1860 MPa strand before 2026-08-01] Eight analysis endpoints — covering ULS interaction, service stresses, losses, fire capacity, moment-curvature, and their MCP equivalents — were missing the $f_{pu}$ and $f_{py}$ fields in the request payload. Any tendon configured with a non-1860 MPa strand grade (for example, 1720 MPa) was silently evaluated at the 1860 MPa default, producing unconservative capacity results and incorrect tendon stress-strain curves. All eight endpoints were corrected on 2026-08-01. If you have previously computed results for a section with non-1860 MPa strand, re-run the analysis and compare. :::

:::caution[Post-tensioned analysis refused when a tendon has no strand material (#4934)] Every tendon defined in the section must have a strand material selected before any post-tensioned concrete analysis will run. Previously, a tendon with no strand material caused the engine to substitute a default silently; §874 requires an explicit refusal instead.

When a tendon carries no material, the analysis is refused with a message identifying the tendon. To resolve it, open the tendon dialog and choose a grade from the strand library, or configure a custom material. Sections analysed without an explicit strand selection before this change may have been evaluated under the 1860 MPa default regardless of the intended grade. Re-running the analysis after selecting the correct material is recommended.

This applies only when a tendon is present. A purely reinforced section with no tendons is unaffected. :::

:::caution[Post-tensioned analysis refused when a tendon material resolves to a synthetic fallback (#5808)] When a tendon’s assigned material cannot be resolved from the catalogue — for example, after a grade is retired or a custom material is deleted — the analysis now refuses rather than substituting a synthetic fallback material.

Previously, an unresolvable tendon material caused the engine to construct a synthetic stand-in and continue silently, which produced results under undisclosed assumptions (§874). The analysis now surfaces an explicit error identifying the tendon whose material could not be resolved.

If you see this error:

1. Open the tendon dialog for the identified tendon.
2. Select a current strand grade from the library, or configure a replacement custom material.
3. Re-run the analysis.

A section that was previously analysed without error under a synthetic fallback may have been evaluated with substituted material properties. Re-running after assigning the correct grade is recommended.

This is distinct from the #4934 case (no material selected at all). Both are now refused: the earlier case catches a tendon with an empty material slot; this case catches a tendon whose material slot references a material that can no longer be loaded. :::

Concrete strength at transfer ($f'_{ci}$)

Set the concrete strength at transfer in the f’ci field below the tendon table. This is the compressive strength of the concrete at the time the prestress is applied — typically 0.75—0.85 $f'_c$ for pretensioned members and may be lower for early-age transfer.

Prestress losses

Prestress losses reduce the initial jacking stress $f_{pi}$ to the effective prestress $f_{pe}$ over time. ACS computes each loss component individually and displays them in a breakdown table.

Immediate losses

:::caution[Friction and draw-in only run when you supply their inputs] Both are properties of the tendon along its length, and ACS analyses a single cross-section, so it cannot derive them from the section alone --- you supply them.

Section-level, in the PT panel. The default anchorage distance is the switch: with it blank, neither loss is computed. It is the distance from a stressing anchorage to the section that any tendon declaring none of its own falls back to --- for a member stressed from one end only, with the section taken at the far anchorage, that is its full length. It feeds nothing else: ACS analyses a cross-section and has no slenderness or buckling check, so this value never reaches the interaction diagram. Beside it sit the friction coefficient $\mu$, the wobble coefficient $\beta_p$ and the anchorage draw-in, which fall back to the AS 3600 Cl. 3.4.2.4 advisory values shown in each box. Those advisory values are not uniformly conservative --- $\mu = 0.20$ is the high (flat-duct) end of its range while $\beta_p = 0.008$ rad/m is the low end of 0.008—0.024 --- so review them against your duct system. Whenever a value is assumed rather than entered, the results say so.

Per tendon, in the tendon dialog. Each tendon carries its own angle change $\alpha$ and its own anchorage distance. Friction accumulates from that tendon’s stressing anchorage along that duct, so two tendons on one section genuinely differ whenever their drapes differ, their lengths differ, or they are jacked from opposite ends --- alternate-end stressing, which is standard practice on long post-tensioned floors. Leave the anchorage distance blank and the section-level default anchorage distance is used, which is the single-end-stressed case.

A blank angle change is not a straight duct. With $\alpha$ undeclared, the Friction card reads --- not modelled for that tendon and its $f_{pe}$ is an upper bound. A genuinely straight duct is entered as $\alpha = 0$, which still charges the wobble term. The two are different statements and ACS reports them differently.

If you leave the default anchorage distance blank. The effective prestress $f_{pe}$ ACS reports is an upper bound for a post-tensioned member, and so is everything derived from it --- the cracking moment, the service stresses and the deflection. On a 24 m simply-supported flat-duct beam ($\mu = 0.20$, $\beta_p = 0.016$ rad/m, 6 mm anchorage set) the two omitted losses account for about 9.6% of $f_{pe}$, and 13—17% at 30—36 m. The Friction and Draw-In cards read --- not modelled rather than 0.0, and the panel raises a disclosure beside them.

Ultimate flexural capacity $\phi M_u$ is not materially affected either way: a bonded strand at the ultimate limit state sits past its proportional limit, so $f_{pe}$ barely moves it.

Pretensioned members are unaffected --- they have no duct and no anchorage set, so neither loss applies, and ACS reports a genuine 0 for both once a default anchorage distance is present.

A bonded tendon with no duct diameter cannot be evaluated for either loss. If you supply a default anchorage distance while such a tendon is present, ACS refuses the calculation and names the tendon rather than returning a zero for it.

Prior to the August 2026 release, when the anchorage distance was zero or draw-in input was absent, the engine emitted 0 as a computed loss value rather than null, making the result indistinguishable from a real zero loss (#4198). If a previous result showed 0.0 for friction or draw-in on a post-tensioned member, treat it as not modelled, not as a confirmed zero loss. :::

Friction loss formula

The friction loss along the duct depends on the design code:

$\sigma_{pa} = \sigma_{pj} \cdot e^{-\mu(\alpha + \beta_p \cdot x)} \quad \text{(AS 3600:2018 Cl. 3.4.2.4 and EN 1992-1-1 Cl. 5.10.5.2)}$

$P_{px} = P_{pj} \cdot e^{-(\mu\alpha + K \cdot x)} \quad \text{(ACI 318-19 Cl. 20.3.2.6)}$

Where:

• $\sigma_{pj}$ = jacking stress (MPa)
• $\mu$ = friction coefficient between tendon and duct (default 0.20 for flat duct, per AS 3600 Table 3.4.2.4)
• $\alpha$ = cumulative angular change from the jack to the section (rad)
• $\beta_p$ (AS/EN) / $K$ (ACI) = wobble coefficient per unit length of duct (default $8 \times 10^{-6}$ rad/mm = 0.008 rad/m, the AS 3600 lower bound)
• $x$ = length of duct from the jack to the section (mm)

AS 3600 and EN 1992-1-1 treat wobble inside the friction coefficient: the exponent is $\mu(\alpha + \beta_p x)$. ACI 318 treats wobble as a separate additive term: the exponent is $\mu\alpha + Kx$. The two forms give materially different answers, so use the one belonging to the code you are designing to.

