Research On Cylindrical Plug On Fan Casing Defect Detection System Based On Multi-sensor

The fan of the washing machine is assembled by the fan casing,rubber cylindrical plug,circuit board and stator,etc.The rubber cylindrical plug is inserted into the plunger hole of the fan casing by manual assembly,which is possibly missing,not in right place or assembly tilted.In addition,there may be defects in the end face of the rubber cylindrical plugs,these problems will directly the normal and smooth operation of fan products.If artificial visual inspection is used to inspect the end face defects and the assembly quality of cylindrical plugs,the detection efficiency and accuracy are difficult to meet the requirements.Therefore,in this paper,the research on the automatic detection of the end defect of the cylindrical plugs based on machine vision was carried out.At the same time,the detection method combining the ranging sensor and the optical sensor was used to solve the problem of automatic detection of the assembly quality of the cylindrical plug when the fan casing is tilted.The main research contents are as follows:(1)Aiming at the fan casing cylindrical plug detection system,designed the overall plan and divided the inspection stations.According to the different detection tasks of each station,the corresponding hardware was designed and selected,and the software process of the detection system and the communication method design between the detection system and the production line PLC system were completed.(2)Aiming at the problem of surface defect detection of small-size rubber cylindrical plug,a defect extraction algorithm based on super pixel segmentation was designed.First,Hough transform was performed on the cylindrical plug image to eliminate interference of complex background;then anisotropic filtering was used to remove interference noise such as dust and texture on the plunger surface;finally,super pixel segmentation algorithm was used to achieve accurate extraction of defect contours on plunger surface,which solved the problem that traditional threshold segmentation cannot completely extract the defect contours and reduced the amount of calculation and complexity for subsequent image processing.(3)Aiming at the multi-classification problem of cylindrical plug defects,a random forest classification algorithm based on grid search cross-validation method was proposed,which can accurately classify different types of defects by getting the best parameters of random forests adaptively.(4)A method for inspecting the quality of assembly of cylindrical plugs when the fan casing was tilted was proposed.First,the detection system automatically obtained the center coordinates of the cylindrical plug and the fan casing through the target recognition algorithm.Then,based on the abovementioned center coordinates,a set of detection sampling points was generated,and the shortest detection path was automatically planned by the simulated annealing algorithm.The system collected the height data of the cylindrical plug and the base surface of the fan casing through the detection path;Finally,the height data was processed and calculated to obtain the measured values of the height and parallelism of the end face of the cylindrical plug with respect to the base surface of the casing,and the judgment of the plunger assembly quality was completed in order.(5)Based on the above key technology research,the detection system was set up on site and tested online.The field test and experimental analysis show that the accuracy rate of cylindrical plug defect recognition of the detection system in this paper reaches 97.2%,the probability of miss is 0.6%,the detection beat is less than 13 s,and the expanded uncertainty of the assembly quality detection system is 0.267 mm,which meet the technical indicator requirements.