In this article, we experimentally evaluate three flash and test strategies of automotive ECUs during serial production. These strategies are: (i) a test-oriented firmware in Virgin Mode, (ii) a dual-bank parallel flash strategy, and (iii) a test that is implemented in the production firmware. The latter could not be evaluated experimentally due to cybersecurity concerns. In this section, we employ 506 flashing cycles (actual production cycles and synthetic cycles from an independent reproduction), and report process time, process jitter, and p-value. In particular, we find that the dual-bank scheme provides the best performance with a mean flashing time of 23.3 s which is a 45% improvement over Virgin Mode. The variance was reduced by a factor of 25, showing that the reduced cycle times were appropriate for a takt time sensitive environment. All differences were statistically important (p −5). The data can be used as a basis to aid the decision making on flashing strategies in the automotive industry, and to define the next generation of ECU manufacturing processes considering process efficiency, cyber security limitations, and EoL procedures robustness.
M. Horák (Thu,) studied this question.