Long-term losses

Relaxation class

The relaxation loss depends on the class of the prestressing steel. EN 1992-1-1 Cl. 3.3.2(4)P defines three, each with its own $\rho_{1000}$ — the relaxation measured at 1000 hours and 20 °C, as a percentage of the initial stress (Cl. 3.3.2(6)) — and its own design expression:

$\rho_{1000}$ is not itself the design loss: it is the value at 1000 hours, and relaxation keeps growing. ACS charges the Cl. 3.3.2(7) expression for the class, evaluated at the Cl. 3.3.2(8) long-term reference of $t = 500{,}000$ hours (about 57 years):

where $\mu = \sigma_{pi} / f_{pk}$ is the initial prestress as a fraction of the characteristic tensile strength, $t$ is the time after tensioning in hours, and $(k_1, k_2)$ is $(5.39, 6.7)$, $(0.66, 9.1)$ or $(1.98, 8.0)$ for Class 1, 2 or 3. At $\mu = 0.7$ this gives 19.0%, 3.9% and 8.7% of the initial prestress respectively — roughly 1.6 to 2.4 times $\rho_{1000}$ itself.

The time-dependent (AEMM) analysis uses the same expression at each analysis age, so the relaxation it reports at 50 years and the long-term loss reported here agree.

The class comes from the tendon grade you select. The 26 mm bar declares Class 3, from AS/NZS 4672.1 Cl. 6.4.3 — relaxation at 1000 h from an initial force of 70 % of the characteristic minimum breaking force “shall be not greater than 4.0 % for all bars”, which is the same measurement and the same figure EN 1992-1-1 gives Class 3. Every strand grade in the standard catalogue declares Class 2. For the 15.7 mm strand — an EN 10138-3 product — that follows from the standard: Cl. 3.3.2(4)P notes that Class 1 is not covered by EN 10138, and Class 3 is bar rather than strand. For the 15.2 mm strand it follows from ASTM A416 §6.5, which caps relaxation at 2.5 % when loaded to 70 % of the specified minimum breaking strength after 1000 hours — the same figure EN 1992-1-1 uses for Class 2. The remaining three grades were relabelled in August 2026 once it emerged that they describe no product any of those standards contains, so for those the Class 2 declaration is the platform’s, not a standard’s. It is unchanged from what every analysis has always used, and it remains the low-relaxation value appropriate to 7-wire strand — but if your product’s certificate states a class, enter it rather than relying on the catalogue.

If the selected grade does not declare a class, ACS assumes Class 2 and says so in the loss results rather than assuming it silently — the difference matters, because Class 1 steel relaxes more than three times as much and an unstated assumption in that direction over-states the effective prestress.

Amber disclosure card

When the relaxation class for a tendon cannot be read from the selected strand grade, ACS displays an amber disclosure card in the PT panel beneath the tendon table. The card states the assumed class (Class 2) and explains why the grade does not declare one. This appears before you run the analysis so the assumption is visible at input time, not only in the output.

You will see this card if you add a custom strand type whose product definition does not include a relaxation class field. Every standard catalogue grade declares Class 2 and so does not trigger the card.

:::note[Results change — analyses before July 2026] Prior to the July 2026 release, ACS did not read the relaxation class from the selected strand grade: all tendons used Class 2 ($\rho_{1000} = 2.5\%$) regardless of the steel specified. If you ran a prestress loss analysis before this update, the relaxation component was computed with Class 2 even when a different class would now apply. Re-run any affected analyses to confirm the result is unchanged — for the standard low-relaxation strand catalogue this makes no difference, but Class 1 or custom grades may see a change in the relaxation loss term. :::

:::caution[Results change — analyses before August 2026] Prior to the August 2026 release, ACS charged $\rho_{1000}$ itself as the relaxation loss, rather than the Cl. 3.3.2(7) expression at the long-term reference time. The relaxation component was under-stated on every class — by about 1.6× for Class 2 low-relaxation strand, and up to 2.4× for Class 1 — which over-stated the effective prestress $f_{pe}$ and every service stress, cracking moment and deflection derived from it. A stress-ratio cut-off that returned zero relaxation for lightly stressed tendons has also been removed; it was not stated by either standard, and Cl. 3.3.2(7) charges a small loss there instead of none. Re-run any prestress loss or time-dependent analysis saved before this update — the results will change. :::

Loss results

The loss panel reports:

:::note[Per-tendon jacking_stress (σ_pj) field in loss response — August 2026 (#5187)] The API and MCP loss response now includes a jacking_stress field (in MPa) in each object in the tendons[] array. This is the jacking stress $\sigma_{pj}$ at the jack — the stress immediately before seating losses are applied — derived from the tendon’s jacking force and cross-sectional area. It differs from f_pi (the stress at the active end immediately after anchorage draw-in) by the draw-in loss component.

Previously the response exposed f_pi and the loss components per tendon but not the upstream $\sigma_{pj}$ — callers who needed to verify the jacking force against a strand-package limit had to back-calculate it from the anchorage area. The field is null when jacking stress cannot be determined (for example, when no jacking force is stored on the tendon). :::

:::note[PT-loss derived read available on sections with no load combinations — September 2026 (#6175)] The PT-losses by-id derived read now returns correct results when the section has no load combinations defined. Previously the endpoint failed or returned incorrect data in this state. Prestress losses are computed from tendon geometry and material properties; load combinations are used by ULS and SLS checks but are not an input to the loss calculation itself. You can run the losses analysis at any point in the design process — before defining load combinations. :::

Typical total losses range from 15—25% of the initial jacking stress, depending on the member geometry, concrete properties, and loading history.

PT-aware time-dependent analysis (AEMM)

When prestressing tendons are defined, the time-dependent effects panel runs the Age-Adjusted Effective Modulus Method (AEMM) with the at-transfer prestress as the initial concrete stress state. The AEMM then computes creep, shrinkage, and relaxation increments that develop on top of that initial state — the same three effects that drive long-term prestress losses.

Bonded tendons are included in the transformed cross-section at both the short-term modular ratio $n_p = E_p / E_c$ and the long-term ratio $n_{p,\infty} = E_p / E_{c,\text{eff}}$. Unbonded tendons contribute the prestress force only (no local stiffness, since the tendon is free to slide within the duct).

Tendon stress-loss output

The Effective Stiffness results table gains a Δσ_p (MPa) column for PT sections, reporting the time-dependent tendon stress loss at each analysis age. This column is hidden for reinforced sections and appears automatically when any tendon is present.

