CN-117032081-B - Intelligent shaping processing control method and system based on machine learning

CN117032081BCN 117032081 BCN117032081 BCN 117032081BCN-117032081-B

Abstract

The invention discloses an intelligent shaping processing control method based on machine learning, which comprises the following steps of S1, obtaining a machine learning prediction model, wherein the machine learning prediction model is an MLP model, S2, placing a product to be processed into the machine learning prediction model which is trained offline, further predicting a processing value, feeding back the relevant value to a mechanical arm press head, carrying out shaping processing on the product, S3, measuring product data again after processing, judging to obtain a processing result, and S4, carrying out online training on the machine learning prediction model according to a time interval set by a user. The invention also discloses an intelligent shaping processing control system based on machine learning. Compared with the prior art, the method adopts the data processing and machine learning methods to predict the processing value, and trains and predicts the independent design model of each side of the product to be shaped.

Inventors

ZHAO GUOMAN
ZHANG YINGJIE
DING SUJUN
ZHANG YICAN
HAN ZHONGGE
ZHAO ZHENGANG

Assignees

苏州康鸿智能装备股份有限公司

Dates

Publication Date: 20260508
Application Date: 20230908

Claims (3)

1. The intelligent shaping processing control method based on machine learning is characterized by comprising the following steps of: S1, utilizing all processing data collected before, building a training set by combining existing samples, and performing offline training to obtain a machine learning prediction model, wherein the machine learning prediction model is an MLP model; S2, placing the product to be processed into an offline trained machine learning prediction model, further predicting a processing value, feeding back the related value to a mechanical arm press head, and carrying out shaping processing on the product; s3, measuring product data again after processing and judging to obtain a processing result; s4, performing online training on the machine learning prediction model according to the time interval set by the user; In the step S1, when an MLP model is built, two models are built on the front, left and right sides of the product to be processed, and each side of the product to be processed is correspondingly provided with a primary shaping MLP model and a secondary shaping MLP model, which are respectively aimed at primary shaping and secondary shaping of one side of the product to be processed; the MLP model building process comprises the following steps: S11, introducing an MLP model through Tensorflow. Keras, stacking a multi-layer structure, and building the model; S12, performing independent model training on each predicted machining of each side, wherein data used for MLP model training are original machining data obtained from an actual production process, the original machining data comprise product output information, point position values and straightness of each machining process and target specification interval information of the product, classification processing is performed when the original machining data are processed, classification extraction is performed according to sides and machining times, and six groups of data are obtained, wherein the data are front, left and right three-side twice machining data respectively; S13, after two processing data of front, left and right sides of a product are obtained, defining input parameters of an MLP model, setting a value of a first hidden layer to be 32, setting a value of a second hidden layer to be 64, setting a third hidden layer to be 32, and finally outputting a value as predicted processing amount, wherein the input parameters are defined as follows: S14, performing 2:8 distribution on an original group data set formed by two processing data of front, left and right sides of the product to obtain a test set and a training set, and training the MLP model by using the training set; S15, evaluating the MLP model by using an MAE index, wherein the MAE calculation formula is as follows: , wherein, The total number of the data sets in the test set; The target value corresponding to the test set is obtained; And the predicted value of the corresponding performance in the test set.
2. The intelligent shaping and processing control method based on machine learning according to claim 1, wherein in the step S2, the shaping and processing flow of the product includes the steps of: S21, after a product to be processed is placed into a shaping machine, the product to be processed is firstly fixed on a shaping platform, point positions of the front, left and right three sides of the product to be processed are scanned through a laser scanner, and then the height values of all the point positions are obtained and are transmitted into an industrial computer; s22, after receiving the height value of the material point position, the industrial computer acquires a specification interval input by a user on a front-end interface, and puts the data into an offline trained machine learning prediction model to wait for a shaping processing result; S23, after the data are put into a machine learning prediction model, the data are subjected to model calculation, the corresponding machining value is obtained through prediction, and the machining value is submitted to an industrial computer; And S24, after receiving the predicted processing value, the industrial computer feeds back the predicted processing value to the controller, so as to control the mechanical arm pressure head to perform the pressing operation, and finish the shaping processing.
3. An intelligent shaping machining control system based on machine learning, comprising: An industrial computer for running a shaping control program and a machine learning algorithm, the shaping control program comprising instructions for performing any of the methods of claim 1 or 2; The scanner is used for measuring the height value of the point to be measured on each side of the product to be processed, and then transmitting the height value of the point to an industrial computer to realize machine learning algorithm and shaping control; a shaping platform for placing a product to be processed; A mechanical arm press head for pressing the product to be processed to deform and hold for a period of time; and the controller is used for converting signals after receiving the signals in the industrial computer, transmitting the signals to the mechanical arm and controlling the pressure head of the mechanical arm.

