The first data off the rig | Building a Steer-by-Wire System

27 runs across five instruments, three temperatures, two voltages, and two sweep rates. Five spec failures remain after correcting the on-center slope spec — all concentrated in two metrics at specific conditions.

5 spec failures remain after correcting the on-center slope spec to 0.005 N·m/deg — all concentrated in two metrics at specific conditions.

27 runs across 5 instruments · 3 temperatures · 2 voltages · 2 sweep rates · 135 raw files

Data preparation in five steps

Five instruments logged the same bench simultaneously with incompatible channel names, mixed units, and unsynchronised clocks. The pipeline resolved all of this before any spec check ran.

The overnight pipelineunattended · ~8 min

1Ingest

135 files → 17 canonical channels

5 sources

2Align clocks

cross-correlate to a 100 Hz grid

±1 s

3Sensor QC

drift · dropout · spikes · truncation

58 flags

4Stabilised points

held-point detection at ±200 / ±400°

100 holds

5Spec checks

6 metrics on every run

143 checks

Every step runs in sequence, on every run, with no one carrying files between tools. By morning the cleaned dataset and the first spec pass are on the shared drive.

Step	What happened	Result
1 · Ingest	135 raw files (MF4 + CSV) read. Channel names mapped to one canonical set. Robot rad→deg, rack→rwa via 1.8 mm/deg, ECU-A current→torque via Kt = 0.20 N·m/A.	Complete
2 · Clock alignment	Per-source offset estimated by cross-correlating `hwa_cmd_deg` against the rig master (±3 s window). All sources resampled to a common 100 Hz grid.	Offsets: ±1 s max
3 · Sensor QC	NaN gaps, frozen dropouts, spikes, warm-up masking, and truncation detection across all 17 canonical channels × 27 runs.	58 instances flagged
4 · Stabilised points	Held-point detection at ±200 deg and ±400 deg. 25 of 27 runs have holds; run07 and run18 truncated before the held-point phase.	100 holds detected
5 · Spec checks	6 metrics computed per run and compared against the baseline spec sheet (on-center slope spec corrected to 0.005 N·m/deg).	5 FAILs

Source mapping

Five sources, one dataset27 runs · 135 files

Rig DAQMF4 · 1 kHzSWA_cmd_deg, FbTrq_Nm

Steering robotCSV · rad, mmcmd_angle, rack_pos_mm

Torque / angle sensorsCSV · raw countstrq_raw, ang_raw

ECU-ACSV · currentI_mot_A, HwaCmd

ECU-BCSVtheta_cmd, M_fb

harmonise

Canonical dataset

17 channels · 100 Hz

hwa_cmd_degrwa_degfb_torque_Nmmotor_torque_Nm+ 13 more

Applied automaticallyrad → degrack ÷ 1.8 mm/degcurrent × 0.20 N·m/Aclocks → 100 Hz grid

Canonical signal	Rig DAQ	Robot	ECU-A	ECU-B	Conversion
`hwa_cmd_deg`	SWA_cmd_deg (deg)	cmd_angle (rad)	HwaCmd	theta_cmd	robot: rad → deg
`rwa_deg`	RWA_deg (deg)	rack_pos_mm	—	—	rack / 1.8 mm/deg
`fb_torque_Nm`	FbTrq_Nm	—	TrqFb	M_fb	—
`motor_torque_Nm`	MotTrq_Nm	—	(I_mot_A × Kt)	T_mot	ECU-A: current × 0.20 N·m/A
`motor_current_A`	—	—	I_mot_A	—	—

Channel health — nothing that blocks the analysis

All faults are minor and isolated. No canonical rig channel is compromised on more than one run.

54 spike instances — motor torque/current, 23 runs. Sparse 1–4 sample events. Peak metrics use 99.9th percentile.
3 NaN gaps — rwa_deg__robot on run04 (5 samples) and run12 (4 samples); motor_torque_Nm__rig on run14 (1 sample).
1 frozen dropout — rwa_deg__robot on run03, 540 samples while rig rwa was moving. Rig copy is clean; used as primary.

Where the campaign stands — 5 failures, 2 metrics

Spec-check matrix27 runs · 6 metrics

Steering ratio

FB torque @ lock

On-center slope

Motor torque

Motor current

On-center hyst.

