A method and system for predicting standard working hours of a garment short-water-flow modular processing procedure

Abstract

The application discloses a kind of garment short water modular processing process standard working hour prediction method and prediction system, including the following steps: (1) historical data acquisition;(2) influence factor analysis: in combination with the situation in actual production, determine the three major influencing factors of garment production and processing process standard working hour, including: sewing object, sewing structure and process requirement;According to the numerical type of influence factor and the influence on working hours, encode each factor using the serial number method, construct the influence factor matrix, and objectively evaluate the weight of each influence factor after numerical normalization by entropy method;(3) machine learning model training;(4) working hours result output: input the basic information of new clothes into the model trained in step (3) and fit, to obtain the standard working hours of each process of the current garment;The prediction system includes style selection module, parameter input module, result output module and data storage module.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of garment production and processing, and particularly provides a garment short-flow modular processing procedure standard time prediction method and a prediction system. BACKGROUND

[0002] China is a large country in textile and garment production and manufacturing. As a labor-intensive industry, many enterprises are trying to build an internal lean management system to achieve efficiency and cost reduction and maintain the fundamental competitive advantage of the enterprise. At present, the overall garment market presents a production mode of "multi-variety, small batch, short cycle", which requires the manufacturing party to deliver high-quality finished products in a short period of time.

[0003] With the continuous promotion of industrial engineering (IE) management mode, many enterprises have begun to use the method of "pre-analysis and post-production" for production management, and the overall efficiency has been significantly improved. Among them, standard time is the basis of industrial engineering and an important indicator in the garment production process. Many enterprises use the GSD action code analysis method to determine the standard time of each garment and then implement process arrangement, but in actual production, each garment needs to be analyzed separately. The analysis of a single garment takes more than one hour, a large number of repetitive actions are performed during the analysis process, and in the short-flow scenario, the pre-process analysis of the garment is particularly important. At the same time, as the number of styles increases, the database gradually becomes large, and historical information cannot be effectively utilized, which results in waste of manpower and time. SUMMARY

[0004] The purpose of the present application is to provide a garment short-flow modular processing procedure standard time prediction method and a prediction system. In view of the deficiencies of the prior art, on the one hand, a garment short-flow modular processing procedure standard time prediction method is proposed to improve the efficiency of garment standard time determination, and on the other hand, a garment short-flow modular processing procedure standard time prediction system for implementing the method of the present application is proposed to simplify the overall time analysis operation process.

[0005] In order to solve the above technical problems, the following technical solutions are adopted:

[0006] A garment short-flow modular processing procedure standard time prediction method and a prediction system, characterized by comprising the following steps:

[0007] (1) Historical data collection: Collecting garment processing workshop production data in recent years, including garment processing basic information and standard time corresponding to each process, garment processing basic information including garment style, parts, fabric, process and machine; the determination of standard time adopts the method of combining GSD action analysis and D-station digital station time collection;

[0008] (2) Influence factor analysis: combined with the actual production situation, three categories of influence factors of garment production and processing standard time are determined, including: sewing object, sewing structure and process requirement; according to the numerical type of influence factors and the influence on the length of time, the order number method is used to code each factor, the influence factor matrix is constructed, and the weight of each influence factor is objectively evaluated by entropy method after numerical normalization;

[0009] (3) Machine learning model training: a BP neural network model is constructed, which consists of input layer, hidden layer and output layer, and is trained by gradient descent method. The number of neurons in the input layer and output layer is determined by the input influence factors and output results respectively. The number of layers and neurons in the hidden layer is adjusted based on the experience value and the training results. The obtained garment style information and time data samples are input into the BP neural network, and the trained model is obtained and saved after repeated training;

[0010] (4) Time result output: the basic information of the new garment is input into the trained model in step (3) for fitting, and the standard time of each process of the current garment is obtained.

[0011] After optimization, the garment style in step (1) is divided into parts or parts according to the garment structure line as the minimum operation unit of process; the parts are the scattered small area cutting pieces such as collar, pocket, front collar, etc. which need to be made separately in advance; the parts are the garment structure line parts such as shoulder seam, side seam, sleeve pocket and bottom swing.

[0012] After optimization, the historical data collection in step (1) is the historical data of the completed production in the enterprise, and the standard time data is determined by professional IE analysis personnel according to the sample clothes, and the process required for each part or part is determined. Before actual production, GSD action analysis is used to determine the time, and then the action code is adjusted according to the actual production situation in the production process, and finally the standard time database is formed; wherein the GSD action analysis uses the GSD standard time analysis system commonly used in the enterprise, and the time before and after the completion of the process is counted in real time through D-station digital station in actual production, and the previously analyzed time is corrected.

