Design Automation

Design and Optimize

Move from a failing or oversized design to a passing, well-sized option without manual trial and error. AutoDesign searches the design space you define, while you keep control of what can change and whether the final result is applied.

Less iteration

One run replaces dozens of manual size-and-recheck cycles.

Same checks

Uses the exact calculation logic behind your design report.

Engineer in control

You define the design space; the result is reviewed before it applies.

Recommended workflow

Search first, then Design and Optimize

Suggest is AutoDesign's quick search — it finds a passing starting point in seconds. Run it before launching the full Design and Optimize. The two-step pattern is dramatically faster than asking the optimizer to escape a failing design from scratch.

1. Click Suggest. Get a passing parameter set quickly.
2. Review and adjust. Deselect impractical values, or run Suggest again on the kept set.
3. Click Design and Optimize. Refine the suggested set into a clean, well-sized design.

Need to compare many cases at once? Pair Design and Optimize with Batch Analysis to auto-design every row in a worksheet.

How It Works

Think of Design and Optimize as a guided iteration loop. It uses the same checks that appear in the design results, changes only the inputs you allow, and keeps evaluating candidates until it finds a passing design or runs out of practical moves.

Start from your current design

AutoDesign reads the current inputs, active calculation settings, and the checks already shown in the result summary.
If any checks are failing, the tool highlights the inputs that are most relevant to those failures.
You decide which inputs AutoDesign is allowed to change before the run starts.

Redesign until it passes

When the current design fails, AutoDesign looks for the governing check and tests practical input changes.
It chooses changes that reduce the governing demand/capacity ratio most efficiently — and also accepts changes that flip an individual failing check to passing, even when the global ratio is unchanged.
If the heuristic search exhausts its iteration budget without producing a passing design, AutoDesign falls back to a combinatorial sweep across the unvisited combinations of selected inputs (see Combinatorial fallback).

Reduce the passing design

After a passing design is found, AutoDesign tries to step selected inputs back down.
A reduction is kept only when the design still passes all active checks.
This helps avoid stopping at an oversized design just because it was the first passing option.

Review and apply

The run shows the final status, governing ratio, iteration count, and calculation results.
You can inspect the log to see how the search moved through the design space.
The final passing result is applied back to the design form only when you choose Apply.

Engineer Control

The tool is meant to reduce repetitive iteration, not replace engineering judgment. Before running, you define the design space AutoDesign is allowed to explore.

What You Decide

Choose which inputs are allowed to change and leave critical project requirements fixed.
Set minimums, maximums, and step sizes for numeric inputs.
Choose which list options are allowed, such as available sizes, grades, or product choices.
Mark selected inputs as Maximize when the calculation should push them toward their maximum passing value after optimization.
Keep the pass threshold tied to the calculation settings so the optimization uses the same acceptance criteria as the design report.
Review warnings and geometry constraints before applying a result.

What AutoDesign Does

Evaluates each candidate using the live calculation model.
Rejects candidates that fail required checks or blocking validation warnings.
Looks for a practical passing design before attempting to reduce it.
Returns the final result for review instead of silently changing your form.

Running an Analysis

Clicking Design and Optimize opens the analysis modal and starts the search without freezing the rest of the interface. The optimizer yields between search chunks so the progress modal, log, and status indicators can keep updating while the calculation is running.

What stays locked while it runs

Parameter selection, Add all, Remove all, and selected-parameter removal are disabled.
Numeric min, max, step, and Maximize fields are disabled.
List option selection, ordering, removal, and Maximize fields are disabled.
The Settings, Suggest, and Design and Optimize buttons are disabled until the run finishes.
The analysis modal cannot be closed while the run is active, preventing a half-finished run from leaving stale state behind.

What keeps updating

The progress modal shows the current running state.
The run log continues to receive messages from redesign, optimization, fallback, and option-sweep phases.
When the run finishes, the status, governing DCR, iteration count, and limit-state table update in place.
Passing runs publish Other passing options to the panel without requiring you to apply the final result first.

Suggest

Beta — algorithm being improved

The Suggest button (next to Add all in the Selected Parameters panel) asks AutoDesign to propose a parameter set that may make the design pass, before you launch the full Optimize run. Use it to get a sensible starting point when you are not sure which inputs to unlock.

