CN-121981326-A - Intelligent UPH prediction method, system, equipment and medium for electronic manufacturing production line

CN121981326ACN 121981326 ACN121981326 ACN 121981326ACN-121981326-A

Abstract

The application discloses an intelligent UPH prediction method, system, equipment and medium for an electronic manufacturing production line, which can learn deep interactive relation between line bodies and products through countermeasure training by a generator for generating a countermeasure network through collaborative filtering, so that even if a new product lacks enough historical data, the generator can find a mode similar to other line bodies in a potential space and generate relatively reasonable UPH, and for an old product with theoretical UPH and actual UPH, only the current line body feature vector and the theoretical UPH of the product in the corresponding line body need to be correspondingly input.

Inventors

WANG WEI
HE YANGYANG
JIANG MINGFENG
QIU YUE
ZHANG DINGJIE
GUO YING
LUO ZHENGYI
Ke Xifeng
HE XIN
GAO DONG

Assignees

四川九州电子科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. An intelligent prediction method for a UPH of an electronic manufacturing production line is characterized by comprising the following steps: judging whether the product is a new product or not; if the model is not a new product, inputting the current line body characteristic vector and the product history theory UPH vector into a pre-trained UPH intelligent generation model to generate a corresponding UPH value so as to dynamically update the original UPH value; If the product is a new product, further judging whether the new product is produced in other line bodies or not; if the new product is produced in other line bodies, inputting the characteristic vector of the current line body and the theoretical UPH vector produced in other line bodies by the new product into a pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference; if the new product is not produced in any line body, inputting the characteristic vector of the current line body and the theoretical UPH vector of the similar product into a pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference; The line characteristic vector is determined by analyzing main influence factors determined by personnel and equipment influence factors related to the line yield, wherein the main influence factors form line characteristics, and preprocessing line characteristic data to obtain the line characteristic vector; the UPH intelligent generation model is obtained by training a collaborative filtering generation countermeasure network.
2. The intelligent prediction method for UPH of an electronic manufacturing line according to claim 1, wherein the analysis process of the line body features comprises: The method comprises the steps of obtaining personnel and equipment influencing factors influencing the yield of a line body, wherein the personnel and equipment influencing factors comprise personnel age, personnel average time, new personnel proportion, personnel on-line post matching rate, personnel post skill number, equipment average fault interval period, equipment using time length and whether equipment is currently maintained or not to form an initial data set; And 5 most relevant features are screened out as line body features by utilizing a principal component analysis method, wherein the line body features comprise post matching rate, average job time, new employee proportion, fault interval period and equipment use duration.
3. The intelligent prediction method for UPH of electronic manufacturing line according to claim 2, wherein the generating process of the line body feature vector and the UPH vector comprises: Acquiring theoretical UPH, actual UPH and corresponding line body characteristic data of a line body corresponding to a product; Obtaining theoretical UPH vectors corresponding to each line body based on line body coding, product coding and theoretical UPH construction; Obtaining an actual UPH vector corresponding to each line body based on line body coding, product coding and actual UPH construction; constructing a corresponding line body characteristic vector according to the actual UPH vector corresponding to each line body; And carrying out normalization processing on all the constructed vectors, namely, the values in the finally obtained vectors are normalized values.
4. A method for intelligent prediction of UPH on an electronic manufacturing line according to any one of claims 1 to 3, wherein the training process of the UPH intelligent generation model comprises: Obtaining an actual UPH vector, a theoretical UPH vector and a line body characteristic vector corresponding to a certain line body based on historical data processing, and constructing a training data set; Training a pre-constructed collaborative filtering generation countermeasure network by utilizing the training data set to obtain the UPH intelligent generation model; The collaborative filtering generation countermeasure network comprises an embedded layer, a generator, a dot multiplication module and a discriminator, wherein the input of the embedded layer is a theoretical UPH vector and a line body characteristic vector, the line body characteristic vector and the theoretical UPH vector of a product produced correspondingly by the line body are learned through the embedded layer, specific intermediate characteristic representations are abstracted, fusion is carried out on the two intermediate characteristic representations to form fusion characteristics, the fusion characteristics are input to the generator, the generator generates a false UPH vector according to the fusion characteristics, the dot multiplication module carries out dot multiplication processing on the false UPH vector and the line body product production state vector, the value of the product which is not produced by the line body is filtered, the vector after the dot multiplication processing and the actual UPH vector are input to the discriminator, the generated false UPH and the actual UPH are distinguished through the discriminator, and the discrimination result is reversely propagated to the generator, so that the generator continuously optimizes the generated data.
5. The intelligent prediction method for UPH of electronic manufacturing production line according to claim 4, wherein the adopted loss function in the training process comprises two parts, and the first part is a loss function of the generator, which is expressed as: ; Wherein, the Representing the loss of the generator; representing the result of the discriminator correctly distinguishing the false UPH from the actual UPH; Representing the regular term coefficients; representing line body generated by generator False UPH vectors; Indicating line body A product production state vector; Indicating line body An actual UPH vector; representing the Hadamard product; Representing a product Online body Actual UPH value during production; Representing a product Online body False UPH values generated during production; the second part is the loss function of the arbiter, expressed as: ; Wherein, the Representing the loss of the arbiter; Representing the correct classification result of the actual UPH vector by the discriminator; The total loss is a combination of the generator and arbiter losses, and model training is aimed at minimizing the generator losses and maximizing the arbiter losses, which are progressively optimized by way of countertraining.
6. The intelligent prediction method of UPH in electronic manufacturing production line according to claim 4, wherein in the training process, small batches of random gradient descent and back propagation are adopted to train the generator and the discriminator so as to minimize the loss of the generator and maximize the loss of the discriminator, thereby obtaining optimal parameters, and substituting the optimal parameters into the countermeasure network to obtain the intelligent generation model of UPH; During training, the parameters of the generator and the arbiter are updated alternately.
7. An electronic manufacturing production line UPH intelligent prediction system, comprising: the first judging unit is used for judging whether the product is a new product, if so, driving the first predicting unit to work, and if not, driving the second judging unit to work; The first prediction unit is used for inputting the current line body characteristic vector and the product history theory UPH vector into the pre-trained UPH intelligent generation model to generate a corresponding UPH value so as to dynamically update the original UPH value under the condition that the current line body characteristic vector and the product history theory UPH vector are judged to be not new products; The second judging unit is used for further judging whether the new product is produced in other line bodies or not under the condition that the new product is judged, if so, the second predicting unit is driven to work, and if not, the third predicting unit is driven to work; The second prediction unit is used for inputting the characteristic vector of the current line body and the theoretical UPH vector of the new product produced by the new product in other line bodies into the pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference under the condition that the new product is judged to be produced by the other line bodies; And a third prediction unit, configured to input, when it is determined that the new product is not produced by any line body, the current line body feature vector and the theoretical UPH vector of the similar product to the pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference.
8. The electronic manufacturing line UPH intelligent prediction system of claim 7, further comprising: and the output unit is used for outputting the UPH generated by the UPH intelligent generation model to a third party platform or system to assist in the establishment of the production line output standard.
9. An electronic device comprising a memory and a processor, the memory storing a computer program, wherein the processor, when executing the computer program, implements the electronic manufacturing line UPH intelligent prediction method of any one of claims 1-6.
10. A computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements the electronic manufacturing line UPH intelligent prediction method of any one of claims 1-6.