Warnings

An amber warning banner appears when:

• Unbonded tendon approximation — the AEMM seeds the initial stress state from an area-weighted average effective prestress across all unbonded tendons. This approximation is stated explicitly so you can judge whether it is acceptable for your geometry.
• High at-transfer stress ($\sigma_{ci} > 0.5 f'_c$) — the AS 3600 creep model is calibrated for stresses up to half the 28-day characteristic strength; results above this level should be treated as indicative.
• Creep coefficient or shrinkage strain estimated — no override was supplied, so that value came from the design code’s creep or shrinkage model rather than from you. Set the override to use your own value; an explicit 0 supplied through the API or MCP tool is honoured as zero.

Transfer and service stresses

ACS checks concrete stresses at two critical stages:

At transfer

Immediately after the prestress force is applied. At this point:

• Concrete has only reached its transfer strength $f'_{ci}$
• A coexisting moment $M_{sw}$ acts alongside the eccentric prestress
• Both top and bottom fibre stresses must be within allowable limits

Entering the transfer-stage moment. $M_{sw}$ is an input on the PT panel, not something ACS derives. Deriving it from a span would require a load pattern, a support condition and a critical-section choice that belong to a member-level tool, so ACS receives the moment the same way it receives the service-stage $M^*_x$ and $M^*_y$. It is predominantly self weight, but construction and handling load act on the same term — precast lifting is the canonical reason the transfer check exists.

The value is signed, on the platform convention: positive = sagging (compression at the top face), negative = hogging. The sign is load-bearing, because leaving the moment out is only conservative where it opposes the prestress moment:

The two unconservative rows are not contrived: a precast member with a low straight strand pattern lifted at inboard points hogs under its own weight while the strand stays low, and the same holds near an inflexion point on a continuous member.

Leaving the field blank means not supplied, and both the panel’s working table and the report label the stage as not modelled rather than printing a bare zero. Entering 0 is a different statement — that the section genuinely carries no coexisting moment, as a PT column under pure axial prestress does — and is labelled as such.

Where:

• $P_i$ = initial prestress force (after immediate losses)
• $e$ = eccentricity of the prestress force from the centroid, positive below the centroid
• $A$, $I$ = section area and second moment of area, on the basis described below
• $y_t$, $y_b$ = distances from centroid to top and bottom fibres
• $\gamma_P$, $\gamma_G$ = the load factors of the governing transfer combination, below

The transfer stresses are design actions, not service actions

Cl. 8.1.6.2 is a strength check, and its opening sentence says so: “The strength of a prestressed beam at transfer shall be checked using the load combinations specified in Clause 2.5.2.2 and a strength reduction factor ($\phi$) for the section of 0.6.” The $0.50 f'_{cp}$ / $0.60 f'_{cp}$ stress limits that follow are the deemed-to-satisfy branch of that check, and they are written against “the maximum compressive stress in the concrete, under the design loads at transfer”. “Design load” is a factored quantity in AS 3600’s own vocabulary — Section 1.6.2 defines $F_d$ as the “uniformly distributed design load, factored for strength or serviceability, as appropriate.”

Cl. 2.5.2.2 (as amended by Amendment 2:2021) supplies the combinations:

The clause requires “the more severe of the following”, so ACS evaluates both in full and reports the worse. Neither governs universally: (a) is a uniform 15 % scaling, while (b) removes 10 % of the relief the permanent action was providing at the soffit while still scaling the prestress up. On a 300 × 600 beam with $P = 1000$ kN at $e = 200$ mm and $M_{sw} = +80$ kN·m, (b) gives $\sigma_{bot} = 15.17$ MPa against (a)‘s 14.06 and a service-level 12.22 — so the service-level reading understates the design value by 24 %, and passes a 12.5 MPa allowable the clause fails.

The panel and the report both name the governing combination and its two factors beside the stresses, and the $P$ and $M$ shown in the working table are the factored values the terms were built from — divide by the stated factor to recover what you entered. The eccentricity $e$ carries no factor: it is geometry, and a load combination scales the force on the lever, never the lever itself.

This applies to AS 3600 only. ACI 318-19 Cl. 24.5.3 and EN 1992-1-1 Cl. 5.10.2.2(5) are permissible-stress clauses written against the actions immediately after transfer, so those arms are evaluated unfactored. The service stage is unfactored under every code — Cl. 2.5.2.2’s first sentence includes prestress in service combinations “with a load factor of unity”.

The whole of $M_{sw}$ is treated as $G$. Cl. 2.5.2.2’s transfer combination names only $G$ and $P$ — there is no imposed-action term at transfer — so the clause offers no separate bracket for construction or handling load that you choose to carry on the same input.

Section-property basis — gross or net. For a post-tensioned member the duct is a hole in the concrete, and whether it is deducted changes the answer materially. ACS applies the conventional per-stage rule:

The rule is applied per duct, so a section mixing bonded and unbonded tendons deducts all of them at transfer and only the unbonded ones in service.

The direction matters, because deducting a duct is not a small symmetric correction. A duct follows the tendon, so it sits low in the section at midspan; removing low-lying material raises the centroid, which lengthens the prestress lever $e$ at the same time as $A$ and $I$ shrink. All three effects push the same way. On a 300 × 1600 girder web with three Ø100 ducts — 4.9 % of the gross area — the soffit stress at transfer rises by 16.9 %, which on a $0.5 f'_{cp}$ limit is about a third of the margin.

Both the panel’s working table and the report name the applied basis beside the $A$ and $I$ they used, together with how many ducts were deducted and their total area, so the calculation can be reproduced by hand. Sections with no ducts — reinforced or pretensioned — show no basis note: the gross and net sections are then the same section, not two choices.

Sign convention. Concrete stresses are reported with positive = compression, negative = tension — the same convention ACS applies everywhere, including the SLS stress checks and the stress map, and set out in full on the platform sign conventions page. Each stress card in the PT results panel names its own sense (“compression” or “tension”) beside the value, so the sign never has to be inferred. The allowable limits below are unsigned magnitudes: a fibre passes when $-\sigma_{t,\text{allow}} \leq \sigma \leq +\sigma_{c,\text{allow}}$.

