This report presents a discussion of existing models for reliability evaluation of electronics systems in the development phase. The discussion is based on published literature and practical experiments. The focus of the product segment investigated has been harsh environment and it is concluded that at least a workmanship model for reliability testing has to be developed, since existing models all have some disadvantages.

A model based on test data (partly top-down similarities analyses) for reliability evaluation has been used on a demonstrator and showed the limitations of the model. A comparison between the funnel model used at Kitron Development AB for development of products using new technologies and existing models for reliability prediction is performed.

The current models for reliability prediction are based on either knowledge from previous products - field data (top-down similarities analyses); test data from environmental testing (partly top-down similarities analyses), stress and damage modelling (bottom-up physics of failure), or handbooks (bottom-up statistical). Top-down similarities analyses are regarded as a good approach. However, existing field data is limited and with the fast development of electronics hardware the elapsed time before the designer has the necessary data available is mostly too long.

In comparison, accelerated testing of prototypes (partly top-down similarities analyses) are faster. However, the cost and the problem to determine the correct acceleration factor are two of the more severe drawbacks that have to be dealt with before the method could be recommended.

Simulation might be a good way to overcome the problem with long lead-time and high cost. A simulation model needs a lot of input data, which can be hard to find unless experimental work is performed. Hence, simulation is not a perfect tool either.

Finally, the conclusion from the literature is that the era for handbooks like MIL-HDBK-217 is over. The literature is amazingly coherent in their conclusion that the standards should be withdrawn, exactly in the way that MIL-HDBK-217 has been.

A recommendation is therefore that a new methodology involving the combination of the methods mentioned should be developed.

Results from reliability implementation of microvia board and microvia board assembly is presented and discussed. The way of working during the implementation was similar to the testing methodology found in the literature. Application suitable tests were identified or developed during the implementation. Nevertheless, the fact that the projects lasted for several years confirmed the conclusion that accelerated testing of products at different development steps is time consuming and expensive.

Based on the overview and the practical work a possible way of working for design for reliability, i.e. virtual prototyping and qualification has been proposed. The proposed methodology will be verified in the final industrial PhD thesis within a time frame of two years from now. The fields that will be included in the thesis are: Failure analysis and relevant failure mechanism, Modelling, Experimental verification of the models, Optimisation, and Implementation in product.