In the recent years, deep learning has been widely used in process monitoring due to its strong ability to extract features. However, with the increasing layers of the deep network, the compression of features by the deep model will lead to the loss of some valuable information and affect the model’s performance. To solve this problem, a fault detection method based on a discriminant enhanced stacked auto-encoder is proposed. An enhanced stacked auto-encoder network structure is designed, and the original data is added to each hidden layer in the model pre-training process to solve the problem of information loss in the feature extraction process. Then the self-encoding network is combined with spectral regression kernel discriminant analysis. The fault category information is introduced into the features to optimize the features and enhance the discrimination of the extracted features. The Euclidean distance is used for fault detection based on the extracted features. From the Tennessee Eastman process experiment, it can be found that the detection accuracy of this method is about 9.4% higher than that of the traditional stacked auto-encoder method.

Intelligent algorithm has been widely implemented to effectively diagnose faults in industrial instrument, electrical equipment and mechanical equipment. In addition, the rapid development of sensing technology generated enormous time series signal. Accordingly, diagnosing faults by analyzing time series signal has been widely developed. This paper aims to diagnose faults by applying improved Convolution Neural Network with Compression Enhancement (ICECNN-1D) to analyze time series signal, which effectively considers time series property of signal while diagnosing faults by artificial intelligence. Additionally, a large number of trend features and fluctuation features in high-frequency time series are also considered. the recognition rates of almost other machine learning algorithm are less than 90% in the experiments. Other methods may provide high rate of recognition, but their fluctuation of the recognition rate has varied obviously with different loads, and results provide undesirable ability of generalization under different working conditions. Comparatively, ICECNN-1D model provides high recognition rate and terrific ability of generation while processing time series with high frequency, and its accuracy of the recognition rate fluctuates inconspicuously with different loads.

Permanent magnet metal flowmeters are implemented to monitor the flow of liquid metal coolant in a nuclear reactor, whose measurement accuracy plays a significant role to ensure the safety and normal operation of the nuclear reactor. According to a theoretical analysis of the permanent magnet metal flowmeter, several factors such as temperature, nonlinear degree, and zero potential will affect the accuracy of measuring the flow. However, for a heavy-caliber permanent magnet metal flowmeter, the influence of the magnetic Reynolds number provides obvious nonlinearity, which affects the measurement accuracy of the flowmeter. Consequently, we use the finite element method to calculate the magnetic field of the flowmeter and analyze the cause of the nonlinearity. Additionally, the influence of nonlinear error is significantly reduced by designing the structure of the flowmeter and the appropriate arrangement of the electrodes.