Fault Diagnosis Based on Prior Knowledge for Train Air-Conditioning Unit

The fault diagnosis of train air-conditioning unit is becoming extremely necessity because train is often occupied with many passengers for a long time. However, limited diagnostic accuracy has become a bottleneck of train air-conditioning unit fault diagnosis. In the paper, two new fault diagnosis methods based on prior knowledge (PK) for train air-conditioning unit was proposed. First of all, taking KLD-29 as an example, according to distribution characteristic of data samples from data acquisition scheme on train air-conditioning unit 6 super spheres were constructed. Secondly, every super sphere was incorporated to its responding optimization problem as constraint and six diagnosis models were obtained. At last, we diagnose the fault of train air-conditioning unit by 1-v-6 diagnostic scheme based on the six models. In experiment, we chose a baseline method and the proposed methods as comparisons. Experimental results showed the PSSVM-based ESOP scheme is more appropriate for single-label failure while for fault diagnosis of multi-label, 1-v-6 PSSVM-based scheme is more suitable than 1-v-6 PSSVM-based ESOP scheme.