CN-121981424-A - Self-adaptive scheduling method and system for multi-specification pipe stock in discrete scene

CN121981424ACN 121981424 ACN121981424 ACN 121981424ACN-121981424-A

Abstract

The invention relates to the technical field of enterprise production intelligent management, and particularly discloses a self-adaptive scheduling method and system for multi-specification pipe stock in discrete scenes. The method comprises the steps of collecting parameters of the excess materials, constructing an attribute file in an RFID mode to obtain an excess material real-time gene library, reading order demand parameters, aligning standardized demand vectors with dimensions of excess material genes, constructing order demand gene expression in priority and delivery time to obtain an order queue to be expressed, constructing a welding defect probability prediction model and an equipment health supplement model based on historical data and real-time equipment data, obtaining a scheduling scheme by the excess material real-time gene data, the order queue to be expressed, the welding defect probability prediction model and the equipment health supplement model, and performing closed-loop execution, model update and gene library update by the scheduling scheme. The invention provides an effective solution for the production scheduling intellectualization in the discrete manufacturing field, and has remarkable practicability and economic value.

Inventors

DING JIANWEI
YIN QINGWEN
LIN RUIQI
WU WENLIANG
WU BO
TIAN YUAN
WU HAIFENG
ZHU HAIHAO

Assignees

东方电气集团数字科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251210

Claims (10)