:::note[Curvature axis labels in the stress map — corrected 2026-08-06] Prior to 2026-08-06, the M-$\kappa$ stress map viewer displayed $\kappa_x$ and $\kappa_y$ labels swapped relative to the platform sign convention. The curvature about the X axis ($\kappa_x$, positive for sagging) was shown as $\kappa_y$, and vice versa. The signs of each curvature were always correct — only the axis label was wrong. Both labels now match the platform convention. Saved designs are unaffected: the curvature values themselves were never swapped, only their display names. :::

:::caution[Results change — PT transfer analyses before August 2026] Prior to the August 2026 release, ACS evaluated the Cl. 8.1.6.2 transfer stress check against unfactored actions, using the service-stage load factors instead of the Cl. 2.5.2.2 factors ($1.15G + 1.15P$ or $0.9G + 1.15P$, whichever governs). The check was therefore under-stated: a design that passes the unfactored check can fail the factored one, as the worked example above demonstrates (12.22 MPa unfactored vs 15.17 MPa factored, against a 12.5 MPa allowable). Re-run any AS 3600 transfer stress analysis saved before this update — results will change where Cl. 2.5.2.2(b) governs and the coexisting $M_{sw}$ is relieving. ACI 318 and EN 1992 are evaluated unfactored by their own clauses and are not affected. Existing cached PT transfer analyses were automatically invalidated (CalculationVersions.ConcretePtStresses advanced from 2.0 to 2.1); the corrected result is computed on first open. :::

At service

Under long-term sustained loads with the effective prestress $P_e$ (after all losses):

Where $A$, $I$, and the extreme-fibre distances $y$ are properties of the transformed section — the composite section with each steel area replaced by an equivalent concrete area using the modular ratio $n = E_s/E_c$. Measuring $y$ from the transformed centroid (rather than the gross centroid) correctly accounts for the neutral-axis shift due to the tendon and reinforcement area. The eccentricity $e$ is the distance from the transformed centroid to the tendon resultant.

The effective tendon force $P_e$ and the eccentric prestress moment $P_e \cdot e$ are both included in every SLS stress check, using the post-loss $\sigma_{p.ef}$ from the losses panel. ACS evaluates this at all defined load combinations and identifies the governing one.

When multiple tendons are present the resultant eccentricity is force-weighted: $e = \Sigma(A_p \sigma_{p.ef} y) / \Sigma(A_p \sigma_{p.ef})$ (#4171). For sections with a uniform effective prestress the area- and force-weighted eccentricities are identical; for staged or varied-prestress arrays — where tendons carry different effective stresses — the force-weighted eccentricity correctly reflects the true prestress resultant, and PT stress results will differ from prior releases.

Allowable stress limits

Not every value above is mandated by the standard whose column it sits in. Where ACS applies a limit the standard does not itself codify — or where we have not been able to verify the attribution against the primary text — it is named as an import or as an unsourced platform value, never presented as a clause requirement. The annotations below cover the AS 3600 column and the two ACI tensile values.

1. AS 3600:2018 Cl. 8.1.6.2(b) — the general branch. Cl. 8.1.6.2(a) permits the more generous $0.60 f'_{cp}$ for a section that is rectangular in cross-section and where the stress distribution is triangular in shape; ACS applies branch (b) to every section, which is the conservative reading. (AS 3600 writes the transfer strength $f'_{cp}$; the ACS input field labels the same quantity $f'_{ci}$.)

2. Not assessed — AS 3600:2018 codifies no tensile limit at transfer. Cl. 8.1.6.2 is the transfer clause and it addresses compression only. Nor is this an omission in the standard: Cl. 8.1.6.2 is a strength check — “checked using the load combinations specified in Clause 2.5.2.2 and a strength reduction factor ($\phi$) for the section of 0.6” — whose compressive stress limits are a deemed-to-satisfy shortcut, not an allowable-stress envelope. Transfer-stage cracking is instead governed by Cl. 8.1.6.1 ($M_{uo} \geq 1.2 M_{cr}$) and by the Section 8.6 crack-control clauses under service loads. ACS previously applied $0.50\sqrt{f'_{ci}}$ here and described it as an imported convention; no source was found to import from — the coefficient appears nowhere in the standard — so it has been removed rather than re-cited. The transfer verdict now reports “compression OK — tension not assessed”, and the transfer tensile stress is still displayed for you to judge.

3. Not codified by AS 3600:2018 as a service stress limit. $0.45 f'_c$ is the Cl. 3.1.8.4 linear-creep threshold, imported from EN 1992-1-1 Cl. 7.2(3) by analogy under the quasi-permanent combination. The same import, with the same reasoning, governs the RC service stress check.

4. AS 3600:2018 Cl. 8.6.3 — codified, and conditional on your detailing. The clause deems flexural cracking controlled up to $0.25\sqrt{f'_c}$ outright; the more generous $0.6\sqrt{f'_c}$ of route (a) applies only where reinforcement or bonded tendons are provided near the tensile face at a centre-to-centre spacing not exceeding 300 mm. ACS now tests that precondition against the section you have drawn: bars and bonded tendons lying in the tension zone are collected, their centre-to-centre spacing is measured perpendicular to the strain gradient, and the $0.6\sqrt{f'_c}$ tier is applied only if at least one qualifying element is present and no gap exceeds 300 mm. Unbonded tendons carry no bond stress and do not count. Sections without that detailing are measured against $0.25\sqrt{f'_c}$ — 2.4× stricter. The PT crack-control panel described below tests the full Cl. 8.6.3 ladder, including routes (b) and (c), and reports which route governs.

5. Unsourced — an ACS platform value, not a citation. The two ACI tensile limits are the coefficients ACS applies, but we have not verified them against the primary text of ACI 318-19, which is not in our reference corpus. On review the clause numbers previously shown beside them appear to address compressive stress, and $0.5\sqrt{f'_c}$ does not match the Class U threshold under any reading available to us. Rather than publish an attribution we cannot stand behind — an engineer reasonably reads a cited clause as the standard’s requirement — the panel labels both as Heuristic and names no clause, and the calculation report states the same beside each limit rather than printing a bare number. The values themselves are unchanged from earlier releases. If you are designing to ACI 318, check these two limits against the standard yourself.

Biaxial bending

The stress field of an uncracked prestressed section is a plane, and ACS solves it in full:

with the two gradients $\alpha$ and $\beta$ found from the section’s full inertia tensor:

This matters more often than it looks. Three independent things tilt the neutral axis, and any one of them means there is no single “top fibre stress” — the stress varies along the top face:

1. A moment about the y-axis. $M^*_y$ stresses the left and right fibres, not the top and bottom.
2. A tendon group eccentric in $x$. A single-sided cable, a laterally draped one, or a group following a curved soffit applies a prestress moment $P \cdot e_x$ about the y-axis, whether or not any applied $M^*_y$ exists.
3. A section with a non-zero product of inertia $I_{xy}$ — an L-section, a single-bevelled beam, or any freeform outline without an axis of symmetry aligned to $x$ or $y$. Such a section bends biaxially even under a pure $M^*_x$.

Where the field is genuinely uniaxial (no $M^*_y$, tendons on the centroidal y-axis, $I_{xy} = 0$) the expression above reduces exactly to the classical $\sigma = P/A \mp P e y / I \pm M y / I$, and the panel reports the top and bottom fibre stresses as before.

Where it is not, ACS reports the section’s peak compression and peak tension instead — each with the corner it was read at — and says which of the three causes applies. It does not report a top- or bottom-fibre value, because on such a section neither is a single number. Because the stress field is a plane and the outline is a polygon, those extremes are located exactly: a plane attains its maximum and minimum on a polygon at vertices, so sweeping the corners is not a sample.

