On-line Detection Of Fiber Bundle Broadening Quality Based On Machine Vision

Due to the advantages of high performance,carbon fiber has been widely used in aerospace,military industry,and new energy.Carbon fiber beam widening is a key technology of the carbon fiber manufacturing industry.The role of this technology is to widen the carbon fiber beam to become thinner.The width and non-defective width beam are the quality assurance of subsequent pre-immersed materials.Because the carbon fiber beam is affected by factors such as production technology and equipment in the width,the widening is not uniform,and defects such as splitting,fibrous damage,and twisting will also occur,which will not guarantee the width quality of the carbon fiber beam.At present,many domestic companies' detection of carbon fiber beam widening qu ality depends on human eyes to identify and detect.The error rate,high labor intensity,and subjective influence of subjective factors.Therefore,it is urgent to use machine vision technology to conduct a testing of carbon fiber beam widening quality.This article designs a visual detection platform for carbon fiber beam widening quality.It develops a software system with carbon fiber beam widening quality with the width and uniform measurement algorithm of carbon fiber beams and the width of the extension detection algorithm.The main points of this article are as follows:(1)Clarify the detection requirements of carbon fiber beam widening quality,according to the edge shape and physical characteristics of the carbon fiber beam,the camera and optical lens of the image collection module,the light source and the light source controller of the light source controller,the motor and motion control module of the image collection module,the motor and motion control module and the motor and module and the motor and module of the module.Motor controller and other specific parameters and performance analysis of hardware,complete the selection,determine the lighting method,and complete the hardware construction of the platform in the laboratory to obtain a carbon fiber beam image with a certain speed.(2)Analyze the image characteristics of carbon fiber beam widening,and propose a measuring algorithm with carbon fiber beam widening uniformity.First of all,we use Open CV to achieve the calibration of came ra calibration and pixels;compare the study of image filtering and image enhancement algorithm,and use Gaussian filtering forces and gamma transformation operators with outstanding characteristics of the edge of the beam edge edge.The image is filtered and exfoliated and enhanced the image;after comparison of the processing effect of the traditional edge detection algorithm,the Canny operator is used to extract the complete line edge of the beam to perform pixel-level edge detection and positioning;use morphology processing to complete the complete extraction of the beam trunk;Finally,the improved Zernike Merglem pixel edge detection algorithm is used to obtain the fine edge.Through the experimental measurement,the measurement results of the above method meet the requirements of 0.2mm detection accuracy of the beam width.(3)Detecting algorithms for carbon fiber silk beams to design YOLOV5 models.In response to the imbalance of the proportion of width deficiencies,the paper use the enhancement method is proposed to prevent the model from overfitting;Tune and training;compare the four different algorithms of the YOLOV5 series,the accuracy of the optimal YOLOV5 X model trained in this article is higher than the other three algorithms;the final t est verification results are: 100% The detection recognition rate was 97.1%,the detection recognition rate of the twist defect was 98.8%,and the average image took 25 ms each image.(4)According to the detection process and testing requirements of the quality of carbon fiber beam,the data and information interaction of each module is designed.Quality detection software system provides a online monitoring visual software platform for carbon fiber beam widening quality detection.Finally,multiple groups of experimental design and analysis were performed to verify that meetting the needs of the width uniformity measurement and the detection of the detection of width defect detection at 0.23m/s.