[0013] After optimization, in the influence factor analysis of step (2), the weight of each factor on the time is determined by combining the basic parameter information with the expert knowledge in the field, wherein the sewing object includes three factors of fabric grade, cutting piece layer number and fabric pattern, the sewing structure includes three factors of length size, sewing shape and sewing number, and the process requirement includes three specific influence factors of needle step type, machine type and process difficulty, totaling three categories and nine influence factors for standard time determination.

[0014] Preferably, the encoding mode of the influencing factor in step (2) adopts the serial number method, and the nine influencing factor data are divided into classified data and sequenced data, wherein the needle type and the machine type are classified data, and the rest are sequenced data; the encoding of the classified data is sorted and encoded according to the length of the time consumption under the condition of uniform other factors; the sequenced data is directly encoded according to the length of the time consumption from small to large, and finally the influencing factor matrix is constructed.

[0015] Preferably, the BP neural network model constructed in step (3) has the number of input layer neurons determined according to the input influencing factors, the output result in the output layer is one data of standard working hours, and the number of corresponding output layer neurons is set to 1; the number of hidden layer neurons is determined according to the empirical value formula w is a constant between 0 and 10, which is preliminarily determined, and then adjusted according to the prediction result.

[0016] A garment short flow modular processing process standard working hour prediction system, characterized in that: the prediction system is used to implement a garment short flow modular processing process standard working hour prediction method, and the prediction system comprises a style selection module, a parameter input module, a result output module and a data storage module;

[0017] The style selection module is used for selection and editing of garment styles, parts and process types, and generates basic garment style information; first, the basic garment style is selected, and the system style library contains common basic styles such as shirts, T-shirts, dresses, skirts, casual pants and western pants; after selecting the style, the corresponding part information is associated; after the style is determined, the style type of each part and the corresponding process are further selected;

[0018] The parameter input module is used for further determination of the manufacturing process after the style of the sample garment is determined, selection of nine influencing factors, and input of the length value of the process operation to perfect the basic information of the process;

[0019] The result output module directly displays the standard working hours predicted by the corresponding process after the influencing factors are input, and writes the standard working hours into a data table; after all the processes are predicted, the results can be directly exported;

[0020] The data storage module stores the basic information of the style and the corresponding standard working hours after the current sample garment analysis is completed, which is used for calling similar styles later.

[0021] Preferably, the style selection module is provided with a subordination logic, the parts belong to the garment style, and different styles correspond to different part types; the process type is a two-level label selection, which is divided into a process category and a detailed needle type.

[0022] With the technical scheme, the following beneficial effects are achieved.

[0023] 1. According to the clothing part processing similarity principle, different categories and styles of clothes are classified and arranged to construct a clothing style, process procedure and standard working hour database.

[0024] 2. The influence factors of the quantitative standard working hour are quantified, the machine learning algorithm is used for data mining, the operation process of the IE analysis personnel is simplified, the repeated actions during analysis are effectively reduced, the clothing process analysis time is shortened, and thus the efficiency of the standard working hour determination is improved.

[0025] 3. The computer system is used to store the historical data, and the similar clothing style and process can be quickly called. BRIEF DESCRIPTION OF DRAWINGS

[0026] The application will be further described below in combination with the drawings:

[0027] Figure 1 A clothing production standard working hour prediction method and system based on a neural network;

[0028] Figure 2 A summary of the types of parts (positions) of a shirt style;

[0029] Figure 3 A shirt style part schematic diagram;

[0030] Figure 4 A D-station digital station working hour data acquisition result schematic diagram;

[0031] Figure 5 A standard working hour influence factor and coding result;

[0032] Figure 6 A standard working hour BP neural network structure schematic diagram. DETAILED DESCRIPTION

[0033] The application aims to provide a clothing short flow modular processing procedure standard working hour prediction method and prediction system to improve the efficiency of clothing standard working hour determination; at the same time, a clothing short flow modular processing procedure standard working hour prediction system for implementing the method of the application is proposed to simplify the operation process of the overall working hour analysis.

[0034] The technical scheme of the application will be further described below in combination with specific examples.

[0035] Example 1

[0036] Taking a shirt style as an example, the short flow modular processing procedure standard working hour prediction method mainly includes the following steps:

[0037] (1)Collect the styles of shirts produced by the enterprise and the corresponding motion analysis to obtain basic data and standard time information of the styles, determine the styles of basic parts (positions) of shirts based on the structure line, Figure 2 arrange the common types of each part of the shirt style. For example, the short-sleeved shirt style in Figure 3 , the style of each part (position) is as follows: three-clamp collar, round corner pocket, continuous front lapel, folded sleeve cuff, rolled hem bottom swing, and the assembly seam uses the method of single-needle computerized sewing and then three-line sewing machine hemming. The data collected by the digital workstation is shown in Figure 4 , the specific operation time of each workstation and the interval time between work procedures are obtained, which are compared with the standard time analyzed to obtain the final standard time data of the garment process, which is used for subsequent analysis.