This feature is currently in beta.

The search algorithm is still being refined. The result is a hint, not a final design — always review the suggested parameters and bounds before running Optimize or applying them back to your form.

How AutoDesign finds the suggestion

When you confirm the run, AutoDesign tries up to three increasingly aggressive strategies — stopping as soon as one of them finds a passing design.

Try your current design. Evaluate the inputs as they are. If they already pass, AutoDesign just trims values that can be reduced and stops.
Pre-search around your design. Probe each selected parameter both up and down from your current values. Pick the change that lowers the governing demand/capacity ratio the most per step. Spread picks across parameters so one input doesn't dominate the suggestion.
Brute-force from minimums. If the pre-search can't find a pass, start every numeric parameter at its smallest practical value and work failing checks one at a time — raising parameters that affect the failing check first, then those shared with other checks if needed.

In step 3, AutoDesign uses your current design as a scale reference per parameter — a 25-step jump on an input that started at 4 is treated as a much bigger drift than a 25-step jump on one that started at 200. This keeps suggestions on a reasonable order of magnitude for each parameter.

Trim and rebalance

Whichever strategy passes, AutoDesign always cleans up the result before showing it to you:

Trim — step each parameter back down one at a time while the design still passes, so the suggestion is not oversized.
Rebalance — try pairwise swaps that reduce one parameter while raising another, accepting the swap only if the design still passes and the largest drift across all parameters goes down. This avoids "one parameter pushed to an extreme while everything else sits at minimum".

What happens to your inputs

Selected iterator min is set to the suggested value, so a follow-up Optimize run starts from a known-passing point.
Iterator max widens upward only when the suggested value exceeds it; bounds also widen to include the engineer's current values so they stay inside the search space.
List parameters are narrowed to the option the search settled on.
A toast confirms when suggested parameters were loaded, and it calls out when AutoDesign had to widen back to previously deselected parameters to make the design pass.
If no strategy passes, your prior selection and bounds are restored unchanged so you can adjust and try again.
While Suggest is running, the selection and parameter controls are locked and the log remains visible so you can monitor progress.

Helpers that protect against bad starts

Warning-clearing warmup. When the near-zero starting state would trigger a blocking validation warning (e.g. a geometric clearance is violated), AutoDesign first finds the smallest single raise on each parameter that clears the warning, then begins the main search from there — so the brute force never starts in an infeasible state.
Selection-aware retry. If you deselect parameters and Suggest can't pass with what you kept, it will widen back to all parameters as a fallback — and the toast will tell you it had to.
Cancellation. Closing the modal mid-run cancels the search cleanly and restores the iterator selection and bounds you had before clicking Suggest.

The work budget for steps 2 and 3 is set by Pre-Search Iterations in Optimization Settings. Higher budgets improve the chance of finding a passing configuration on difficult designs at the cost of a longer Suggest response time.

Suggested workflow

Open Design and Optimize on a failing design.
Click Suggest and confirm the beta warning.
Review the suggested values. Deselect any parameter whose proposed value is impractical for your project (e.g. a bolt count that is too high to detail).
Click Suggest again. AutoDesign now optimizes only the parameters you kept, and only widens back to the deselected ones if the narrower set can't pass.
Once the suggestion looks reasonable, click Design and Optimize to refine it with the full search.

Combinatorial Fallback

The redesign engine uses a sensitivity-guided heuristic that converges quickly when the design space is well-behaved. Some designs, however, only pass at specific combinations of values that the heuristic doesn't explore on its own. To catch these, AutoDesign automatically falls back to a combinatorial sweep when the heuristic search finishes without producing a passing design.

How it works

While the heuristic search is running, AutoDesign tracks every combination of selected-parameter values it has already evaluated.
If the run finishes still failing, AutoDesign enumerates the cartesian product of values within each selected iterator's range — every min → max step for numerics, and every option for lists.
For each unvisited combination, AutoDesign evaluates the design (still respecting candidate validation warnings). The first combination that passes is taken as the result.
The sweep stops when a passing combination is found, when all combinations have been tried, or when the configured budget is reached — whichever comes first.

