Sample-equalizable crankshaft surface defect detection method and detection system

Abstract

The invention discloses a sample-equalizable crankshaft surface defect detection method and a sample-equalizable crankshaft surface defect detection system, and the detection method comprises the following steps: constructing a training sample data set: applying magnetic powder to the surface of a crankshaft and irradiating with fluorescent light to obtain a surface image of the crankshaft; judging and marking the surface state to obtain normal image data and defective image data, defining an image as a data sample, marking the sample as a positive sample when all the images are normal images, otherwise, marking the sample as a negative sample, forming a training sample data set with the value of (N + 1) / (M + 1); training based on a deep learning convolutional neural network model; and forming a data sample by y images of the crankshaft to be detected and x-y images in N, and determining the surface condition of the crankshaft through a deep learning recognition model. According to the invention, image data with defects and normal image data can be kept balanced, and a sample expansion effect is achieved under the condition of the same number of images.

Description

technical field [0001] The present invention relates to the technical field of crankshaft surface inspection, in particular to a crankshaft surface defect inspection method and inspection system capable of equalizing samples. Background technique [0002] The crankshaft is the drive shaft of the engine and is the most important part of every engine, generally produced by forging or casting. After the crankshaft is forged or cast, it also needs to undergo processes such as heat treatment, shot blasting, and flaw detection. In forming production and heat treatment, crankshafts are prone to various production defects, such as shear cracks, forging folds, quenching cracks, grinding cracks, etc. The existence of these defects can easily cause the engine to break during operation, resulting in safety accidents and property damage. Therefore, in the final process of production, it is extremely important to detect crankshaft defects and find quality problems in the production proc...

AI Technical Summary

Problems solved by technology

In actual production, the acquired image data with defects is far less than the normal image data. The traditional sample making method makes the positive and negative samples seriously unbalanced, which will seriously affect the performance of the algorithm.