Lifetime modelling in KEEPEMALIVE Sample Clauses

Lifetime modelling in KEEPEMALIVE. The establishment of a software tool for lifetime prediction has not been achieved in Task 5.4. The main reason for this is insufficient and inconclusive data for an establishment of a model. There are several aspects that contribute to this fact and this is discussed below. The development of new materials, and changes to new components within the project has rendered much of the data not comparable statistically. These changes have nevertheless been necessary in order to achieve sufficient durability. Especially a change to the reinforced membrane has improved durability. Significant changes to the test programme have also been required. Break-in requirements for the reinforced membrane changed drastically due to the finding that a lot of test data where middle of test performance was better than at beginning of test. Positive degradation rates are of course useless for lifetime estimation purposes. The use of experimental design for evaluation of stressing conditions is an excellent approach to quantifying the rates of the individual stressors, but also the possible interactions between stressors. The requirement of these 23 factorial designs is that all eight experiments need be completed in order to determine the interaction between effects. In KEEPEMALIVE, only a few complete sets have been completed. Statistical analysis of these sets, see D5.2 and 0 in this deliverable, have shown that the magnitude of the interactions are very often large and significant. An inherent challenge to parametric modelling of degradation is to relate individual degradation mechanism to lifetime. Published work has so far mainly focused on the effect of single stressors: Arrhenius for temperature, humidity transformation and fluoride emission rates to name a few. Multivariate relationships like Eyring have been applied in other applications than PEM fuel cells [10]. The challenge with application of Eyring to KEEPEMALIVE is that the data is not from discrete stressors but a multivariate combination, where interactions are not fully resolved. The problems associated with the execution of the AST protocols have been the ability to conduct experiments at some combinations of operational set points. This was remedied by introducing introductory scans through all combinations of set points before start of the accelerated stress test. It was further complicated by a change to reinforced membrane that required further revision of the protocol. It has proven to be challenging to...