Tool Broken Detection Based On GoogleNet Model

The gradual advancement of Industry 4.0 has played an important role in the production of traditional industries.Using artificial intelligence to solve some problems in the industry has also become an important part of the development process.This paper introduces a method to test the small broken tool with a radius of less than 5mm.The tool would be easily broken,not wear.As a result,the whole design has a good value for detecting this conditions.The deep learning method is used to detect the broken tool.The unrelated background is removed by template matching.And the processed image are scaled to the same size,which can effectively remove the influence of the size of the tool on the final result,and also enhance the adaptability of the model to the tool size.Then the processed images are used as the training set.During the training process,the parameters are set.However,the over-fitting problems appears during the training process.Finally,the accuracy of this model is about 90%.But the whole model is not performed well in the case of poor shooting conditions.What’s worse,the overall accuracy rate has a certain degree of decline.In view of these problems,the inception model,a kind of migration learning model,is added to improve the original network model,which increases the depth and width of the original model network,and solves over-fitting the existing models.The improved model reaches better score,and the training speed is 3/2 of the original speed during the whole training process.At the same time,the images with poor shooting conditions get better learning results.The accuracy of the improved model can be kept stable at around 96%.Compared with the original model,there is a big improvement in accuracy and operating efficiency.