When it helps

Designs where two or more parameters need specific values together (e.g. raising bolt diameter only helps if plate thickness is also at a specific step).
Cases where the heuristic stalls because no single-parameter change improves the governing ratio, but a multi-parameter change does.
Designs that pass at a non-obvious corner of the search space the heuristic doesn't naturally explore.

In the Design and Optimize UI, the fallback uses a fixed internal evaluation budget of 5000. When the budget is reached without finding a passing combination, AutoDesign returns the best result from the prior redesign attempt so you can adjust your bounds and try again.

Other Passing Options

A passing Design and Optimize run also fills the Other passing options panel. This panel is for browsing nearby passing combinations without rerunning the optimizer. Applying an option from this panel updates the design form immediately and marks the row as applied, but it does not show a toast notification.

Where the options come from

The final optimized result is always added as an option when the run passes.
AutoDesign remembers the first passing state found before post-pass reduction.
Starting from that first passing state, it sweeps selected-parameter combinations up to each parameter's configured max.
Numeric parameters are checked step-by-step from the first passing value to max.
List parameters are checked from the first passing option through the last selected option.
Only combinations that still pass every active check appear in the panel.
In the Design and Optimize UI, the sweep reuses the same fixed 5000 evaluation budget as the fallback search.

Sorting and applying

Small sorts by the smallest selected-parameter size score.
Eff sorts by highest governing DCR first, showing options closest to the pass threshold.
Close sorts by the smallest drift from your starting design.
The enlarged modal adds per-parameter sorting. Click a parameter once for ascending order and again for descending order.
Parameter ordering chosen in the enlarged modal remains active when the modal closes, so the compact panel uses the same order.
When the panel or modal list has focus, ArrowDown applies the next option and ArrowUp applies the previous option.
Keyboard navigation scrolls the list so the applied option stays visible.

Compact panel

The compact panel has a fixed maximum height and scrolls internally.
In fullscreen drawing mode, the panel is capped at one third of the drawing height.
In fullscreen, the panel starts minimized by default after Design and Optimize so it does not cover the drawing.
Restoring a minimized panel scrolls to the applied option when one exists.

Enlarged modal

The enlarge button opens the same option set in a larger modal.
The modal can still apply options to the current design form.
Opening the modal scrolls to the applied option when one exists.
The modal is the best place to use per-parameter ascending and descending ordering.

Optimization Settings

The Settings button controls how hard AutoDesign searches before stopping. These values are not engineering requirements by themselves; they are search controls that balance speed, thoroughness, and predictability.

Max Redesign Iterations

Default: 300

Limits how many redesign moves AutoDesign can make while trying to turn a failing design into a passing design.

If increased: Increasing it gives difficult designs more chances to pass, especially when several inputs must change together. The tradeoff is a longer run when the selected bounds still cannot produce a passing design.

Max Optimize Iterations

Default: 100

Limits how many reduction passes AutoDesign can make after it has already found a passing design.

If increased: Increasing it lets AutoDesign keep trimming a passing design for more passes. This can help when many selected inputs can be reduced one at a time, but it may add time after the design is already acceptable.

Pass Threshold

Default: Calculation Settings DCR

Uses the same demand/capacity acceptance threshold as the visible calculation settings, so optimization and reporting agree.

If increased: If the project threshold is increased, designs with higher demand/capacity ratios can pass. For example, raising the threshold above 1.00 makes the acceptance criteria less conservative, so this should match your intended design basis.

Sensitivity Probe Steps

Default: 15

Controls how far ahead AutoDesign looks when estimating which input change is most likely to improve a failing design.

If increased: Increasing it lets the search look past a small first move. For example, if one size step does not improve a ratio but two or three steps do, a higher probe value can identify that direction sooner. Values beyond an iterator range simply stop at the available bound for that parameter.

Max Expand Steps

Default: 15

Allows larger jumps when one-step changes do not improve the design, which helps cross flat spots in the design space.

If increased: Increasing it allows larger fallback jumps. This can help when a design only improves after a bigger change, such as moving several size increments at once, but it can also skip over smaller intermediate options.

Pre-Search Iterations

Default: 50

Iteration budget used by the Suggest button when it runs a quick pre-search to propose a passing parameter set.