:::note[Signed moment components — results change in 2026-07] The biaxial stress solve uses the signed values of $M^*_x$ and $M^*_y$ as they enter the right-hand side of the inertia-tensor system above. Prior to 2026-07, the calculation used scalar magnitudes instead, which produced incorrect results for sections under moments of mixed sense — a sagging $M^*_x$ combined with a hogging $M^*_y$, or vice versa. Sections under single-axis bending are unaffected; sections with combined biaxial loading should be re-checked in the current version. :::

What the results panel shows

The results panel shows the fibre stresses at both stages with pass/fail status against the applicable limits — the top and bottom faces for a uniaxially-bent section, or the two peak points for a biaxially-bent one (see above). Each stage’s verdict card is accompanied by an expandable disclosure — click it to see the allowable limit, the actual demand, and the reasoning behind the pass or fail. For each stage it also reports:

• The governing comparison — the worse of the peak-compression and peak-tension comparisons, as a value/limit pair with its location (“top” / “bottom”, or a corner such as “upper-left”) and sense (“compression” / “tension”). A PT fibre stress sits inside $-\sigma_{t,\text{allow}} \leq \sigma \leq +\sigma_{c,\text{allow}}$ at two independent extremes, so “the utilisation” is undefined until a rule picks one; the pass/fail verdict remains the engine’s, and this ratio explains it rather than deciding it.
• A stress band placing each reading inside the admissible range. Where a stage has no tensile allowable, the tension side is hatched and labelled not assessed rather than drawn to a bound — nothing in the graphic asserts a limit the standard does not set.
• The term-by-term working behind each stress: $P/A$, $\mp P e_y y / I_{xx}$ and $\pm M^*_x y / I_{xx}$, plus $\mp P e_x x / I_{yy}$ and $\pm M^*_y x / I_{yy}$ where the y-axis participates, which sum exactly to the reported $\sigma$ — alongside the $P$, $e$, $A$, $I$, $y_t$, $y_b$ and $M$ they were substituted from. A stage given an $M^*_x$ but no $M^*_y$ says so rather than printing a bare zero, on the same footing as the transfer-moment disclosure: an unsupplied moment and a moment asserted to be zero produce identical stresses and mean opposite things.
• Which limit applied, and where it came from. Each allowable is labelled with the sense it governs ($\sigma_c$ / $\sigma_t$) and carries its provenance: the clause where one genuinely mandates the value, the source standard where it is an import by analogy, and Heuristic where it is unsourced. For AS 3600 at service this includes which Cl. 8.6.3 tier your section earned — $0.6\sqrt{f'_c}$ or the 2.4× stricter $0.25\sqrt{f'_c}$ — and, when it did not earn the permissive tier, what the detailing would have to be (see note 4).
• A substitution notice when no transfer strength was entered and $f'_{ci}$ was taken as $0.75 f'_c$. Because the transfer compressive allowable is $0.5 f'_{cp}$, that substituted input sets the number your section is measured against, so it is stated beside the limit rather than left implicit.

The same stage tables, verdicts, limits and disclosures appear in the Transfer / Service Stresses section of the generated calculation report, computed from the same inputs by the same kernel — the report cannot show a different governing face, a different allowable or a different verdict than the panel. That includes the biaxial branch: where the panel reports peak points rather than faces, so does the report.

PT cracking moment

The cracking moment for a prestressed section is higher than for an equivalent reinforced section because the applied moment must first overcome the prestress precompression before the tension fibre can crack. Per AS 3600:2018 Cl. 8.5.3.1:

Where:

• $Z_{tr}$ = transformed section modulus to the tension fibre (mm³), computed using the transformed gross section (concrete + modular-ratio contributions of reinforcement and tendons)
• $f'_{ct.f}$ = flexural tensile strength of concrete (MPa)
• $\sigma_{cs}$ = shrinkage-induced restraint tensile stress at the tension fibre (MPa, AS 3600 only; zero for EC2 and ACI 318)
• $N^* / A_g$ = axial precompression from the applied design axial force (MPa)
• $P / A_g$ = prestress axial decompression (MPa), where $P = \sigma_{p.ef} \sum A_{pt}$ is the total effective prestress force
• $P \cdot e$ = eccentric prestress decompression moment (kN·m), where $e$ is the eccentricity of the tendon resultant resolved toward the tension fibre

The physical interpretation: the prestress applies a compressive stress $(P/A_g + P \cdot e / Z_{tr})$ at the tension fibre. Before the section can crack, the applied moment must overcome this precompression and reach the flexural tensile strength $f'_{ct.f}$.

ACS computes $M_{cr,t}$ using the post-loss effective prestress $\sigma_{p.ef}$ — the same value used in the service stress and ultimate capacity checks. The cracking moment is reported in both the SLS results panel and as the cracking point on the moment-curvature curve; both show the same value for the same section.

Which section $Z_{tr}$ is taken over

$Z_{tr}$ is the modulus of the uncracked transformed section: the concrete section with each steel area replaced by an equivalent concrete area at the modular ratio, $(n-1)A_s$ for a bar and $(n_p-1)A_p$ for a tendon. The $-1$ removes the concrete the steel displaces, which is what makes the substitution exact for a reinforcing bar or a pretensioned strand — each displaces its own area and nothing more.

An unbonded post-tensioned tendon is the exception, on both counts, and ACS treats it accordingly:

The second row is the same rule the AEMM analysis already applies: an unbonded tendon is free to slide within its duct, so there is no strain compatibility between the strand and the concrete beside it at any one section. Its prestress acts on the section as an external force, not as a local stiffness, so it does not belong in the transformed section at all.

Both differences run the same way — each would stiffen the section, raising $Z_{tr}$ and therefore $M_{cr,t}$, which claims the member cracks later than it does. On a 300 × 1600 girder web with three Ø100 unbonded ducts (4.9 % of the gross area), $Z_{tr}$ at the soffit falls from 143.2 × 10⁶ mm³ to 112.5 × 10⁶ mm³ — a 27 % difference carried straight into $M_{cr,t}$, and from there into the deflection transition and the crack-width check.

The same transformed section is used for the section-level and the four directional cracking moments, for crack width, and for the linear-elastic moment-curvature and deflection branches, so these never disagree about one section.

How far the unbonded rule reaches

The rule above governs the uncracked transformed section, and the cracked-section solve now applies it too — see Unbonded tendons in the cracked section below. Between those two updates the SLS stress panel and the stress map carried a notice saying the cracked solve did not model an unbonded tendon; that notice has been removed along with the limitation it described.

The ultimate-strength interaction surface now applies $\sigma_{pu}$ too, matching the single-point flexure check. It takes the member span-to-depth ratio entered on the PT panel and selects the AS 3600 Cl. 8.1.8 (or ACI/Eurocode) branch from it — on the N—M curve, on every angle slice of the 3D surface, and at the pure-bending point alike. Until that landed the surface held an unbonded tendon at its effective prestress $\sigma_{p.ef}$, which understates capacity. Where a duct has been deducted, the report’s transformed section table names the basis in its heading. Reinforced, pretensioned and bonded post-tensioned sections are unaffected: there is no unbonded duct to deduct, and the gross and net sections are then the same section rather than two choices.

