The Health State Assessment of the Benzene-Ethylene Ratio Control System Based on STA-BiGRU

The health state of the Benzene-Ethylene ratio control system is critical for ensuring product quality and improving overall efficiency. However, due to complex structure, hierarchical interactions and loop coupling, it is challenging to obtain the true health state of the system. In order to overcome this problem, a spatiotemporal attention-based Bidirectional Gated Recurrent Unit(STA-BiGRU) is proposed to assessment the health state of the Benzene-Ethylene ratio control system. First, the Spatial Attention and BiGRU are embedded in the encoder to identify the important process variables at each time step related to the health state. Then, the Temporal Attention and BiGRU are embedded in the decoder to adaptively discover the health-related hidden states all time steps. The combination of spatiotemporal attention and the BiGRU using information from both the past and future to capture the health-related process variables and time points, thus constructing a feature extraction model. Subsequently, a state recognition model is constructed by SoftMax classifier to obtain the health state level. Finally, the effectiveness and practicality of the proposed is demonstrated on the Benzene-Ethylene ratio control system platform.