Estimation of Passenger Car Brake Pad Wear and Remaining Useful Life From On-Road Usage Data Considering Variable Vehicle Mass

The braking system of a car is essential to ensure the safety of its passengers and surroundings. Cars can possess significant kinetic energy, attributed to their high mass and ability to travel at high speeds. The possibility of frequent and severe applications of the brakes means that the braking system must be capable of dissipating such energy reliably and repeatedly. The braking action of cars is primarily provided by friction brakes, either drum or disc brakes, with the latter being the most common on modern cars. This is due to the higher tolerance for high temperatures and more linear and predictable operation. A disc brake consists of a disc mounted on a rotating wheel hub, clamped by hydraulically actuated brake pads. Once actuated, the pads press against the rotating disc, generating a friction torque and converting the kinetic energy into heat.

This braking action, however, results in wear of the friction components. Frequent physical inspections are thus necessary to ensure the components remain within specifications. The useful life of brake pads is sometimes expressed in terms of distance travelled, providing only a rough estimate, as the wear is highly dependent on how the car is driven, e.g., highway driving with long distances and infrequent braking or city driving with more frequent braking over shorter distances. Other factors, such as road gradients and the presence of additional vehicle loads, will also affect the wear rate. Depending on the vehicle, extra load can constitute a significant part of the total vehicle mass, e.g., a small hatchback weighing around 1,100 kg, being loaded with four passengers and luggage.

Some brake pads are equipped with wear sensors, triggering an alarm once a certain thickness has been reached. Sensors capable of measuring the actual wear can also be employed, but the harsh operating environment makes such sensors expensive and typically reserved only for larger heavy-duty vehicles. Access to the brake wear can allow for condition-based maintenance, providing indications of the current state of the brake pads. Furthermore, estimation of the remaining useful life can allow for improved maintenance planning through predictive maintenance. In the case of fleet management, where a large number of vehicles is typically considered, condition-based and predictive maintenance can aid in obtaining a more efficient use of maintenance resources.

The overall topic of this thesis is the problem of estimating passenger car brake pad wear, based on usage data from vehicle mounted sensors, including an inertial measurement unit, global navigation satellite system, and vehicle on-board diagnostics data. A vehicle longitudinal dynamics model accounts for effects such as aerodynamic drag and road gradients. Additional vehicle loads are considered through real-time mass estimation. The proposed methods are validated using onroad data from passenger cars, considering a total of almost 20,000 km of driving across three different vehicles.

The thesis is based on five original journal papers on mass estimation [P0, P0, P0] and wear estimation [P0, P0]. In prior works considering vehicle mass estimation, the proposed methods typically require manufacturer-specific controller area network (CAN-bus) parameters, e.g., related to engine torque and current gear. In [P0] vehicle mass estimation is carried out based on standard vehicle parameters, available on all modern passenger cars. Gear and engine torque are estimated from the available data, and validation is performed using two different cars. The same CAN-bus data constraint is considered in [P0], but with the addition of satellite data. With this inclusion, an extended Kalman filter based sensor fusion is employed, providing accurate estimates of vehicle position, speed, and attitude angles. An experimental investigation is carried out, considering the sensitivity of the mass estimates to variations in tire pressures, window openings, and ambient winds. To investigate the possibility of reducing instrumentation requirements, mass estimation without reliance on any vehicle data is covered in [P0]. A novel approach to gear estimation based on spectral data of the vehicle acceleration is proposed. Due to the lack of engine torque information, vehicle mass is only accurately estimated during full engine load, based on engine torque curves estimated from prior training data. Improvements to road gradient estimates based on vehicle pitch angles are also proposed, by compensating the pitch angles for pitching from vehicle loads and inertial forces, resulting in improvements to the vehicle mass estimates.

In [P0] the estimation of brake pad wear is considered, proposing a wear model based on an estimated brake work. The brake work is defined as the product of distance travelled by the vehicle during braking and the longitudinal braking force, estimated from the vehicle model. Validation of wear coefficients estimated from on-road data is carried out with a pin-on-disc laboratory test-rig. The final paper [P0] combines all the previous works, covering brake pad wear and RUL estimation, while also considering real-time mass estimation to account for varying vehicle loads. A total of 31 brake pad wear measurements over 10,200 km of driving is considered.