:::caution[Results change — unbonded PT analyses before August 2026] Prior to the August 2026 release, an unbonded tendon was included in the transformed section as though it were bonded, and its duct was not deducted. Cracking moments, crack widths, deflections and moment-curvature cracking points for sections with unbonded tendons were therefore unconservative — $M_{cr,t}$ over-stated by up to about 27 % on the geometry above. The same assumption applied to the cracked section, where it over-stated $I_{cr}$ by up to about 79 % on the same geometry, so cracked neutral-axis depths, the cracked branch of the moment-curvature curve and SLS deflections were affected as well. Re-run any unbonded post-tensioned analysis saved before this update. Reinforced, pretensioned and bonded post-tensioned sections are unchanged. :::

Cracked-elastic branch for prestressed sections

Past the cracking moment, the M-κ curve for a PT section uses the effective prestress in every step of the cracked-section solve. The precompression shifts the neutral axis toward the compression face, stiffening the cracked-elastic branch relative to an equivalent RC section with the same reinforcement layout. For reinforced sections (effectivePrestress = 0), the cracked-section solver skips the prestress block entirely — RC M-κ curves are bit-identical before and after this change.

Unbonded tendons in the cracked section

The rule that keeps an unbonded tendon out of the uncracked transformed section, above, applies unchanged once the section has cracked: there is no strain compatibility between a greased strand and the concrete beside it, so the tendon contributes no area to the cracked transformed section, and its prestress continues to act as an external force on that section. Two things follow, and ACS applies both:

• An unbonded tendon in the tension zone no longer contributes $n_p A_p$ of cracked-section stiffness. This is the larger of the two, because it acts at the far fibre — on the girder web above it is worth 79 % on $I_{cr}$ and 44 % on the reported neutral-axis depth.
• An unbonded duct above the neutral axis is a real hole in the compression block, and is deducted from it. Unlike the uncracked case this deduction depends on where the neutral axis lands, so it is resolved during the solve rather than applied once: a duct below the neutral axis falls in concrete that is already being ignored and correctly costs nothing.

Both differences again ran the same way — each over-stated $I_{cr}$, which under-predicts curvature and deflection past cracking and stiffens the Bischoff cracked-elastic branch. $I_{cr}$ and the cracked neutral-axis depth appear on the Properties tab (as the four directional cases) and in the report, and feed the cracked branch of the moment-curvature curve and the SLS deflection check.

Tension-stiffened variant. The optional Bischoff tension-stiffened curve does not yet apply this correction. Updating it in isolation would create a disagreement between the M-κ panel and the SLS deflection check, which consume the same Bischoff effective-stiffness core. Both will be corrected together under a dedicated issue.

:::caution[Results change — PT M-κ analyses before August 2026] Prior to the August 2026 release, the effective prestress was applied to the cracking-moment calculation but omitted from the cracked-section solves used to trace the M-κ curve past $M_{cr,t}$. This made the cracked-elastic branch approximately 2.5 times too flexible for PT sections; SLS curvature estimates and crack-width checks derived from the M-κ curve may have been affected. Re-run any prestressed concrete analysis saved before this update. Reinforced sections are unchanged. :::

:::caution[Results change — PT section properties, M-κ interaction, and SLS stress-field before August 2026] Prior to the August 2026 release, the section-properties, M-κ P-M-M interaction, and SLS stress-field helpers read the tendons’ persisted jacking stress $f_{pi}$ as though it were the post-loss effective prestress $\sigma_{p.ef}$. Because losses were never deducted, these three results overstated the prestress applied to the section — which is unconservative for tension-governed serviceability checks and invisible from the response (the numbers are internally consistent and physically plausible). The correction threads tendonStressIsPreLoss: true and derives $\sigma_{p.ef}$ from the persisted PtLossConfig through the same path ConcreteReportPipeline has always used, so these surfaces now agree with the PDF report by construction. Re-run any PT section analysed through section properties, M-κ interaction, or SLS stress field before this update. Ultimate capacity and prestress-loss results are unchanged. :::

PT crack control (AS 3600 Cl. 8.6.3)

For sections with prestressing tendons, the Crack Width panel implements the deemed-to-comply ladder of AS 3600:2018 Cl. 8.6.3 rather than the RC direct crack width calculation. The ladder tests four routes in order from least to most onerous; the section passes as soon as any route is satisfied.

Routes and criteria

The name of the satisfied route appears in the results panel alongside the utilisation ratio.

Table 8.6.3 limits (steel-stress increment route)

For the steel-stress increment route, ACS looks up the permissible $\Delta\sigma_{scr}$ from AS 3600:2018 Amendment 2:2021 Table 8.6.3 based on the governing crack-control element and the design crack width limit $w'_{\text{max}}$:

The decompression load fraction $\alpha_{\text{dec}}$ — the ratio of the decompression moment to the governing service moment — is computed automatically from the prestress geometry.

Results

When tendons are present, the Crack Width panel displays a Cl. 8.6.3 crack control design check card showing:

• Concrete tensile stress $\sigma_{ct}$ compared to $0.25\sqrt{f'_c}$ and $0.6\sqrt{f'_c}$
• Steel-stress increment $\Delta\sigma_{scr}$ and the Table 8.6.3 limit for the governing element
• Utilisation $\leq 1.0$ = OK, with the satisfied route name as the result label

PT crack control — EN 1992-1-1 (Cl. 7.3.1, Table 7.1N)

For sections designed to EN 1992-1-1, crack control for prestressed members follows Cl. 7.3.1(6) and the associated National Annex table. The approach differs from AS 3600’s deemed-to-comply ladder: EN 1992 specifies a maximum crack width $w_{k,\max}$ per exposure class and relies on the general RC crack width formula (Cl. 7.3.4) for sections that are permitted to crack, while decompression replaces the crack width limit for the most aggressive exposure classes.

Exposure class limits (Table 7.1N): Under the quasi-permanent load combination:

Decompression check: ACS verifies that every fibre in the critical zone (within 100 mm of the exposed face) remains in compression ($\sigma \leq 0$) under the quasi-permanent combination. When the decompression criterion is not met, the exposure class is violated — there is no fall-back to a crack width limit for XD/XS sections.

Tendon stress increment $\Delta\sigma_p$ (Cl. 7.3.4): For bonded post-tensioned sections remaining uncracked under the characteristic load combination, ACS additionally checks that the stress increment in the prestressing steel above the decompression state does not exceed the nationally determined limit (default 200 MPa for post-tensioned, 150 MPa for pre-tensioned). See Section analysis — EN 1992-1-1 crack control for prestressed members for full formulation.

