Multi-Material Chassis Optimisation for Crash and Stiffness Targets

The problem

Future EV structures mix materials, putting high-strength steel where crash loads are high, aluminium where stiffness per kilogram matters and composites in panels. Choosing which material goes where, and how thick, is a combinatorial problem that a few manual iterations cannot settle, and crash performance is hard to include early.

Objectives

• Parameterise a chassis so each member group can take one of several materials and gauges.
• Define stiffness, mass and a crash-energy proxy as objectives and constraints.
• Run a design of experiments of Simulation studies driven by Design Tables.
• Fit surrogate models in MATLAB and find a Pareto front with multi-objective optimisation.
• Verify selected Pareto designs with full Simulation and discuss joining and cost limits.

What you need

Key concepts

Design of experiments

A planned set of design points, such as a Latin hypercube, that covers the design space efficiently so a surrogate can be fitted with few expensive runs.

Surrogate model

A fast approximation, such as a Gaussian process or polynomial, fitted to Simulation results. It allows thousands of evaluations in an optimiser.

Pareto front

The set of designs where no objective can be improved without making another worse. It shows the trade-off between mass and stiffness or crash performance.

Crash-energy proxy

A simplified measure, such as energy absorbed in a quasi-static crush of a rail to a fixed displacement, used early when full crash simulation is too costly.

Step-by-step guide

1. 1

   Frame the optimisation problem

   Choose an example EV front structure plus floor: front rails, crash box, cross members, sills and floor panel. Group members into four to six groups. For each group, allow example materials (mild steel, high-strength steel, 6xxx aluminium, a quasi-isotropic laminate) and a gauge range. Define objectives: minimise mass and maximise torsional stiffness, with constraints on first mode frequency and crash-energy proxy.

   Checkpoint: A problem statement with variables, ranges, objectives and constraints written down before any modelling.

2. 2

   Build the parametric model

   Model members as surfaces or sheet bodies so they can be shells with thickness from a parameter. Use a Design Table to drive thickness and material for each group. Connections between groups use bonded contact, with a note that real joints add flexibility. Check that every row rebuilds and that Mass Properties updates correctly. Add a column in the Design Table for density so mass is always correct even when a laminate is approximated by an equivalent material.

   Checkpoint: Ten random Design Table rows rebuild without error and give the expected masses.

3. 3

   Set up the response studies

   Create a Static torsion study and a Frequency study using the same mesh settings. Run mesh convergence on the baseline for torsional stiffness and first frequency, refining until change is below about 2 per cent. For the crash proxy, set up a Nonlinear quasi-static crush of the front rail to a fixed displacement with plastic material models, or, without Premium, use an analytical folding-model estimate for thin-walled tubes from published methods.

   Checkpoint: Converged baseline responses and a working crash-proxy calculation.

4. 4

   Plan the design of experiments

   In MATLAB, generate a Latin hypercube for continuous gauges with lhsdesign and combine it with material choices as categorical levels. Choose around ten times the number of variables as a starting sample size. Export the plan as rows of a Design Table or as an Excel sheet read by SOLIDWORKS Design Study. Keep a seed so the plan is reproducible.

   Checkpoint: A DOE file with design points that SOLIDWORKS can read and that spans every variable range.

5. 5

   Run the design points

   Run all points through the studies, using Design Study to batch the runs or a SOLIDWORKS API macro if you know VBA or C#. Record mass, torsional stiffness, first frequency and crash proxy for each point. Check a few points by hand for sanity, such as mass, and look for outliers caused by mesh failure rather than physics. Store results with the DOE row number and a time stamp so failed runs can be repeated without confusion.

   Checkpoint: A results table with every DOE point and no unexplained outliers.

6. 6

   Fit and test surrogates

   In MATLAB, fit a surrogate for each response, for example fitrgp for Gaussian process regression. Treat material choice as categorical or fit separate surrogates per material combination if there are few. Use k-fold cross-validation and report prediction error. Add infill points in regions of high error and refit until the error is acceptable for each response. Plot predicted against actual values for each response so poor fits in parts of the space are easy to see.

   Checkpoint: Cross-validated error for each surrogate below a stated limit, such as 5 per cent for stiffness.

7. 7

   Find the Pareto front

   Run gamultiobj on the surrogates with constraints on frequency and crash proxy, or a dense grid search if Global Optimization Toolbox is not available. Plot mass against torsional stiffness for feasible designs and highlight the Pareto front. Group Pareto designs by material pattern to see which material goes where on the efficient designs. Check that constraint handling works by confirming no infeasible design appears on the reported front. Report the knee point where extra stiffness costs most mass.

   Checkpoint: A Pareto plot with feasible designs and a table of material patterns along the front.

8. 8

   Verify selected designs

   Pick three Pareto designs, such as lightest feasible, stiffest and a knee point. Rebuild each in SOLIDWORKS and run the full studies. Compare Simulation and surrogate predictions. If they differ by more than the surrogate error, add these points to the training set and refit. Only report verified results as design outcomes. Present a table of predicted and verified values for each response and each verified design.

   Checkpoint: Three verified designs whose Simulation results agree with surrogate predictions within the stated error.

9. 9

   Assess joining, cost and practicality

   For each verified design, list material interfaces and possible joining methods, such as self-piercing rivets, flow-drill screws or structural adhesive, and note galvanic corrosion risks between steel, aluminium and carbon fibre. Estimate relative cost from mass and example material prices. Identify which designs would need a new production process. Flag designs that need curing ovens or new tooling, since these raise investment. Note where an insulating layer or coating is needed to stop galvanic corrosion between materials.

   Checkpoint: A practicality table for the verified designs covering joining, corrosion and relative cost.

10. 10

    Write up the method

    Document the problem formulation, model assumptions, mesh convergence, DOE, surrogate accuracy, Pareto results and verification. Discuss limits: bonded joints, quasi-static crash proxy, and linear material for composites. State that full crash performance must be checked with explicit crash simulation and physical tests. Provide scripts and Design Tables so the method can be reproduced. Include the problem statement written in step one so readers can see that objectives were fixed before results were known.

    Checkpoint: A reproducible package of model, DOE plan, scripts, results and a report with limitations.

Testing & validation

• Mesh convergence on baseline responses.
• Cross-validated surrogate error reported for each response.
• Pareto designs re-verified in full Simulation.
• Crash proxy method stated and its limits discussed.
• DOE seed and scripts provided for reproducibility.

Expected results

A reproducible multi-material optimisation workflow for an example EV chassis, with validated surrogates, a Pareto front of mass against stiffness under frequency and crash-proxy constraints, three verified designs and a practicality assessment of joining and cost.

Take it further

• Replace the crash proxy with explicit crash simulation in a dedicated crash solver for the verified designs.
• Model joint flexibility with spring connectors calibrated from tests.
• Add cost and embodied carbon as optimisation objectives.
• Compare surrogate types for accuracy versus number of training points.

Viva and review questions

Learn it at DIYguru

These DIYguru courses and programmes teach what this project needs.