| Aquaculture is a major industry in developing countries.The scale of aquaculture is constantly expanding every year,and the aquaculture industry has developed rapidly.The promotion and use of sensors has brought greater convenience to farmers,and intelligence has become a general trend in modern aquaculture.Dissolved oxygen is the most important parameter in water quality.The concentration of dissolved oxygen is related to the survival of aquatic organisms(such as fish,crabs,etc.).The dissolved oxygen-monitoring sensor communicates the dissolved oxygen concentration by generating a large amount of data during operation.Due to the complex water quality environment and the high intensity of the dissolved oxygen sensor,it will cause the sensors in the water to fail more and more frequently.As an important device for obtaining water environment information,the decreased accuracy of the dissolved oxygen sensor will lead to people not being able to obtain the actual concentration of dissolved oxygen in the water.Improper control of the dissolved oxygen content in a timely manner will result in a decline in food intake,slow growth of farmed animals in other cases,a large number of deaths will occur,and the losses will be very serious.The research on fault detection and fault-tolerant control of dissolved oxygen sensor has very important research significance and practical value.Therefore,how to use developed information technology to monitor sensors in real time to ensure the accuracy of measurement data,fault detection and compensation for water quality sensors used in aquaculture waters has become the focus of attention.Here,the polarographic sensor is used as the research object,and the principles and technologies of deep learning are used to perform fault diagnosis on the data collected by the sensor.This method can automatically complete the feature extraction and fault identification of fault data.The main work here is as follows:(1)By analyzing the working principle of the polarographic dissolved oxygen sensor,this paper explores the failure rule of the polarographic dissolved oxygen sensor under the influence of complex environment,including the types of faults that may occur in the sensor,the characteristics of the fault and the causes of various types of faults.Studying the failure modes of polarographic sensors can provide support for fault diagnosis models.(2)By comparing various commonly used basic models for deep learning,combining the data characteristics of polarographic dissolved oxygen sensors,eventually Convolutional Neural Networks(CNN)was finally determined as the fault diagnosis model of the sensor.In order to improve the detection performance of the convolutional neural network,the data was pre-processed,and then the influence of the convolutional neural network with different hidden layers on the detection performance was studied in combination with the actual training set data.Finally,a convolutional neural network model with eight layers was proposed to diagnose the data of the dissolved oxygen sensor.(3)The actual collected data is tested to verify that the recognition rate of the model reaches more than 98%.The results show that this method is feasible and achieves good results.Compared with the traditional diagnosis method based on feature extraction and classifier,it is verified that the method has better detection effect. |
