Intelligent Fault Diagnosis of Consumer Electronics Sensor in IoE via Transformer

IoE era is coming with the development of information and communication technology. As a typical representative of the IoE intelligent era, consumer electronics products have maintained a momentum of rapid development. A variety of complex sensors are built into consumer electronics. Due to the influence of temperature, humidity, vibration, impact, and other factors, sensors in consumer electronic products are prone to failure. To ensure the normal operation of sensors, it is necessary to study sensor fault diagnosis methods. This work proposes a transformer-based consumer electronic sensor fault diagnosis network (CESFDNet). First, CESFDNet utilizes multi-layer convolution to extract the local correlation information among adjacent data, then fuses it into the global dependency information in transformer. Second, CESFDNet improves self-attention mechanism of transformer. This reduces the impact of noise in the sequence on the diagnostic results. Thirdly, CESFDNet combines absolute and relative position coding information to improve transformer. This is conducive to paying attention to the absolute position relationship and relative position relationship between sensor data. Fourthly, IN standardization and GeLU activation are embedded to optimize the model. Fifthly, this work conducted systematic and comprehensive experiments on CESFDNet. The experimental results fully prove the superiority and reliability of CESFDNet.