Cracked regime scope: EN 1992-1-1 crack width calculation for cracked prestressed sections (where the quasi-permanent combination places the cross-section in tension) requires combining bonded tendon area with conventional bar area in the $s_{r,\max}$ expression (Cl. 7.3.4(3)). ACS currently implements this combined-area path for sections with bonded tendons plus supplementary reinforcement. Sections with unbonded tendons or complex duct configurations may produce conservative results pending refinement — these are noted in the results panel.

PT ultimate capacity

The ultimate moment capacity of a prestressed section accounts for the increase in tendon stress beyond the effective prestress as the section deforms toward failure. ACS uses the effective (post-loss) prestress $f_{pe}$ as the starting point for all ultimate capacity calculations — both in the live results panel and in generated PDF reports. The initial jacking stress $f_{pi}$ is never used as a substitute.

The capacity itself is reported on the ULS tab, as $M_{uo,x}$ / $\phi M_{uo,x}$ — the same pure-bending solve, prestress-aware, that a reinforced section uses. The PT tab reports the tendon state at that ultimate: $f_{ps}$, and — for bonded tendons — the peak tendon strain $\varepsilon_p$ against its limit. There is one capacity engine, so the two tabs cannot disagree.

$f_{ps}$ is an area-weighted mean across the tendons, and that is why the strain readout sits beside it rather than replacing it. A single over-strained strand is diluted by its neighbours in a mean — and reporting a maximum would not help either, because the constitutive curve is capped at $f_{pb}$, so a strand at its strain limit and one well inside it read within a couple of percent of each other. Stress stops discriminating exactly where discrimination is needed; strain does not. See Tendon strain at ultimate.

Where:

• $f_{pe}$ = effective prestress after all losses (not the jacking stress $f_{pi}$)
• $\Delta\varepsilon_{ps}$ = additional strain at the tendon level at ultimate
• $f_{pu}$ = ultimate tensile strength of the strand (1860 MPa for standard strands)

:::caution[Results may change — tendon constitutive model corrected (#4155)] Prior to the August 2026 release, the PT engine entered the tendon constitutive law using a linear model but exited via the PCI power curve — two different laws applied to the same quantity in a single solve. Tendon stress is now held consistently: linear entry uses a linear exit, power-curve entry uses a power-curve exit. Additionally, sections where the declared $\sigma_{p.ef}$ has no admissible strain on the power curve are now refused with a clear diagnostic rather than returning a physically meaningless result. Re-run any PT ultimate capacity or interaction analysis to confirm results are unchanged — sections with a conventional $\sigma_{p.ef}$ near or below $0.70 f_{pu}$ are unlikely to be affected. :::

When $f_{pe}$ is unavailable, the capacity is withheld — not estimated

If a section declares tendons but the effective prestress resolves to zero — no $f_{pe}$ was supplied, the loss calculation failed, or the computed losses consumed the whole jacking stress — ACS does not fall back to a reinforced-only number. It withholds the PT ultimate capacity and says so in the section warnings.

A bare reinforced capacity reported for a section that declares tendons would be inconsistent with the rest of the same analysis: the transformed section properties ($A_{tr}$, $I_{tr}$, $Z_{tr}$ and the transformed centroid) always count a tendon as transformed steel, because they depend only on the modular ratio $n_p = E_p/E_c$ and not on the prestress level. The ultimate, interaction, moment–curvature and stress-distribution surfaces, by contrast, are driven by the tendon prestrain and so carry no tendon at all at $f_{pe} = 0$. The two would not describe the same section, and the difference is not small — on a typical beam a single strand is worth a few percent on $I_{tr}$ but several hundred percent on $M_{uo}$ if it were admitted to the ultimate equilibrium as passive steel.

Rather than choose one reading and present it as the answer, the analysis declines to report a prestressed capacity. Supply the post-loss effective prestress, or remove the tendons to analyse the member as reinforced.

An over-prestressed $f_{pe}$ is rejected, not analysed

The same principle applies at the other end of the range. $f_{pe}$ is the stress remaining in the tendon after all losses, so it cannot exceed the stress the tendon was jacked to — and no supported design code permits jacking above $0.85 f_{pb}$ (AS 3600:2018 Cl. 17.3.4.6(b) for stress-relieved post-tensioned tendons, the most permissive of the three arms; ACI 318-19 and EN 1992-1-1 are both tighter). An $f_{pe}$ above that limit is rejected with an explanatory error rather than analysed.

The reason is not only that the value is inadmissible, but that the engine has no honest answer for it. The tendon prestrain is obtained by inverting the strand’s stress-strain curve, and that curve is capped at $f_{pb}$ — so a stress at or above the breaking strength corresponds to no strain at all, and even approaching it the implied prestrain becomes physically meaningless well before it becomes undefined (at $0.90 f_{pb}$ it is 2.1 %, at $0.95 f_{pb}$ it is 5.7 %, past the rupture strain of any real seven-wire strand).

If you see this error, check whether a pre-loss jacking stress has been entered where an effective one was expected — that is much the commonest cause.

One section, one $f_{pe}$ — and it says so when that costs you something

Every ambient, serviceability and fire engine analyses a section at a single effective prestress. In the editor that is exactly what you supplied: the tendon panel carries one $f_{pe}$ and writes it to every tendon, so nothing is lost.

The REST API and the MCP tools are different — their tendon arrays accept a per-tendon effectiveStress, which is the honest shape for a member whose ducts were stressed in stages or from alternate ends. When those values differ, they are reduced to one area-weighted value before the analysis runs:

The area weighting is not a convenience. It makes the total prestress force exact — $\bar{f}_{pe} \sum A_{p,i} = \sum A_{p,i} f_{pe,i}$ — so the only residual is in the lever arm, which is area-weighted rather than force-weighted. Where the tendons sit at one depth (a band beam, alternate-end stressed) the result is exact at any spread; where they differ in depth as well as stress, expect sub-percent effects on $M_{cr}$ and effectively none on $\phi M_{uo}$, whose tendon sits past its proportional limit at ultimate regardless.

Because a substitution you cannot see is worse than one you can, responses disclose it. When the supplied values differ, the response carries a prestressCollapse object naming the spread, the tendon count and the value actually analysed; on the analyze read, the batch check and the PDF report it appears instead as a section warning. A section whose tendons all carry the same $f_{pe}$ — every editor session — collapses nothing and carries no advisory.

The over-prestress check above is applied to $\bar{f}_{pe}$, the value the engines actually receive, for the same reason: the number that is validated must be the number that is used.

