Verification
Are the stated equations and numerical methods implemented correctly? Analytical equilibrium, signs and symmetry detect force-balance errors. Exact model reduction checks correspondence to a simpler trusted model. Deterministic repeatability and Worker/direct parity check the execution path and transferred evidence.
Numerical verification
Independent state-space implementations assemble equations separately from production RK4. Transient references use matrix-exponential propagation; frequency checks use an independent harmonic system and settled transient cross-checks. Step refinement tests convergence on specified smooth cases. Numerical agreement and convergence are complementary checks.
Full-Car compares all fourteen states and includes a final partial step. Its four prescribed roads and Half-Car reduction remain explicit. Inspect FC-001 through FC-008 and their acceptance criteria.
Published numerical reproduction
An independently published mathematical-model result provides external numerical evidence. The source, parameter mapping, coordinate convention, event alignment and reproduction scope must be traceable. Digitization uncertainty limits comparison to plotted data; independently calculated consistency references serve a different purpose.
The Zhao 2015 source audit retains its DOI, extracted curve provenance and quantitative digitization-aware criteria.
Physical validation and correlation
This requires measured data, controlled and traceable test conditions, suitable instrumentation and intended-use acceptance criteria. It has not been performed and is not currently claimed for VehicleLab.
Acceptance criteria
Canonical cases declare tolerances and comparison operators before interpretation. Model pages expose the actual ordered bounds and exact parity assertions from the existing registry or canonical tests. Each state retains its physical units; a mixed-unit error norm does not replace per-state acceptance.
Computed plot PASS/FAIL uses full-resolution arrays. Deterministic plotting/downsampling occurs only after acceptance. Registry and regression statuses describe reviewed automated checks; the static page does not rerun the entire browser or Worker suite.
Frequency results retain separate solver identities. A successful check supports its stated case and reference type, not every possible parameter set or physical vehicle.
Choose the relevant evidence
Quarter-Car evidence · Half-Car evidence · Full-Car evidence
Read the evidence boundaries →