Summary of expert opinion - collision reconstruction

Claim reference CLAIM-A · prepared 26 August 2026 · CapSeal Recon 0.2.0

1 Identification and status

This is a draft summary of expert opinion prepared in the structure of a summary under Rule 36(9)(b) of the Uniform Rules of Court. It has been generated by CapSeal Recon from the sealed evidence listed in section 4 and is placed before an independent reconstruction expert for review, correction and, if the expert is satisfied, adoption and signature.

Until it is adopted and signed by that expert it is a draft and it is not expert evidence. The expert, not the software and not the party, is the witness.

2 Duty of the expert

The expert who adopts this report does so on the following terms, which are reproduced here so that they are part of the document the court reads.

• My duty is to assist the court on the matters within my expertise. That duty overrides any obligation to the party by whom or through whom I am engaged.
• I have set out in this report every assumption on which my opinion depends, and the range over which each assumption was tested.
• Where the physical evidence does not establish a matter, I have said so rather than expressing an opinion the evidence does not support.
• I have not been asked to, and do not, express any opinion on the apportionment of fault. Apportionment under the Apportionment of Damages Act 34 of 1956 is a matter for the court.

3 Instructions and scope

I was instructed to determine, so far as the physical evidence allows, the speeds of the two vehicles at the instant of contact, the order in which they entered the intersection, and whether each driver's account of the collision is capable of producing the physical evidence recorded at the scene.

The analysis is a two-vehicle planar reconstruction. It does not address occupant kinematics, injury causation, vehicle defect, or the conduct of either driver beyond what the physical evidence establishes.

4 Evidence considered

Every document and file I considered is listed below. Each was hashed on receipt and recorded in a sealed manifest with a chain-of-custody log; the manifest travels with this report as 00_manifest and can be re-verified independently.

All 10 evidence files in this bundle are intake-sealed. They were hashed and timestamped on receipt by PayFar; integrity is asserted from intake onward only. No file in this bundle is capture-sealed, so capture-time integrity is not asserted for any of them. This is the expected position for a retrospective POC claim.

• Every file in this bundle is intake-sealed. Capture-time integrity is not asserted for any of them; this is the expected position for a retrospective POC claim (section 6, step 1).

Provenance tiers: capture-sealed material was hashed at the moment of capture; intake-sealed material was hashed on receipt and no assertion is made about its integrity before that moment.

5 Assumptions

The table below lists every numerical assumption the analysis depends on, the value adopted, the range over which the analysis was repeated, how confident I am in the value, and where it came from. Nothing else was assumed.

The following parameters were not documented for these vehicles and were taken from a specification database or a segment-class default: cg_from_front_axle_ratio, cg_height_m, front_overhang_share, height_m, kerb_mass_kg, length_m, occupant_mass_kg, payload_mass_kg, track_m, wheelbase_m, width_m, yaw_inertia_factor. Each was swept across the range shown, so the effect of that estimation on the answer is visible in section 9 rather than hidden in a single figure.

6 Inputs

The scene was modelled to scale from the source stated below together with the assessor's site measurements. Accuracy is bounded by those measurements; no survey-grade scan was carried out.

• friction: no site measurement; a published range for asphalt / dry is used and swept. This is expected for a retrospective claim and is stated in the report.

Where no published crush stiffness coefficients exist for a model, no crush-energy analysis was performed for that vehicle and the reconstruction rests on momentum and trajectory matching alone.

7 Method

The reconstruction was carried out in four steps.

• Impact. A planar impulse-momentum analysis of the two vehicles at the contact configuration recorded in section 6, with a coefficient of restitution along the impact-plane normal and an impulse-ratio condition along the tangent. Linear and angular momentum are conserved exactly; the analysis rejects any solution that would create kinetic energy.
• Separation. Each vehicle was then integrated to rest as a planar rigid body on four tyre contact patches, each patch developing friction opposing its own slip velocity. The final heading of each vehicle is therefore an output of the analysis, not an assumption.
• Fit. The pre-impact speeds were solved by repeating the above until the simulated rest positions and headings matched the documented ones, each weighted by the uncertainty recorded for that document. The fit was started from 25 different starting points to test whether the answer is unique; 1 distinct solution(s) were found.
• Uncertainty. Every assumption in section 5 was then varied, first one at a time and then all together by Monte Carlo, with the whole fit re-run each time. The speed ranges in section 8 come from that sweep, not from an error estimate placed around a single answer.

8 Results

The reconstruction reproduces the documented rest positions to within the figures in the second table. That agreement is what gives the speeds their weight: they are not read off a formula, they are the speeds that put both vehicles where they were actually found.

Vehicle B crossed its stop line 1.14 s before vehicle A crossed its own, on a 5th-to-95th percentile interval of 1.02 to 1.27 s. The interval does not include zero, so the entry order is established by the physical evidence at this confidence.

