Research On Industrial Heavy Metal Pollution Detection Model Base On Machine Learning And LIBS

While the development of the manufacturing industry has brought people a high-quality life,it has also brought serious industrial pollution.Because industrial pollution has endangered human health,corresponding measures must be taken to deal with this problem.All kinds of environmental pollution caused by industrial pollution will eventually evolve into soil pollution through ecological cycles,and even lead to soil pollution with heavy metals.Land is the main source of human food,so it is urgent to repair soil pollution.Using LIBS technology for pollution monitoring has unique advantages such as fast speed and almost no pretreatment.Deep learning is currently in a period of rapid development and has the advantage of being able to quickly extract data features.This paper combines LIBS technology with deep learning to construct a mechanical device for real-time monitoring of pollution in the manufacturing industry,and research on real-time monitoring methods for pollution in the manufacturing industry.The main important content of this article includes four parts:First,the design of mechanical devices for pollution monitoring in the manufacturing industry.Introduce the design of the mechanical device for pollution monitoring in the manufacturing industry,including the design of the optical path part,the spectrum collection part and the sample stage,and the function,working principle and related parametersettings of each part of the experimental instrument.Second,research on improving the quality of pollution monitoring signals in the manufacturing industry.Aiming at the errors in the LIBS spectra obtained through experiments,in order to reduce the impact of the errors,a combination of rough penalty smoothing noise reduction and segmented baseline correction is proposed to improve the quality of pollution monitoring signals in the manufacturing industry.By comparing with S-G smoothing and wavelet denoising methods,the results prove that the method has the best improvement effect and the highest recognition accuracy.Third,improve the manufacturing industry pollution monitoring combined model of AdaBoost algorithm.In view of the instability of the single classifier,the classification results of the algorithm are not good.In order to improve the prediction results of the classification and make the prediction results more accurate,the combined model is studied and an improved AdaBoost algorithm is proposed for the manufacturing industry pollution monitoring combination model.On the basis of the original fixed weight,a variable weight is added,and the fixed weight and the variable weight are combined as the final voting weight value.Through comparative analysis,the results prove that the improved AdaBoost algorithm combination model has better classification results and good stability.Fourth,research on pollution monitoring in manufacturing industry based on CNN-SVM classification algorithm.The processing method for LIBS spectrum is traditional machine learning algorithm,which is rarely combined with deep learning.Therefore,it is proposed to combine LIBS technology and deep learning,and propose a CNN-SVM classification algorithm model,using CNN as the feature extractor and SVM as the Classifier,to conduct experimental research on it.Through comparative analysis,the results prove that the research on pollution monitoring in the manufacturing industry based on the CNN-SVM classification algorithm is feasible and the classification results are good.