Reliability analysis regarding product fleets in use phase: Multivariate cluster analytics and risk prognosis based on operating data

The increasing complexity of product functionality and manufacturing process parameters often leads to complex failure modes and reliability problems within the product life cycle. Especially in the case of mass production of consumer goods - e.g. automobiles, washing machines, computer - an increasing percentage of damaged products within the product fleet can lead to garage or recall actions. If the manufacturer receives knowledge about the first damage claims based on a field observation, a risk probability prognosis is the base of operations regarding further actions. State of the art concerning risk calculation methods consider the failure behaviour and allow the univariate determination of the risk probability regarding the product fleet. These methods do not consider the load or usage profile of the products based on any life span variable. In fact, current technical complex products save a lot of life data (“Big Data”), which can be additionally used for risk analysis within product fleets. This paper outlines an approach to determine the risk probability in product fleets based on a combined multivariate analysis of the product failure behaviour and the customer product usage profile. The theory and application of the approach is shown with the help of a synthetic data set within an automotive case study, which includes real effects of typical field failure behaviour and usage profiles of an automobile fleet.