Reliability is one of the properties that customers of heavy trucks value highest.Dependent on all parts and functions of the vehicle, reliability is a complexproperty, which can normally be measured only towards the end of a developmentproject. At earlier development stages, forecasts can give valuable decision supportfor project planning.The main function of a heavy truck is to transport goods, but the truck also hasinteractive functions as the working environment of the driver. Interactivefunctions are functions experienced by the driver. They are subjective, in the senseof being person dependent, so that a system can be experienced as inadequate byone user but satisfactory by another. Examples of interactive functions of heavytrucks are climate comfort and ergonomics, which are experienced differently bydifferent drivers. Failures of these functions lead to costs and limited availabilityfor the customer. Therefore it is important to include them in reliability forecasts.The work described in this thesis concerns some elements of the system reliabilityforecast. Two studies are presented, one proposing a qualitative systemarchitecture model and the other reviewing and testing methods for evaluating theimpact of varying operating conditions. Two case studies of a truck cab in a systemreliability test were made. The first case study shows that the system architecturemodel supports reliability forecasts by including interactive functions as well asexternal factors, human and environmental, which affect function performance.The second case study shows that modelling uncertainty is crucial for interactivefunctions and recommends a method to forecast function performance while takingvarying operating conditions into account.

Heavy truck customers attach great importance to reliability, which make reliability assessments essential in product development projects. Since changes are easier and less expensive in early project stages, early reliability assessments are valuable. At these early stages, complete vehicle testing cannot yet be made. System reliability assessments must be made based on test data from component and subsystem tests, sometimes performed with different operating conditions than the system will be used in. Test data must be translated to the new situation, which requires information about how various factors affect reliability. Furthermore, the uncertainty in the forecast increases when the assessment is made for new operating conditions. Therefore, we also need information about how uncertainty propagates. The question is how this translation can be made, when data is sparse and expert judgement must be used, and how the increasing uncertainty can be reasonably modelled. In this paper, current methods to take into account varying operating conditions have been reviewed, and four methods have been tested in a case study. These methods are one based on fuzzy logic, a first-order second-moment reliability method (VMEA), and two variants of the proportional hazards model. The study shows that several methods are capable of handling sparse data, but only VMEA can model how uncertainty increases when operating conditions vary. It has however the drawback of being quite sensitive to uncertainty in the input data.