[0038] (2) Analyze the factors affecting the working hours and encode them to determine the weight of the influencing factors. According to the results of motion analysis, the factors affecting the working hours can be divided into three categories: sewing objects, sewing structures, and process requirements, and there are nine specific influencing factors: fabric grade, number of layers, fabric pattern, length and size, sewing shape, sewing quantity, needle step type, machine type, and process difficulty level. Nine kinds of influencing factor data are divided into categorical data and ordinal data, among which needle step type and machine type are categorical data, and the rest are ordinal data. The encoding of categorical data is sorted and coded after the length of time consumed by the operation unit is counted under the condition of uniformity of other factors. Fabric grade, process difficulty, sewing shape, and fabric pattern have the property of grade, and different parameters of each factor will affect the length of standard time. For example, the higher the fabric grade, the more difficult the sewing operation, and the longer the time consumed. The fabric grade is coded from 1 according to the length of time consumed. There are five levels. Length and size, number of layers, and sewing quantity have corresponding numerical values, i.e. the actual numerical values are input. Needle step type and machine type are coded in order according to the length of time consumed by the same process of the operation unit. The specific coding results are shown in Figure 5 . The weights of the influencing factors are determined by the entropy method after normalization of the numerical values. The calculation results of the weights of each factor are shown in Table 1. The basic attributes of the sewing object have a relatively large impact on the working hours, while the impact of the process requirements is relatively small.

[0039] Table 1 Calculation results of the weights of nine influencing factors

[0040]

[0041]

[0042] (3) The neural network model adopts a basic BP neural network, which is composed of an input layer, a hidden layer and an output layer. The structure of the hidden layer and the number of neurons are adjusted according to the prediction results, the gradient descent method is used for training, and the best neural network model structure is saved. The number of input layer neurons is 9, the number of output layer neurons is 1, the number of hidden layer is determined to be 2 after repeated experiments, and the number of neurons is 8 and 6 respectively. The neural network structure is as shown in Figure 6 The correlation coefficients between the predicted values and the sample values in the training, validation and test sets are 0.97289, 0.97097 and 0.9669 respectively, which have significant correlation. The overall R value is 0.97164, and the fitting effect is good. The relative error of 71% of the process time prediction results is within 10%, and 87.1% is within 15%, which basically meets the acceptable float range in actual production. The Figure 3 The relative error of the predicted results of the style splicing part is within 5% for most of the processes, except for a few processes.

[0043] Table 2 Standard process time prediction results of part of the splicing part

[0044]

[0045]

[0046] Step 4, time result output: input the basic information of the shirt into the trained model for fitting to obtain the standard process time of each process of the current shirt.

[0047] Example 2

[0048] The garment short flow modular processing process standard time prediction system is established on the basis of the garment short flow modular processing process standard time prediction method and prediction system of example 1. A large amount of historical data is stored in the system, including four modules of style selection module, parameter input module, result output module and data storage module.

[0049] The style selection module is used for selecting and editing the style, parts and process type of the garment, and generating basic garment style information. First, the basic style of the garment is selected. The style library of the system includes common basic styles such as shirts, T-shirts, dresses, skirts, casual pants and western pants. After selecting the style, the corresponding part information is associated. After the style is determined, the style type of each part (or part) is further selected. The selection module sets the dependent relationship logic, and the operation part (part) belongs to the garment style. Different styles correspond to different part (part) types. The process type is a two-level label selection, which is divided into process category and detailed needle step type.

[0050] Parameter input module, after the style of the sample clothes is determined, further manufacturing process determination is needed, nine categories of influence factors are selected, and specific length values of process operation are input, so as to perfect the basic information of the process.

[0051] Result output module, after the influence factors are input, the interface directly displays the standard working hours predicted by the corresponding process after clicking the determination, and the standard working hours are written into the data table, and the results can be directly exported after all the process prediction is completed.

[0052] Data storage module, after the current sample clothes analysis is completed, the basic information of the style and the corresponding standard working hours are stored while the result output is performed, and are used for subsequent calling of similar styles.

[0053] The analyst selects the basic information of the process corresponding to the style and inputs length size and other parameter information, clicks the determination, the system automatically outputs the prediction result and stores the prediction result, and subsequent calling is facilitated.

[0054] The above is only a specific embodiment of the present application, but the technical features of the present application are not limited thereto. Any simple change, equivalent replacement or modification made on the basis of the present application to solve the basically same technical problem and realize the basically same technical effect is covered in the protection scope of the present application.