:::caution[Results may change for API and MCP callers only (#4191)] Prior to the August 2026 release, differing per-tendon effectiveStress values were not area-weighted consistently: most surfaces used the first tendon in the array and discarded the rest, /api/v1/concrete/fire-capacity used the largest, and only the former /api/v1/concrete/crack-width payload route area-weighted. Because nothing in that contract fixed tendon array order, the first-tendon rule meant that re-ordering an unchanged tendon array could change the answer — on a four-duct beam, by +10.9 % or −15.2 % on the bottom-fibre stress depending on the order sent. Re-run any API- or MCP-driven analysis that supplied differing per-tendon effective stresses. Editor sessions are unaffected — they have always sent one shared value — and cached results for them are unchanged. :::

Bonded tendons

For bonded tendons, strain compatibility applies — the tendon strain increases with the surrounding concrete strain, providing a significant stress increase at ultimate. The tendon stress at ultimate is computed from the full nonlinear strain profile up to the limiting concrete strain — which is the code-specified $\varepsilon_{cu}$ unless the tendon’s own strain limit governs first (below).

Tendon strain at ultimate

:::note[Tendon strain at ultimate checked against A_gt — September 2026 (#5864)] When a bonded tendon’s strain at ultimate would exceed its $A_{gt}$ — the strain at which the constitutive curve reaches $f_{pb}$ and stops being valid — ACS now constrains the extreme-fibre concrete strain $\varepsilon_{cu}$ so that no bonded tendon exceeds that limit. The constrained $\varepsilon_{cu}$ is disclosed on the PT tab and in the PDF report. Before September 2026, a strand strained past $A_{gt}$ returned exactly $f_{pb}$ with no indication that the strain was outside the constitutive curve’s declared validity domain — the section would appear to carry its full capacity while a conservative disclosure was owed (§874). The check applies to bonded tendons under all three design codes; unbonded tendons carry the member-average Cl. 8.1.8 $\sigma_{pu}$ and have no local strain to limit. :::

A bonded tendon’s strain at ultimate is checked against its strain at maximum force, $A_{gt}$ — the value declared on the strand grade, or the disclosed platform default of 5 %.

$A_{gt}$ is not an arbitrary allowable. It is the strain the tendon’s stress-strain curve is constructed to reach $f_{pb}$ at, so it is the upper bound of that curve’s validity: past it the engine is no longer computing a stress, it is returning the cap. A strand strained beyond $A_{gt}$ would therefore report exactly $f_{pb}$ and contribute full force to the equilibrium, and the section would report its full capacity with nothing to say so.

When the limit would be exceeded, ACS constrains the solve rather than reporting that number: the extreme-fibre concrete strain $\varepsilon_{cu}$ is reduced until no bonded tendon exceeds its limit, so tendon strain governs the ultimate instead of concrete crushing. The reduced $\varepsilon_{cu}$ is disclosed on the PT tab and in the PDF report — an ultimate solved away from the code strain is not something you should have to infer.

Capacity typically moves very little — the tendon’s stress is nearly flat approaching $f_{pb}$, so bounding its strain barely changes the force it carries. The value of the check is that an over-strained strand is now visible instead of silently reported as intact.

Unbonded tendons are excluded, and the exclusion is not an omission: an unbonded tendon carries the member-average Cl. 8.1.8 $\sigma_{pu}$, which no local strain compatibility produced, so there is no local strain to limit.

If a tendon is over its limit at every positive $\varepsilon_{cu}$ — i.e. under its locked-in prestrain alone, before any load — the section has no ultimate to report and $M_{uo}$ is withheld with the reason tendon_over_strained (see API reference). The fix is the strand grade, its declared $A_{gt}$, or $\sigma_{p.ef}$ — not the applied load.

Unbonded tendons

For unbonded tendons, the tendon is free to slide within the duct, so the stress increase is limited. AS 3600 Cl. 8.1.8 provides two simplified formulae depending on the span-to-depth ratio of the member:

Item (a) — span/depth ratio ≤ 35 (compact members):

Item (b) — span/depth ratio > 35 (slender members):

In both cases $f_{ps}$ is additionally capped at $f_{py}$ (the tendon yield stress).

Where:

• $f_{pe}$ = effective prestress after all losses (MPa)
• $f'_c$ = characteristic compressive strength of concrete (MPa)
• $\rho_p$ = tendon reinforcement ratio ($A_p / (b \cdot d_p)$)

Enter the member span-to-depth ratio in the Applied Loads panel. ACS automatically selects Item (a) or Item (b) based on this value. For span/depth > 35 the Item (b) cap of $f_{pe} + 200$ MPa is significantly lower than the Item (a) cap of $f_{pe} + 400$ MPa — using an incorrect branch overstates the ultimate capacity of slender post-tensioned members.

Results

The PT tab reports the tendon stress and strain; the ultimate capacity itself is on the ULS tab, as $M_{uo,x}$ / $\phi M_{uo,x}$. Both come from the same strain-compatibility solve, so the two tabs cannot disagree.

Tips and best practices

• Verify that $f_{pi}$ does not exceed 0.85 $f_{pu}$ (AS 3600 Cl. 3.4.2) or 0.80 $f_{pu}$ (ACI 318) — exceeding these limits risks strand relaxation or rupture during jacking
• If transfer stresses exceed allowable limits, consider debonding some strands near the ends or raising $f'_{ci}$
• For post-tensioned members, place the tendon at the location of maximum eccentricity (typically near the bottom at midspan)
• Check that the effective prestress $f_{pe}$ provides sufficient precompression to raise the cracking moment $M_{cr,t}$ above the maximum service moment — cross-reference the PT cracking moment result with the SLS stress check
• For slender post-tensioned slabs (span/depth > 35), the Item (b) formula applies and the achievable tendon stress at ultimate is lower; ensure the capacity check accounts for this
• The bond type significantly affects ultimate capacity: bonded tendons develop higher stresses at ultimate than unbonded tendons
• The library’s 1860 MPa strands declare $f_{p0.1k} = 1636.8$ MPa (prEN 10138-3 White Draft May 2006, Table 4 note (c) — a draft; see the strand library section), so their $f_{py}$ is that value under every design code; only a product with no declared proof stress — the 1720 MPa grades — falls back to $0.82\,f_{pb}$ per AS 3600 Cl. 3.3.1(b)(iii); a tendon with no strand material at all is refused before the analysis runs (see above)
• The moment-curvature (M-$\kappa$) curve and interaction diagram both incorporate the tendon prestrain — the cracking point on the M-$\kappa$ curve equals $M_{cr,t}$ from the PT cracking moment panel
• Use flat ducts for slab-band tendons with 1–5 strands; the wider footprint lowers the required cover relative to a round duct of equivalent area, and ACS uses the actual flat-duct width in the Cl. 8.2.1.5 shear deduction
• For crack control on PT slabs, the Cl. 8.6.3 decompression route often governs; if the decompression check fails, adding slightly more prestress (or reducing the service moment) typically brings the section into compliance without changing the reinforcement layout
• Review the Δσ_p column in the time-dependent panel alongside the losses breakdown — the two analyses use independent paths and should be compared to confirm consistent loss estimates

Related pages

• Reinforcement — placing tendons in the section
• Section analysis — overview of all analysis types
• Time-dependent effects — creep and shrinkage analysis
• RC beam example — worked example (reinforced, not prestressed)