Advanced statistics improves brake pad manufacturing quality

The University of Manchester was the first to develop a new method of nonlinear statistical analysis which can be used to design efficient product development experiments. Global vehicle parts supplier, Federal-Mogul Friction Products, used this approach to reduce product testing by up to 88% and reduce variation in its braking products, currently used in a variety of mainstream and premium brand cars.

Banner image

Under any driving conditions, you expect your car brakes to work perfectly – every time. When our lives depend on the safety of technology, a consistent safety and quality standard is essential.

Federal-Mogul supplies an extensive line of automotive and heavy-duty original equipment and aftermarket brake components. As with all safety critical products, the company must be sure that its products demonstrate consistent high quality with minimal variation between and within batches. During the development phase of its latest braking products, Federal-Mogul’s R&D team began reaching the limits of off-the-shelf analysis software and asked Dr Alexander Donev of The University of Manchester for assistance. The goal was to maximise the accuracy of the statistical methodologies to predict and repeat the outcomes of product testing.

Rather than  follow the 'six sigma' approach used by most automotive suppliers, Dr Donev developed a new statistical methodology which would help Federal-Mogul to design physical experiments for testing new products in development, specifically disc brake pads. Comparative analysis between the testing regime and the company’s previous approach showed that it could obtain reliable estimates of variation from far fewer experimental trials. Indeed, in one case the experimental designs reduced the number of required tests by 88%.

The new experimental designs soon become standard procedures within the company’s braking products business and were first used to optimise the production of a new brake pad for a Ford pick-up truck; a vehicle with an annual volume of around 350,000 per year in the US since 2010. Importantly, Dr Donev's new statistical model identified the optimal time a brake pad should remain on a press; this led to a 30% reduction in the overall manufacturing time, generating a significant efficiency improvement for Federal-Mogul.

In addition to maximising production efficiency, the modelling work helped to minimise variability and maximise product quality by accurately identifying optimal manufacturing conditions.

Following the outstanding success of the Ford project, further brake products have been brought into production. In 2012, new disc brake pads were developed for key premium German and US brands – all using the new statistical model.

The major contributions of the research are:

The statistical software reduces the need for expensive experiments by 88%


The statistical software reduces the need for experiments by up to 88%.
Dr Donev’s new statistical model reduced manufacturing time by 30%.


Dr Donev's new statistical model reduced manufacturing time by 30%.


350,000 Ford pick-up trucks are sold annually in the US.
Growth infographic


By 2012, new pads were developed for key premium German and US brands.


Dr Alexander Donev
Dr Alexander Donev

Dr Alexander Donev and his team of PhD student researchers at The University of Manchester developed novel statistical methodologies to improve Federal-Mogul’s design of physical product testing experiments. The new approach took account of existing information from previous testing to evaluate sources of variability in the manufacturing process. The team also developed ways to display the results as graphs which improved the ease of use for Federal-Mogul’s braking Research and Development team.

The researchers developed a range of models to study products made from different materials. They were the first to adopt nonlinear models incorporating 'estimable flexible regressors' so the models could handle fast and localised events. These models are potentially applicable to a huge range of other industrial processes.


Lead Academic:
  • Dr Alexander Donev


PhD students:
  • Liam Brown
  • Sergio Loeza-Serrano
▲ Up to the top