2. THOR-M instrumentation and sensor suite

The THOR-M dummy is only as useful as the measurement chain that surrounds it. Its biomechanical fidelity — the geometry of the head, the compliance of the thorax, the kinematics of the knee sliders — matters only if the resulting motion and loading can be captured with sufficient bandwidth, range, and channel integrity to reconstruct injury-relevant quantities. This section describes the sensor suite that turns the physical dummy into a quantitative instrument: where the accelerometers, load cells, and potentiometers sit, what each measures, how the signals are conditioned and digitized, and how to verify that the acquisition chain is trustworthy before a sled or vehicle test begins.

Coordinate Convention and Sign Discipline

Before discussing individual sensors, fix the reference frame. THOR-M follows the SAE J211 sign convention for automotive crash testing: the x-axis is forward (positive toward the front of the vehicle), the y-axis is lateral (positive to the dummy's left), and the z-axis is vertical (positive upward). Rotations follow the right-hand rule about these axes. Every accelerometer, load cell, and potentiometer channel in the data set must be documented with its axis orientation and polarity, because a single inverted channel — for example, a chest accelerometer mounted with its sensitive axis reversed — will silently corrupt the injury criteria computed downstream. The convention is stated once here and used consistently throughout the course and in the figure below.

Accelerometer Locations and Roles

THOR-M carries triaxial accelerometer clusters at the head, chest, and pelvis, plus additional single- or multi-axis units depending on the build configuration. The head cluster is mounted at the head center of gravity and provides the linear accelerations used to compute Head Injury Criterion (HIC) and the 3 ms clip. The chest cluster sits on the anterior thorax and supplies the accelerations that, together with chest deflection, feed thoracic injury assessment. The pelvis cluster captures the occupant's lower-body response and is essential for interpreting submarining and belt-to-pelvis load transfer.

Typical specifications for these channels are a measurement range of ±500 g to ±2000 g for head and chest, and ±200 g to ±500 g for pelvis, with a frequency response flat to several kilohertz. The practical consequence is that the accelerometers are not the bandwidth bottleneck; the acquisition system and the mechanical mounting are. A loosely torqued accelerometer block will produce mounting resonance that appears as high-frequency content in the 1–3 kHz band and inflates HIC. Mounting torque, thread locking, and cable strain relief are therefore part of the measurement, not an afterthought.

Load Cells: Belt, Neck, and Structural Channels

Load cells in THOR-M measure forces where the dummy interacts with the restraint system and where structural loads are diagnostically useful. The most consequential are the belt load cells — shoulder belt and lap belt — which quantify how much restraint force the belt delivers to the dummy. These are typically tension load cells inserted in line with the webbing, with ranges on the order of 10–25 kN and a calibrated sensitivity expressed in mV/V. The upper neck load cell measures the axial force, shear, and bending moment transmitted between head and torso, and is the basis for neck injury criteria such as Nij. Additional load cells may be present at the lower neck, in the lumbar spine, or at structural interfaces depending on the test objective.

Load cell signals are small — a few millivolts at full scale — and are therefore the most vulnerable part of the chain. They require stable excitation (typically 5–10 V DC), short shielded leads, and a bridge completion and amplifier stage with low noise. A drifting zero or a slowly decaying offset after the test usually indicates moisture ingress, a damaged cable, or thermal drift in the amplifier, not a real mechanical event.

Potentiometers and Deflection Measurement

THOR-M uses potentiometers and rotary sensors to measure chest deflection and, in some configurations, knee slider displacement and pelvis rotation. The chest deflection potentiometer is a spring-loaded string or rotary device that converts the sternum-to-spine displacement into a voltage proportional to deflection, typically calibrated in millimeters. Because chest deflection is a primary input to thoracic injury criteria, its calibration and zeroing are critical: the sensor must be set to a known reference position with the dummy in its nominal seated posture, and the full-scale range must comfortably exceed the expected deflection (often 0–100 mm) without mechanical bottoming.

Potentiometers are ratiometric devices: their output scales with the excitation voltage. If the data acquisition system uses a different excitation than the calibration reference, the deflection will be scaled incorrectly. Always record the excitation voltage used during the test and confirm it matches the calibration certificate.

Data Acquisition Channels and Signal Conditioning

The acquisition system must handle three signal families simultaneously: high-level, high-bandwidth accelerometer outputs (single-ended or differential, ±5 V or ±10 V full scale); low-level bridge outputs from load cells (mV/V); and ratiometric potentiometer outputs. A practical channel budget for a full THOR-M test is on the order of 40–60 channels, sampled synchronously at 10–20 kHz with at least 16-bit resolution and anti-alias filtering appropriate to the channel type. SAE J211 recommends channel filter classes (CFC 60, CFC 180, CFC 600, CFC 1000) matched to the physical quantity; accelerometers typically use CFC 1000, load cells CFC 600, and deflection channels CFC 60 or 180. Applying the wrong filter class changes the computed injury values, so the filter must be documented per channel.

All channels must share a common time base and a common trigger. A hardware trigger derived from the sled or barrier event, distributed to every acquisition module, prevents the channel-to-channel timing skew that would otherwise distort cross-correlations between, for example, belt force and chest acceleration.

Channel Map and Pre-Test Verification

The following channel map is a representative example for a THOR-M frontal sled test; actual channel numbers and ranges depend on the dummy build and the laboratory's acquisition hardware. Treat the values as illustrative configuration, not as measured results.

Verification follows a fixed sequence. First, confirm the channel map against the physical dummy: every cable labeled, every axis orientation checked against the SAE frame, every polarity confirmed with a tap test or a known static input. Second, zero all channels with the dummy in its nominal posture and record the offsets. Third, run shunt calibration on load cells and known-input checks on accelerometers and potentiometers, and log the results. Fourth, arm the trigger and confirm that all modules report a common time base. Only then is the test ready to run.

Common Failure Modes and How to Recognize Them

Three failure patterns account for most bad THOR-M data sets. Inverted polarity on an accelerometer or load cell produces a time history that looks plausible in isolation but yields a negative belt force or a chest acceleration that opposes the sled pulse; the tap test during setup is the defense. Excitation mismatch on a potentiometer or load cell scales the entire channel by a constant factor, which is invisible in the waveform but wrong in the injury metric; recording the excitation voltage and comparing it to the calibration certificate catches it. Ground loops and cable microphonics appear as high-frequency noise or baseline wander that is absent in the shunt calibration but present during the test; shielded twisted-pair cabling, a single-point ground reference, and strain-relieved cable routing are the standard mitigations.

When a channel fails verification, the correct response is to isolate the fault before the test, not to compensate after it. Swap the cable, re-seat the connector, re-run the shunt calibration, and confirm the expected step. A channel that cannot be brought within tolerance should be flagged and, if it is injury-critical, the test should be postponed rather than run with a known-bad measurement.

The sensor suite described here is the measurement foundation for everything that follows in this course. Restraint optimization, injury-criterion computation, and test-to-test comparison all depend on a channel map that is correct, a signal chain that is calibrated, and a data set whose timing and polarity are trustworthy. Establishing that foundation deliberately — before the first sled run — is what separates a defensible test program from a collection of waveforms.