−10 °C

25 °C

80 °C

Pass

Fail

Not evaluated

138 of 143 checks pass. The five failures sit in just two metrics (feedback torque at lock, all at 80 °C, and on-center hysteresis at a single condition). The other four metrics are clean across the campaign.

Metric	Spec target	Tolerance	PASS	FAIL
Steering ratio	15.0	±5 %	27	0
FB torque at lock	3.5 N·m	±1.5 N·m	22	3
FB on-center slope	0.005 N·m/deg	±50 %	18	0
Motor torque peak	≤ 8.0 N·m	—	27	0
Motor current peak	≤ 40.0 A	—	27	0
On-center hysteresis	≤ 0.4 N·m	—	16	2

Four of six metrics are clean campaign-wide. Steering ratio holds at 15.000 ± 0.001 across all 27 runs. Motor torque peaks at 2.44 N·m — well below the 8.0 N·m limit.

Feedback torque at full lock — 3 FAILs at 80 °C only

Feedback torque at full lockspec 3.5 ± 1.5 N·m

Three runs at 80 °C (×) reach ~5.38 N·m, just past the 5.0 N·m limit and about 20% above the temperature-scaling model. Every other condition sits inside the band. A temperature trend, not random scatter.

Spec: 3.5 ± 1.5 N·m (band: 2.0–5.0 N·m). All three failures are at 80 °C — runs 19, 20, and 25 — measuring ~5.38 N·m against the 5.0 N·m upper limit. Every other condition is within the band. Runs run07 and run18 are truncated before the held-point phase and cannot be evaluated for full-lock torque, so 25 of 27 runs are scored — 22 PASS / 3 FAIL.

The temperature-scaling model predicts ~4.46 N·m at 80 °C. The measured value of ~5.38 N·m is 20% above that prediction. The −10 °C conditions sit near 2.8 N·m — within band but on the low side.

On-center hysteresis — 2 FAILs, both at 25 °C / 13.5 V

On-center hysteresislimit ≤ 0.4 N·m

Only two runs exceed the limit, both at 25 °C / 13.5 V and both carrying operator notes (a rerun and a reported clunk). The other runs at that same condition pass, so this reads as run-specific, not a property of the actuator.

Measured within |HWA| < 0.5 deg. Limit: ≤ 0.4 N·m. Only run09 and run10 fail, both at 25 °C / 13.5 V, measuring ~0.65 N·m. All other rateA runs are within limit (0.17–0.39 N·m). Both failing runs carry operator notes.

On-center slope — all 18 rateA runs PASS (spec corrected)

The previous spec value of 0.014 N·m/deg was incorrect. The correct spec is 0.005 N·m/deg ±50 % (band: 0.0025–0.0075 N·m/deg). All 18 evaluated rateA runs measure 3.51–6.82 × 10⁻³ N·m/deg.

The prior version of this report showed 18 FAILs on this metric. Those were entirely an artefact of the incorrect spec value (0.014 instead of 0.005 N·m/deg). With the corrected spec, this metric is clean across the full campaign.

What the data shows

5 failures across 2 metrics. 4 of 6 metrics are clean campaign-wide.

FB torque at full lock — 3 FAILs at 80 °C. Runs 19, 20, 25 measure ~5.38 N·m vs upper limit 5.0 N·m. All other conditions pass.
On-center hysteresis — 2 FAILs at 25 °C / 13.5 V. Runs 09 and 10 measure ~0.65 N·m vs limit 0.4 N·m. Both have operator notes.
Steering ratio, motor torque, motor current, on-center slope — all PASS. Steering ratio is 15.000 ± 0.001 across all 27 runs.

Spec correction. On-center slope spec updated from 0.014 to 0.005 N·m/deg (±50 % tolerance). This change converts 18 previous FAILs to PASSes. 27 runs × 5 sources = 135 files.

All data in this series is synthetic, generated to mirror a real steer-by-wire validation program. No customer data is used.

Next in the series: the fault-injection campaign. If the week that disappears before the first plot sounds familiar, we would like to talk: founders@movedot.com, or www.movedot.ai.