The Real-time Detection Of Valve Components Assembly Quality Based On Machine Vision

As an important part of the engine valve train, valve’s assembly quality directly affectsthe integral performance and service life of the engine. Currently, the valve assembly workmostly completed in the semi-automated assembly line, the assembly quality is mainlydependent on human eyes. The traditional inspection method that has some problems on thelarge quantities of continuous production lines,such as labor-intensive, low efficiency etc.It’salso can not be met the stability and reliability requirements.With the title “The Real-timeDetection of Valve Assembly Quality Based On Machine Vision”,the thesis systematicallyanalyses the key technology of the valve assembly quality detection,such as image acquisition,image processing and algorithm optimization. An automated quality inspection system wasdesigned which was used for valve’s assembly quality based on machine vision.THSI paper introduce domestic and foreign automotive equipment manufacturingindustry background and machine vision technology in industrial detection research and itsapplication.In view of valve components assembly quality detecting automation level isrelatively backward, the author put forward a method of the valve components assemblyquality detection based on machine vision technology.Considering the actual conditions ofproduction lines, equipped with vision inspection system in different assembly stations todetect the assembly quality of valve components.The main work include the followingparts：1）Firstly,thesis proposed general solution of embedded vision inspection system on thebasis of demands analysis.According to the detecting requirements to complete visualinspection system hardware selection, including cameras, lenses, lighting and input andoutput modules and so on. Thesis put forward a solution to completed the set up of hardwareplatform.2）In the image processing algorithm design,image contrast for valve assembly and thepresence of noise problems,image enhancement and image filtering algorithms are introducedand analyzed. After comparing different methods in each algorithm, gray stretch and gaussianfiltering algorithms were chosen for system.Image feature location，choose PatMax templatelocalization algorithm based on geometric shapes. Design image threshold segmentation, edgedetection and color extraction processing in view of the characteristics of valve componentsimage.3）Design the valve components image processing algorithms in the In-Sight softwareplatform. Through the production of standard parts, verify the stability and detection accuracy of the visual system, and analysis for the reason of misjudgment occurred during the detectionprocess.The vision inspection system has been formally put into production.After a long run toverify, the vision system is stable and reliable operation. The detection accuracy rate reached99.7%, and the rest of the indicators are reached corporate’s expectations. Experimentalresults show that the use of machine vision technology can be quickly and accurately realizethe detection of the valve components assembly quality, and has important significance forimproving the degree of automation of production.