Study on an Optimization Algorithm of Generating Test Vectors for Digital Reactor Protection System Testing

As a safety-critical system for the NPP, the digital Reactor Protection System (RPS) has replaced the traditional analog Reactor Protection System in the most newly-built NPPs. A new type digital RPS developed by INET, Tsinghua University, must pass the hardware qualification and software Verification & Validation (V&V) to satisfy the requirements of quality criterion and safety laws. The stimulation/response testing method is always used in the integration testing phase of software V&V. The test vectors group would be very large if the digital RPS has many input variables. Therefore, In order to find out all of the failure of software, the less testing vectors would be benefit to limit the testing time and cost. A black box model is always be used for those systems with few known information for the Conner. All testing vectors would be generated by nature sequence. The black box model has good features. It does not rely on any prior knowledge about the objective system. However, the black box model may increase the average number of test vectors and average time to find out all of the failure. If a grey box model can be adopted in the testing process, a lot of known information of the objective system can be used and the test time would be saved prominently. As independent developed digital RPS by INET, there is enough information of the testing objects, which can be used to apply the grey box model on the digital RPS testing procedure and to generate the test vectors. An optimization algorithm of test vectors generating is as follows: a) Firstly, a different weight factors would be set to different combination of input variables by expert knowledge and logic design rules; b) Secondly, a particle movement algorithm is used to optimize, compare and select random test vectors by weight factors. The primary simulation results indicate that the average testing time and the number of test vectors are both less than the normal test strategy which based on the black box model. The optimization algorithm of test vectors generating based on the particle movement may be more efficient to find out all of the failure. Therefore, the testing cost and time would be saved in consequence.