Line type identification method for auto-preset mould of car body action frame

In the case-based design of the intelligent CAD system for automobile panel molds, in order to facilitate the implementation of the system, when comparing the similarity of two CAD models, several section lines are taken as characteristic curves on each panel CAD model, respectively. Calculate the similarity of these characteristic curves, and then calculate the similarity of the entire cover from the similarity of these curves. In this way, the case retrieval in the case-based automobile panel mold CAD is realized.

In 1989, MathaiG. applied BAM neural network to solve the classification problem of power spectral density function PSD in the fiber manufacturing process. The specific method is to first preprocess the discrete PSD curve to obtain a new discrete point sequence yi(i), where i is the frequency value, and then divide the yi(i) curve into equal parts r along the amplitude axis and s along the frequency axis Make equal parts to get a square. According to whether the discrete yi(i) value falls within a certain square, assign the corresponding binary matrix element to 1 or 0, as shown. Connect the rows of the above matrix to obtain a unipolar binary vector of one dimension, then double-polarize it and input it into the designed BAM network to realize the classification of the PSD curve.

In the case-based intelligent CAD system for forging die intelligent CAD developed by Nakajima Shoji <3> and others, the following method is used to calculate the similarity of the CAD model: the two-dimensional part drawing is divided into small spaces with 5 in the horizontal and 4 in the vertical. , Then set the inside of the part to 1 and the outside of the part to 0 to make a binary model. Apply the binarization model to calculate the feature quantities such as the area ratio and circumference ratio of each region, and then use these feature quantities to weight the similarity of the parts.

The common point of the above algorithms is that the curve is converted into a binary graph, which requires a large amount of memory for calculation, and does not have the invariance to the translation, rotation, and scaling of the curve. In particular, it is sensitive to the overall difference in the shape of the curve, but not to the difference in the fillet radius, which has a significant impact on the stamping formability.

This paper starts from the influence of the shape of the curve on the stamping formability, uses the vector composed of the radius of curvature of the discrete points on the curve to describe the shape of the curve, uses the interface function provided by UG to directly read the radius of curvature of the curve from the curve digital model, and Use neural network as a classifier to classify and recognize the shape of the curve.

The discretized radius of curvature of the curve data is a parameter that directly reflects the shape of the curve. If the radii of curvature of the two curves are equal everywhere, then the shapes of the two curves must be the same. Using the radius of curvature to express the shape of the curve has the following advantages: (1) The radius of curvature is an invariant of the translation and rotation of the curve, and it can accurately reflect the shape of the curve. (2) Compared with the aforementioned features <2, 3> that reflect the overall shape of the curve, the radius of curvature, as a feature that reflects the local shape of the curve, is more in line with the requirements of the formability analysis for the recognition of the curve shape. (3) Compared with the method of expressing the curve with coordinate points, the radius of curvature can better reflect the formability of the curve. The two curves shown are section lines taken from two square boxes. If you take a series of coordinate points as the characteristic quantity of the curve, then the distance between the two curves in the characteristic space is very small, so it is difficult to distinguish; while taking the radius of curvature of a series of points as the characteristic vector, the two curves are easier distinguish. In fact, it is the radius of curvature of each point on the curve and its changes that really affect the formability of stamping parts.

For the sheet metal stamping forming problem (take the drawing of the square box shown as an example), the main factors affecting the forming performance are: the arc radius at the 4 corners, the slope of the side wall and the height of the side wall. When using coordinate points to represent the curve, it is easy to compare the difference in the height of the side wall, but it is not sensitive to the change of the arc radius and the slope of the side wall. When the curve is expressed by the radius of curvature, the pre-processed figure shows that the 4 wave crests represent the size of the arc radius at the 4 corners. After preprocessing, the large curvature radius value has been transformed to close to zero. Therefore, the deviation of the wave crest position and the distance between the crests have a significant influence on the comparison of the curve shape, and these two just reflect the slope of the side wall and the distance between the crests. The difference in the height of the side wall, so the curvature radius representation method is more sensitive to the three factors that affect the formability, which is an ideal representation method.

The automotive parts and parts machining, PTJ Shop offers the highest degree of OEM service with a basis of 10+ years experience serving the automotive industry. Our automotive precision shop and experts deliver confidence. We have perfected the art of producing large component volumes with complete JIT reliability, backed by the quality and long-term reliability our customers expect.