Data-Driven Toe Misalignment Detection in Single-Unit Twin-Axle Trucks 2024-26-0278

Toe misalignment detection and its correction are important periodic tasks recommended by Original Equipment Manufacturers (OEMs) for Heavy Commercial Road Vehicles (HCRVs) to prevent premature tyre wear and improve fuel economy. Existing misalignment detection methods need skilled professionals to operate sophisticated equipment, while automated methods require additional sensors, which are not readily available in most trucks, making their implementation challenging. This study explores the effectiveness of a data-driven method to detect toe misalignment in single-unit twin-axle trucks with symmetric and asymmetric toe configurations. This method involves continuous monitoring of lateral dynamics variables measurable using sensors present in most trucks making it practically tractable. Ramp steer manoeuvres with a 45° steering angle and a rise time of 3 seconds in two directions (clockwise and anticlockwise) for two toe configurations (symmetric and asymmetric) constituted the four test manoeuvres. For each manoeuvre, a dataset of 192 number of feature vectors with different values of toe at the front axle wheels, containing an equal number of aligned and misaligned cases, was synthesised using IPG TruckMaker^®. Support Vector Classifier (SVC) models were trained on 63.00% and 67.50% of the datasets for symmetric and asymmetric toe configurations, respectively. The symmetric toe misalignment detection precision was 96.77%, 100.00%, with false negatives (FN) of 8.33%, 6.94% and false positives (FP) of 1.39%, 0.00% for the anticlockwise and clockwise steer manoeuvres, respectively. While the asymmetric toe misalignment detection precision was 92.31%, 78.95%, with FN of 30.65%, 25.81% and FP of 1.61%, 6.45% for the anticlockwise and clockwise steer manoeuvres, respectively. This approach indicated better effectiveness in detecting symmetric toe misalignment with lesser FN and FP percentages for both ramp manoeuvres when compared with asymmetric toe misalignment. This study’s outcomes are expected to contribute towards an onboard automated misalignment detection method, including thrust misalignment, to alert drivers in real-time.