Study On Microwave Detection Of Water Content In Textile Based On Machine Learning And Ultra-wideband Antenna

In the textile industry,the moisture content is a very important parameter for the performance of textiles,and it runs through every process in textile production.The moisture content of the fabric is extremely closely related to the stability of the various processes in the production of the fabric.Therefore,research on fabric moisture detection technology featuring non-contact,non-destructive,real-time,and high measurement accuracy is of great significance in improving the ability of the production line to integrate with the Internet,achieving intelligent factories,and accelerating the construction of textile networking.Microwave moisture detection is a widely used method using microwave technology to detect the moisture content.Compared with the conventional fabric humidity detection methods,microwave detection has the characteristics of real-time,online and non-destructive,and has attracted more and more attention of researchers.However,there are many problems in the existing research.First of all,in the study of the moisture content of textiles by microwave method,in order to obtain high penetration signal power and minimize multipath effect,the shape of the tested sample is limited.And,most of the current microwave methods use a single frequency point,and the model relationship between the moisture content and the dielectric value limits the water content value that can be detected to a rather narrow value.In addition,the research at this stage is only for one type of textiles,and for the moisture detection of textiles of different material and size parameters,an additional model needs to be established.This requires a lot of time in real applications.In this article,we propose to combine ultra-wideband antennas with support vector machines and deep belief network to solve the existing problems in the moisture detection of the textile microwave.Based on the study of the related principles of the dielectric properties of the cone yarn and the microwave detection,the overall structure of the fabric moisture microwave detection system is firstly designed.The detection system mainly includes microwave signal source,ultra-wideband antenna,signal acquisition module and the fabric detection data processing module.Then the fabric moisture regression algorithm was designed.The model consists of four modules: data preprocessing module,feature extraction module,training model and model optimization.In order to break the limitation of the size and moisture limits of the tested samples and realize the moisture regression of textiles with different size and material parameters,a moisture regression algorithm based on the UWB-SVR model is proposed.The model is based on Standard Support Vector Regression(?-SVR)and Ultra-Wideband Antenna(UWB)systems to detect the moisture of cone yarn that are widely used in the textile industry in a wide moisture content range.Compared with the single frequency antenna used in the traditional microwave moisture detection,UWB has a wide operating frequency band and better performance.Therefore,the UWB antenna can acquire enough fabric moisture feature data to create a textile moisture model in combination with a machine learning algorithm to realize moisture regression of the fabric in a wide moisture range without being limited by the size of the test sample.However,the moisture regression algorithm based on UWB-SVR model for textiles has a relatively large error in the individual test samples,and the error in the combined moisture regression model established for textiles of different sizes and material parameters is too large.It is not suitable for joint moisture detection and control of various textile raw materials and finished products.Therefore,we also design the moisture regression algorithm based on ultra-wideband antenna(UWB)and deep belief network(DBN).This algorithm is used to overcome overcomes the disadvantages of the microwave moisture regression algorithm based on UWB-SVR model,and improves the stability of moisture regression.Subsequently,this algorithm was used to carry out moisture regression analysis studies on textile size and material parameters independence.This paper validates the algorithm through simulation and experimental methods.We first set up a simulation system using CST Microwave Studio to perform simulation of cone yarn moisture and obtain simulation data.Then,according to the simulation system,a textile moisture microwave test platform was set up using Ultra-wideband antennas,vector network analyzers,turntables,wave absorbing materials,etc.After data collection,multiple linear regression,BP neural network,UWBSVR model-based moisture regression algorithm,and UWB-DBN model-based moisture regression algorithm were used for data analysis,and the results were compared.The results show that the moisture regression algorithm based on the UWB-SVR model accurately achieves the moisture regression of the cone yarn,and solves the defects in the existing methods that limit the size and moisture range of the test sample.The UWB-DBN model based on the moisture regression algorithm more accurate realization of the different size of the material cone yarn joint moisture regression.The moisture regression algorithm based on the UWB-DBN model achieves more accurate regression of single material and single-sized yarn moisture.It is accurate to realize the combined moisture regression of different size and material yarn.It lays the foundation for reducing the modeling work of different sizes and material parameters,and expands the moisture detection based on the microwave.Finally,summarize the full text and discuss the future work.Explain the application advantages and potential of microwave moisture detection in the upcoming textile networking,and combine with other excellent deep learning methods to achieve more accurate and efficient textile moisture detection.