Claims

1. A method for predicting standard working hours for modular processing steps in garment manufacturing, characterized in that... Includes the following steps: (1) Historical data collection: Collect production data of the garment processing workshop in recent years, including basic information on garment processing and standard working hours corresponding to each process. Basic information on garment processing includes garment styles, components, fabrics, processes and machines; the determination of standard working hours adopts a combination of GSD motion analysis and D-station digital workstation time collection method. (2) Analysis of influencing factors: Based on the actual production situation, the three major categories of influencing factors of standard working hours in garment production and processing are determined, including: sewing object, sewing structure and process requirements; according to the numerical type of the influencing factors and their impact on the length of working hours, the numbering method is used to encode each factor, construct the influencing factor matrix, and after numerical normalization, the weight of each influencing factor is objectively evaluated by the entropy method. The weight of each factor on working hours is determined by combining basic parameter information with domain expert knowledge. The sewing object includes three factors: fabric grade, number of cut pieces, and fabric pattern. The sewing structure includes three factors: length, sewing shape, and number of sewing pieces. The process requirements include three specific influencing factors: stitch type, machine type, and process difficulty. There are a total of three categories and nine factors that affect the standard working hours. The influencing factors are coded using a numbering method, dividing the nine influencing factor data into categorical data and ordinal data. Among them, needle step type and machine type are categorical data, while the rest are ordinal data. The categorical data is coded by sorting the time taken under the statistical unit length of operation, while the ordinal data is directly coded according to the time taken of each element in ascending order, and finally constructing the influencing factor matrix. (3) Machine learning model training: Construct a BP neural network model consisting of an input layer, hidden layers, and an output layer. Use gradient descent for training. The number of neurons in the input layer and the output layer are determined by the influencing factors of the input and the output results, respectively. The number of hidden layers and neurons are adjusted based on empirical values ​​and training results. Input the obtained clothing style information and working time data samples into the BP neural network. After repeated training, obtain and save the trained model. (4) Output of working hours: Input the basic information of the new garment into the model trained in step (3) and then fit it to obtain the standard working hours of each process of the current garment.

2. The method for predicting standard working hours for a modular garment processing procedure according to claim 1, characterized in that: In step (1), the garment style is divided into parts or sections based on the garment structure line as the smallest operational unit of the process. The parts are small pieces of fabric that are relatively scattered and need to be made separately in advance, except for the main body pieces, including the collar, pockets, and front placket. The sections are the garment structure lines of the main body pieces, including the shoulder seam, side seam, armhole, and hem.

3. The method for predicting standard working hours for a modular garment processing procedure according to claim 1, characterized in that: The historical data collection in step (1) refers to the historical data of production completed in the enterprise. The standard working time data is first determined by professional IE analysts based on the sample garment to determine the required process for each part or part. Before actual production, the working time is determined by GSD motion analysis. Then, the motion codes are adjusted according to the actual production situation during the production process, and finally a standard working time database is formed. Among them, the GSD motion analysis adopts the GSD standard working time analysis system commonly used in enterprises. In actual production, the time before and after the completion of the process is statistically analyzed in real time through D-station digital workstation, and the working time analyzed in advance is checked.

4. The method for predicting standard working hours for a short-flow modular processing step in garment manufacturing according to claim 1, characterized in that: In step (3), the number of neurons in the input layer of the constructed BP neural network model is determined based on the input influencing factors. If the output result in the output layer is a standard working hour, then the corresponding number of neurons in the output layer is set to 1. The number of neurons in the hidden layer is determined based on an empirical formula. w is a constant between 0 and 10, which is initially determined and then adjusted based on the prediction results.

5. The prediction system for the standard working time prediction method of the short-flow modular processing procedure in garment manufacturing as described in claim 1, characterized in that: This prediction system is used to implement a standard time prediction method for modular processing steps in garment production. The prediction system includes a style selection module, a parameter input module, a result output module, and a data storage module. The style selection module is used for selecting and editing garment styles, components, and process types, generating basic garment style information. First, the basic garment style is selected; the system's style library contains common basic styles, including shirts, T-shirts, dresses, skirts, casual pants, and trousers. After selecting a style, the corresponding component information is associated. Once the style is determined, the style type and corresponding process for each component are further selected. The parameter input module, after the sample garment's style is determined, requires further determination of the production process. Nine influencing factors are selected, and the specific length values ​​for each process operation are entered to complete the basic process information. The result output module, after inputting the influencing factors and clicking "OK," directly displays the predicted standard working hours for the corresponding process and writes them to a data table. After all process predictions are completed, the results can be directly exported. The data storage module, after the current sample garment analysis is completed, stores the basic style information and corresponding standard working hours while outputting the results for future use with similar styles.

6. The prediction system for the standard working time prediction method of the short-flow modular processing procedure in garment manufacturing according to claim 5, characterized in that: The style selection module is configured with a subordinate relationship logic, where components belong to clothing styles, and different styles correspond to different component types. The process type is selected as a secondary label, which is divided into a major process category and a detailed needle step type.

Images