Application And Research Of Intelligent Algorithmsl In Industrial Data Preprocessing

In the industrial field,the accuracy of data is very important.Data acquisition and data transmission will be interfered by external environment,and the data will be partially missing,which causing losses to industrial production.Such type of data is called industrial interference missing data.According to the characteristics of missing data,this paper divides the industrial interference missing data into two types: strong time series missing data and high-precision missing data.For example,in the signal acquisition system,because of the large amount of acquisition,monitoring information is delayed,and even data loss errors occur.Such missing data is called strong time series missing data.While in the positioning system,because the coordinates of the positioning points need to be precise,the missing data caused by the hardware data acquisition and transmission can be called high-precision missing data.So it is necessary to pre-process the data that been received,and check if data loss errors occur,If it happens,the missing data should be supplemented to ensure the accuracy of the data.Therefore,data preprocessing is very important in the industrial field,and how to preprocess data and detect missing data and complete missing data is the focus of this paper.Firstly,this paper analyses a kind of strong time series data in the industrial field,and concludes that the strong time series data has the characteristics of periodicity and time series.According to the characteristics of periodicity,this paper decides to use the recurrent neural network(RNN)to establish a time series model to predict missing values.But standard RNN can neither capture good periods in time series nor deal with missing values.Therefore,attention mechanism is added to RNN structure in this paper.Its design purpose is to capture time and make RNN more robust,so as to capture missing values.By modifying the loss function,the convergence speed of the model was accelerated,and the cosine annealing algorithm was used to train the algorithm to improve the generalization ability of the model and accelerate the convergence speed.Then,through the analysis of high-precision data in industrial field,this paper decides to use kNN machine learning algorithm to build model to predict missing values.According to the characteristics of this kind of high-precision data,this paper decides to use kNN machine learning algorithm to build a model to predict missing values.However,it was found that the standard kNN can not meet the completion needs of high-precision missing data due to its long calculation period and abnormal reference points.In this paper,error filtering algorithm is used to determine the existence of missing values.And through the selection of K value(select k coordinate points near missing values)to reduce the operation time of kNN algorithm.The reference points are weighted to make the reference points which close to the missing values have higher weights and reduce the impact of outliers on the missing values.After completing the missing data,the accuracy can be improved in this way.After that,this paper carries out demand analysis,outline design,detailed design and system implementation of the ceramic power generation monitoring system and UWB wireless positioning system,in order to complete the research and development of the two systems.Finally,the actual experimental data were obtained through ceramic power generation monitoring system and UWB wireless positioning system.This paper compares the missing value completion algorithm with the traditional algorithm,evaluates the advantages and disadvantages of the algorithm,and put forward the improvement strategies.