Board-level Intelligent Built-in Test Design of Digital Input Module by Improved BP Neural Network

Abstract

The built-in test (BIT) technology plays a major role in heavy-duty gas turbine control system. The accuracy of BIT is a guarantee for the normal operation of the equipment. In the actual production process, conventional BIT has a relatively high false alarm rate. In this paper, an intelligent built-in test (BIT) design method was presented for digital input (DI) module using a back propagation (BP), which was optimized by improved adaptive gravitational search algorithm (IAGSA) for quantitative analysis. In order to improve defect of the basic gravitational search algorithm (GSA), a tent map for population chaos initialization was proposed. Firstly, the status information that can reflect the failure of the DI module was chosen as experimental data. Then, these data were handled by normalization. After that, multiple IAGSA-BP models were established with status information about DI module. The output of these models can fully represent the fault status of the DI module. Finally, the experimental results are shown to validate the effectiveness of the proposed approaches.