If increased: Increasing it raises the chance of finding a passing configuration in difficult designs but makes Suggest take longer to respond.

Improving the Chance of a Pass

When a Design and Optimize run finishes without a passing design, walk through the steps below in order. Each step is more aggressive than the previous one — usually you only need the first one or two.

1. Loosen your bounds (cheapest fix)

Widen iterator min and max ranges so the search has more values to try. Tight bounds are the most common reason a run fails.
Make sure the parameters you selected actually affect the failing checks shown in the failing-design-checks panel.
For list parameters, add the options you'd actually accept (e.g. larger plate sizes, higher-grade steel).

2. Give the search more time

Raise Max Redesign Iterations (default 300). Easiest knob — same step granularity, just more chances to find a pass. Doubling it is a reasonable first try.
If you're using Suggest, raise Pre-Search Iterations (default 50) for the same effect.

3. Help the search escape flat regions

When DCR doesn't change after a small parameter move but does change after several steps (for example, a thicker plate that has to clear a discrete bolt geometry threshold), the search can stall. Raise these:

Max Expand Steps (default 15) — lets the engine try 2-, 3-, ... up to N-step jumps when no single step improves the design. Practical ceiling is the size of your largest iterator range.
Sensitivity Probe Steps (default 15) — lets the engine score parameters by what they do over several steps ahead, not just one. Useful when small moves don't show improvement but bigger ones would.

Both knobs make each iteration slower in proportion. Raise one notch at a time (e.g. 15 → 25 → 35) rather than jumping to a very high value.

4. Try Ignore Maximum Values

If the run still fails, the analysis modal offers Ignore Maximum Values. Confirming it lets the search continue past your configured maximums. The result may exceed bounds you originally entered, so review carefully — this is meant to tell you "what would have to change for the design to pass," not to silently override your sizing intent. See Ignore Maximum Values for details.

Try Suggest first when you're stuck

Before tuning settings, click Suggest. It runs a faster search and may already produce a passing configuration with bounds wide enough for the full Optimize to refine. If Suggest can't find a pass either, your bounds or parameter selection probably need to widen before any setting tweak will help.

If nothing works

A run that genuinely cannot pass is sometimes telling you the design isn't feasible in the form you're constraining it. Check that your loads, member sizes, and connection geometry are realistic for the limit states being checked. The combinatorial fallback already searches every unvisited combination of selected values, so if it returns failing too, the design space you've selected truly does not contain a passing answer.

Why This Helps

Many detailed designs are solved by repeated trial and error: change a size, rerun checks, adjust spacing, rerun checks, change material, rerun again. Design and Optimize automates that loop while keeping the final decision visible and reviewable.

Less manual iteration

Run the search once instead of repeatedly editing inputs and scanning ratios.

Better starting point

Get a passing option that can be reviewed, adjusted, and documented.

Consistent checks

The search uses the same calculation logic behind the visible design results.

Ignore Maximum Values

When Design and Optimize finishes without a passing design — usually because the parameter maximums you set are too tight — the analysis modal offers an Ignore Maximum Values option. Confirming it continues the search from the last candidate, this time allowing numeric parameters to grow past their configured maximum.

Only numeric maximums are ignored. Minimums, list choices, and validation warnings still apply.
The reduction pass that follows a passing design still respects the configured minimums.
List parameters still stop at the highest option you selected.
Geometry and validation warnings still block infeasible candidates.
The resulting design may exceed the maximum bounds you originally entered, so it must be reviewed before you apply it back to the form.

Assumptions and Practical Limits

The search is discrete. Numeric inputs move by the configured step size; list inputs move through the selected choices.
The redesign method is sensitivity-guided, not a full exhaustive search of every possible combination.
The optimization pass finds a practical local minimum in the selected design space.
After a passing run, the option sweep checks combinations from the first passing state upward to each selected parameter's max, not combinations below the first passing point.
The Other passing options panel is a browsing and application surface. It does not replace the engineering review of the final design.
Only selected parameters are changed. Unselected parameters remain fixed at the current design value.
Derived values are regenerated through the same calculation flow used by the designer.
Applying a result writes the solved parameters back to the form only when the run status is passing.