Design-Stage Cost Modelling for EV Frames and Enclosures

The problem

Most of a product's cost is decided at the design stage, yet engineers often learn it only after quotes arrive. Automated CAD-based costing promises early feedback, but its accuracy depends on templates, rates and assumptions. For EV start-ups building frames and battery enclosures in low to medium volumes, a validated method for early cost estimates and design ranking is a useful research question.

Objectives

• Define a family of example frames and enclosures with design variants.
• Set up and calibrate SOLIDWORKS Costing templates for weldments, sheet metal and machined parts.
• Build an independent parametric cost model in a spreadsheet for comparison.
• Compare estimates with benchmark quotes or literature cost structures, and quantify error.
• Assess how reliably each method ranks design alternatives, and propose an improved approach.

What you need

Key concepts

Cost drivers

Material mass and price, process time, setups, tooling and overheads. A good model links each to measurable geometry, such as cut length or bend count.

Template calibration

Adjusting rates and times in costing templates so estimates match known costs for a reference set of parts.

Rank correlation

A statistic, such as Spearman's, that measures whether two methods put design options in the same order, which matters more than absolute accuracy for early decisions.

Activity-based costing

Assigning costs to the activities a part needs, such as cutting, bending, welding and painting, rather than spreading overhead evenly.

Step-by-step guide

1. 1

   Frame the research question

   State a question such as: how accurately does CAD-based costing estimate and rank low-volume EV frame and enclosure designs, and how much calibration is needed? Review literature on design-stage cost estimation, parametric and feature-based methods and their reported accuracy. Define the volumes of interest, for example 500 to 5,000 units a year, and the processes: tube cutting and welding, laser cutting and bending, and CNC machining.

   Checkpoint: A written question, hypotheses and a literature summary with the gap your study addresses.

2. 2

   Build the design family

   Model an example two-wheeler frame in Weldments and a battery enclosure in Sheet Metal, then create variants with configurations: two tube profiles, two wall thicknesses, welded versus bolted joints, and a machined versus folded bracket. Use design tables so each variant is reproducible. Keep design intent clear so the variants are fair, for example similar stiffness, checked by a quick static study if needed.

   Checkpoint: At least eight variants across frames and enclosures, each rebuilding cleanly with a recorded cut list or flat pattern.

3. 3

   Collect benchmark cost data

   Obtain reference costs by asking local fabricators to quote a subset of variants from drawings, under an agreed confidentiality note, or by building costs from published process times and rates if quotes are not possible. Record volume, material, finish and what is included. Be explicit that quotes include margin and vary by supplier, so you treat them as a benchmark with uncertainty, not truth.

   Checkpoint: A benchmark table with cost, source type and assumptions for at least six variants.

4. 4

   Set up Costing templates

   In the Costing Template Editor, create weldment, sheet metal and machining templates with material prices, cutting rates, bend costs, weld cost per length, setup times and labour rates appropriate to your benchmark context. Keep a log of every value and its source. Run Costing on each part and assembly, review the Costing Manager breakdown, and fix any misread features, such as cuts treated as machining.

   Checkpoint: Costing runs on every variant with a cost breakdown, and each template value is logged with its source.

5. 5

   Build a parametric model

   Independently, build a spreadsheet or Python model that takes geometric measures exported from SOLIDWORKS, such as mass, cut length, bend count, weld length and hole count from Mass Properties and cut lists, and computes cost from activity rates. This gives a transparent second method. Use the same rates as the templates where they overlap, so differences come from method rather than inputs.

   Checkpoint: A parametric model that reproduces a hand calculation for one simple part exactly.

6. 6

   Calibrate on a training set

   Split variants into a calibration set and a test set. Adjust template rates and the parametric model on the calibration set only, minimising error against the benchmark. Record how many parameters you tuned. Avoid tuning on the test set, because that hides the real predictive error. State the uncertainty of the benchmark itself as part of the comparison. Plot estimated against benchmark cost for the calibration set, and check that the residuals show no obvious pattern with part size or process.

   Checkpoint: A calibrated model with a documented list of tuned parameters and calibration error.

7. 7

   Test accuracy and ranking

   Apply both calibrated methods to the test set. Compute mean absolute percentage error and bias against the benchmark, and Spearman rank correlation between each method's ranking and the benchmark ranking. Plot estimated versus benchmark cost. Check whether either method misranks options that a designer would actually choose between, such as welded versus bolted joints. Repeat the comparison for the uncalibrated templates as well, so the benefit of calibration itself is measured rather than assumed.

   Checkpoint: Error and rank correlation values for both methods on the test set, with any misranked pairs identified.

8. 8

   Analyse sensitivity and drivers

   Vary key inputs such as material price, labour rate and volume one at a time, and see how estimates and rankings change. Identify which geometric features drive cost most, for example weld length in frames or bend count in enclosures. Discuss which drivers are poorly represented by CAD-only estimates, such as fixture costs, finishing and scrap. Use a tornado chart for the most expensive variant so the reader sees at a glance which input moves the estimate most.

   Checkpoint: A sensitivity chart and a list of drivers that CAD costing captures well and badly.

9. 9

   Propose an improved method

   Based on the errors, propose an improvement, for example adding fixture and finishing activities to templates, or a correction model fitted on residuals. Test it on the held-out set with the same metrics. Be careful not to overfit with few data points, and report confidence intervals or bootstrap ranges for errors. Keep the improved method simple enough that a designer could apply it without your scripts, otherwise it will not be used in practice.

   Checkpoint: The improved method shows a measured change in error and ranking on held-out data, with an uncertainty range.

10. 10

    Write up and share

    Write the thesis chapter or paper: question, design family, data sources, templates, parametric model, calibration, results, sensitivity and limitations, including small sample size and benchmark uncertainty. Anonymise supplier data. Provide the models and templates as appendices so others can reproduce the study. Discuss practical guidance for EV start-up designers using early costing. Include a short worked example in which a designer uses the method to choose between two frame variants, showing how the result guides a real decision.

    Checkpoint: A reproducible report with anonymised data and appendices, reviewed by a supervisor.

Testing & validation

• Calibration and test sets are separate.
• Both methods are compared on the same benchmark with stated uncertainty.
• Error metrics and rank correlations are reported with confidence ranges.
• Parametric model reproduces a hand calculation.
• Limitations of benchmark data and sample size are stated.

Expected results

A calibrated set of Costing templates and a parametric cost model for EV frames and enclosures, with measured accuracy and ranking reliability against benchmark costs, and a proposed improved method.

Take it further

• Add environmental impact from SOLIDWORKS Sustainability for a cost and carbon comparison.
• Extend to cast and moulded parts.
• Build a simple design-review dashboard that shows cost as variants change.

Viva and review questions

Learn it at DIYguru

These DIYguru courses and programmes teach what this project needs.