Description

Intelligent shaping processing control method and system based on machine learning Technical Field The invention relates to the technical field of intelligent shaping processing of workpieces, in particular to an intelligent shaping processing control method and system based on machine learning. Background Material shaping is an important part in material processing. Shaping is a process of obtaining the shape of a product by utilizing metal plasticity, and is time-saving, labor-saving and material-saving. In the current shaping system, the mechanical arm pressure head is used for pressing and shaping the front, left and right three sides of an original square material according to a specific stroke amount, so that the height values of the three sides are all within a standard range. Currently, the traditional processing amount is mainly obtained by predicting a processing value through a traditional empirical model. The empirical model is a model for predicting machining accuracy by summarizing and analyzing empirical knowledge accumulated in engineering practice of a production line for a long time. The model starts from the existing data, forms a corresponding table according to experience, and establishes an expert experience library. The content of the expert experience library is mainly the primary measured height values of the front, left and right sides of the material to be processed and the processing amount corresponding to each side, and the processing values of primary shaping and secondary shaping are obtained through table lookup. These values were all manually tested. Each time a new material is put into the processing, the computer table look-up program can search in the manual experience library, and after information consistent with the material data is found, a corresponding processing value is found for processing. The processing mode needs to manually test and reshape each type of workpiece in advance, grasp the proper processing amount of workpieces with different shapes, sizes and materials, needs special engineers to watch the periphery of production line equipment and adjust timely, clearly wastes a large amount of manpower resources, and reduces the material reshaping efficiency. In actual production practice, shaping based on a conventional shaping process prediction model (as shown in fig. 1) has the following disadvantages: 1) The processing process is easily affected by noise interference, elastic deformation of materials and unknown hidden variables, and the factors are difficult to consider based on expert experience libraries, so that the decision and the processing process are generally greatly different from expected ones; 2) At present, the method for obtaining the processing amount by looking up the table through an expert experience library depends on the number of samples, and is difficult to measure all conditions and difficult to sample; 3) The updating of the expert database depends on the adjustment of the latest time and has limited space, so that the expert experience database is suitable for short-term shaping processing and cannot be used for a long time; 4) The method is lack of self-adaptive capability, the processing amount of each side is acquired through an expert experience library, three sides are required to be checked, but the processing of one side of a product also has an influence on other sides, so that the processing amount obtained by the method has errors, manual adjustment is required, and the self-adaptive capability is not realized. Disclosure of Invention The invention aims to provide an intelligent shaping processing control method and system based on machine learning, which adopts a data processing and machine learning method to predict processing values and train and predict an independent design model of each side of a product to be shaped. In order to achieve the above purpose, in one aspect, the invention discloses an intelligent shaping processing control method based on machine learning, comprising the following steps: S1, utilizing all processing data collected before, building a training set by combining existing samples, and performing offline training to obtain a machine learning prediction model, wherein the machine learning prediction model is an MLP model; S2, placing the product to be processed into an offline trained machine learning prediction model, further predicting a processing value, feeding back the related value to a mechanical arm press head, and carrying out shaping processing on the product; s3, measuring product data again after processing and judging to obtain a processing result; And S4, performing online training on the machine learning prediction model according to the time interval set by the user. As a further description of the above technical solution: in step S1, when the MLP model is built, two models are built on the front, left and right sides of the product to be processed, and each side of the prod