Signal phase timing was requested from the municipality and was not obtained. Conclusions are framed on entry order and speed, not on light colour. No clip in this bundle shows a signal head, so this reconstruction makes no statement about which vehicle had a green light. It establishes entry order and speed.

• Sight line, vehicle A. The view became clear 3.9 s before impact and remained clear to impact. Earlier in the window it was obstructed. Plan-view test at an eye height of 1.10 m; obstructions recorded as lower than that are ignored. This states geometric availability of a view, not what a driver observed.
• Sight line, vehicle B. The view became clear 3.9 s before impact and remained clear to impact. Earlier in the window it was obstructed. Plan-view test at an eye height of 1.10 m; obstructions recorded as lower than that are ignored. This states geometric availability of a view, not what a driver observed.

The threshold adopted for this work is 1.5 m and 15 degrees. This reconstruction is within that threshold.

9 Uncertainty

No single figure is offered for either speed. The ranges in section 8 are the 5th to 95th percentile of a Monte Carlo sweep in which every assumption in section 5 was drawn from its stated range simultaneously and the whole reconstruction re-solved for each draw.

The combined uncertainty leaves a workable speed range; the bundle presents the range, not a single figure.

120 of 120 Monte Carlo draws produced a usable solution; draws that failed to converge are excluded and counted, not replaced.

Each assumption was moved to each end of the range in section 5 with all others held at their adopted values, and the whole reconstruction re-solved.

10 The parties' accounts tested against the physical evidence

Each driver's account was then run through exactly the same physics, with the same masses, the same impact geometry and the same road surface. The only thing changed was the speeds each account asserts. The question asked is narrow: does that account put the vehicles where they were found?

The account of the driver of vehicle A cannot produce the documented rest positions. Run through the same physics, with the same masses, the same impact geometry and the same road surface, it leaves vehicle B 15.8 m from where it was found (and the worst heading error is 54 degrees), against a tolerance of 2.4 m and 24 degrees. The gap is what the animation shows.

The account of the driver of vehicle B can produce the documented rest positions: run through the same physics it leaves both vehicles within 0.06 m and 0.3 degrees of where they were found. This account is not excluded by the physical evidence. It does not follow that it is what happened - more than one account can be consistent with the same evidence.

An account that is not excluded by this test is not thereby established. More than one account can be consistent with the same physical evidence, and where that is so this report says so rather than choosing between them.

Tolerance applied: 2.4 m and 24 degrees, being the consistency threshold widened by the recorded uncertainty in the rest positions themselves.

11 Conclusion

On the physical evidence recorded at the scene, the collision is reconstructed as follows. The reconstruction reproduces both documented rest positions to within 0.06 m and 0.4 degrees.

• Vehicle A was travelling at between 49 and 56 km/h at the instant of contact.
• Vehicle B was travelling at between 37 and 43 km/h at the instant of contact.
• Vehicle B crossed its stop line before vehicle A, by between 1.0 and 1.3 seconds.
• The account of the driver of vehicle A is not capable of producing the documented rest positions and is therefore excluded by the physical evidence.
• The account of the driver of vehicle B is capable of producing the documented rest positions and is not excluded by the physical evidence.

11.1 Matters on which I express no opinion

I express no opinion on which driver was at fault, or on the apportionment of damages between them. What the physical evidence constrains is set out above; the legal consequences of those facts are for the court.

Nor do I express an opinion on which vehicle had a green light, except to the extent stated in section 8.

12 Limitations

The following limitations apply to this analysis and should be read with the conclusion.

• This is a two-vehicle planar analysis. It does not model occupant motion, pedestrian or cyclist multibody dynamics, motorcycle dynamics, rollover, or structural crush by finite element analysis.
• The scene was measured by tape and hand laser against aerial imagery. It was not surveyed. Plan accuracy is +/- 0.20 m.
• The tyre-road drag factor was not measured at the site. A published range for the surface and condition was used and swept.
• Signal phase timing was requested from the municipality and was not obtained. Conclusions are framed on entry order and speed, not on light colour.
• Telematics, where present, is treated as corroboration of the reconstruction, not as its basis. Its resolution is stated per vehicle in 04_timeline.
• I express no opinion on apportionment of fault.

12.1 What this analysis does not assert

The following are recorded so that no reader takes them as implied.

• That any bundle will be admitted in any court, or that a court will prefer it over an opposing expert.
• That all eight claims will produce a tight speed solution; one or two may be momentum-only with wide ranges, and will be labelled so.
• That the POC pipeline is the production platform; it is an internal tool built to test the concept and to measure unit economics.
• That a fault percentage will be produced; the bundle constrains physical facts and the animation demonstrates which account is physically consistent.

Declaration and signature