Description

Intelligent UPH prediction method, system, equipment and medium for electronic manufacturing production line Technical Field The application belongs to the field of production line production efficiency evaluation in the electronic manufacturing industry, and particularly relates to an intelligent UPH prediction method, system, equipment and medium for the production line in the electronic manufacturing industry. Background In the electronics manufacturing industry, UPH (Units Per Hour) is a core indicator that measures the production efficiency of a production line or unit of work Per Hour. The formulation and updating of the UPH standard is a dynamic process that is based on extensive time studies, production data, equipment status, and ongoing improvements. Through periodic evaluation and feedback mechanisms, the UPH standard can be ensured to accurately reflect the current production capacity, and the continuous improvement of the production efficiency is effectively promoted. The UPH standard should also be updated synchronously during continuous optimization of the production flow, technique and process. Currently, in the initial production stage of new products, because of the lack of UPH reference, UPH can only be formulated according to the same industry standard or the average hourly yield of the production line. However, the practical significance of such standard references is not great in view of the significant differences in the practice of each producer. In addition, because of the difference of product procedures, large errors are easily generated when the production line is relied on to average the hourly production. Meanwhile, the present manual operation still occupies an important place in the electronic manufacturing industry, especially in complex assembly and testing links. The difference of staff skill level easily causes fluctuation of operation efficiency and quality, and further affects stability and update of UPH standard. In addition, the failure and instability of the equipment in the production process directly restrict the production efficiency. Problems such as equipment maintenance, aging and shutdown often lead to the fact that actual UPH cannot reach expectations, and especially when production equipment frequently breaks down, the accuracy and instantaneity of UPH updating face serious challenges. These factors make UPH update cycles of many production lines often exceed one month, and are not efficient enough, and hysteresis is common. In summary, the existing UPH updating mode has the following challenges that UPH prediction cannot be accurately realized due to lack of enough historical production data when a new product is faced, while the theoretical UPH value and the actual UPH value exist in the production process of an old product, and the theoretical UPH is usually adjusted and updated by adopting a manual adjustment mode, so that the timeliness and the reliability are both insufficient. Disclosure of Invention Aiming at the problems of the existing UPH updating technology of the production line in the electronic manufacturing industry, the application provides an intelligent UPH prediction method, system, equipment and medium of the production line in the electronic manufacturing industry. The application is realized by the following technical scheme: An intelligent prediction method for a UPH of an electronic manufacturing production line comprises the following steps: judging whether the product is a new product or not; if the model is not a new product, inputting the current line body characteristic vector and the product history theory UPH vector into a pre-trained UPH intelligent generation model to generate a corresponding UPH value so as to dynamically update the original UPH value; If the product is a new product, further judging whether the new product is produced in other line bodies or not; if the new product is produced in other line bodies, inputting the characteristic vector of the current line body and the theoretical UPH vector produced in other line bodies by the new product into a pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference; if the new product is not produced in any line body, inputting the characteristic vector of the current line body and the theoretical UPH vector of the similar product into a pre-trained UPH intelligent generation model to generate a corresponding UPH value as an initial UPH reference; The line characteristic vector is determined by analyzing main influence factors determined by personnel and equipment influence factors related to the line yield, wherein the main influence factors form line characteristics, and preprocessing line characteristic data to obtain the line characteristic vector; the UPH intelligent generation model is obtained by training a collaborative filtering generation countermeasure network. In some embodiments, the analysis of the line b