1. The self-adaptive production scheduling method for the multi-specification pipe stock in the discrete scene is characterized by comprising the following technical processes: collecting parameters of the residual materials, constructing a complete attribute file in an RFID mode, and obtaining a real-time gene library of the residual materials; Reading order demand parameters, aligning the standardized demand vector with the dimension of the residue genes, and constructing order demand gene expression by taking priority and delivery date as references to obtain an order queue to be expressed; Based on historical data and real-time equipment data, constructing a welding defect probability prediction model and an equipment health supplement model; Performing scheduling by using the real-time genetic data of the residual materials, the order queue to be expressed, the welding defect probability prediction model and the equipment health supplement model to obtain a scheduling scheme; Performing closed-loop execution in a production scheduling scheme; and updating the model and the gene library according to the closed-loop execution result.
2. The method for adaptively scheduling multi-specification tube stock in discrete scenes according to claim 1, wherein the process for obtaining the real-time gene library of the excess stock is as follows: Step 1, acquiring parameters including outer diameter, wall thickness, length, material, furnace batch number, equivalent depth of surface defects, storage coordinates and material taking time of corresponding excess materials in a mode of combining laser measurement with RFID tracing; Step 2, mapping the acquired parameters into a 42bit binary gene sequence to realize standardized coding of the residue information, wherein coding logic discretizes each parameter according to a value range, and ensures the uniqueness and readability of the coding; The generation of the gene fragments corresponding to the remainders satisfies the following relation: ; wherein Gene is the Gene sequence corresponding to the remainder; d 0 is the outer diameter of the corresponding residue; w 0 is the wall thickness of the corresponding remainder; l 0 is the length of the corresponding residue; M 0 is the material corresponding to the remainder; H 0 is the furnace batch number corresponding to the residue; delta 0 is the equivalent depth of the surface defect corresponding to the remainder; And 3, carrying out reproducible judgment and gene library construction, wherein the method comprises the following steps: The method comprises the steps of (1) carrying out regeneration judgment, and if the length L 0 of the excess material is not less than (the minimum usable length is set and the kerf compensation length is set), marking the excess material as a 'regenerated excess material'; And (3) warehousing a gene library, writing the gene sequence of the renewable clout into a real-time gene library, synchronously updating the RFID tag of the corresponding clout, and realizing full life cycle tracking of the clout.
3. The method for adaptive production of multi-specification tubing stock in discrete scenarios of claim 1, wherein the order demand gene expression is constructed by: Step 1, reading the complete demand parameters of each order, including the outer diameter, wall thickness, length, material quality, number of required pieces, delivery period, priority and weld grade of the pipe; step 2, converting parameters of the order into standardized demand vectors according to the following relation to realize dimension alignment with the residue gene data: ; Wherein DEMANDVEC j is the standardized demand vector of the current order; d j is the outside diameter of the tube for the current order requirement; w j is the wall thickness of the tube for the current order requirement; L j is the pipe length for the current order requirement; M j is the material of the current order requirement; g ́ j is the weld grade of the current order requirement; P j is the priority of the current order; n j is the current order lot; Step 3, generating an order queue to be expressed according to a secondary ordering rule of priority-delivery date, so as to ensure the priority treatment of the emergency order; The sequencing logic is that the priority is firstly increased according to the priority, and the same priority is increased according to the normalized delivery date; the order queue to be expressed satisfies the following relation: ; Wherein Q is an order queue to be expressed; DEMANDVEC 1、 DemandVec 2、 DemandVec No is an order for each batch; NO is the total number of orders.
4. The adaptive production method for multi-specification pipe stock in discrete scenes according to claim 1, wherein the construction process of the welding defect probability prediction model is as follows: step 1, extracting a set amount of historical crater data, wherein each set of historical crater data comprises pipe specifications, welding parameters, environmental parameters and defect labels; step 2, constructing a basis function by adopting a 6-order knowledge-data fusion regression mechanism and combining with metallurgical rules, solving model coefficients by a regularized least square method, and outputting defect probability of any candidate welded junction ; And 3, performing constraint application, including: If it is 1% Is marked as "high risk weld" and the risk is reduced by adjusting the welding current until ≤1%。
5. The method for adaptively scheduling multi-specification pipe stock in a discrete scene according to claim 1, wherein the construction process of the equipment health supplementary model is as follows: Step 1, obtaining an equipment health index based on real-time equipment signals, wherein the following relational expression is satisfied: ; In the formula, Is a device health index; I rms is an effective value of the current of the main shaft of the cutting machine tool; i 0 is rated current; N arc is the number of arc striking times of the welding machine; n max is the maximum number of times per day; Is a solid melting furnace temperature drift; ; and 2, adjusting a compensation strategy by taking the equipment health index as a reference, wherein the following relation is satisfied: If it is <0.85, Adding penalty coefficient in the production objective function, and forcedly inserting preventive type changing window with set time for equipment maintenance; The penalty factor is kappa=0.1 time (0.85- )。
6. The adaptive production scheduling method for multi-specification pipe stock in a discrete scenario of claim 1, wherein the production scheduling scheme is obtained by the following steps: step 1, designing a double-chain chromosome structure, and binding and encoding 'order procedure-remainder matching', wherein the integrated optimization treatment is realized by the following relation: ; Wherein, chrom is a double-stranded chromosome data structure; O 1 、O 2 、O K is the corresponding production task order index in the process chain pi; K is the total number of production pieces; r 1 、r 2 、r K is the corresponding residue index in the residue chain ρ; And 2, quantitatively evaluating the advantages and disadvantages of the production scheme by taking the material utilization rate, the delivery timing rate and the production stability as three targets, and meeting the following relational expression: ; wherein F is the quality evaluation score of the scheduling scheme; η material is the material waste rate, and the concrete calculation relation is that the sum of the length difference values of the excess materials and the total length of the excess materials are less than or equal to 10 percent; Eta deliuery is the order delay rate, and the specific calculation relation is that the sum of delay time and delivery date is less than or equal to 5 percent; p defect,avg is the average welding defect probability, and the target is less than or equal to 1%; H health,avg is the average equipment health index, and the target is more than or equal to 0.85; T makespan is the total production period, and the target after normalization is less than or equal to 1.2; And 3, optimizing in a small sample evolution solving mode to obtain an optimal production scheduling scheme.
7. The method for adaptive production scheduling of multi-specification tubing stock in a discrete scenario of claim 6, wherein the small sample evolution solution process comprises: Firstly, generating 50 initial chromosomes, wherein 5% of the initial chromosomes are knowledge rule individuals; The following genetic operations were then performed: -tournament selection, 3 individuals at a time, retaining the highest fitness; -two-point sequence crossing, exchanging gene fragments, maintaining the procedure-remainder correspondence; reverse sequence variation, adjusting the sequence of the working procedure, and synchronously updating the residual material chain; And finally, continuously carrying out 20 generations without lifting or stopping after iterating 100 generations, and outputting the optimal production scheme corresponding to the chromosome with the highest fitness.
8. The method for adaptive production scheduling of multi-gauge tubing stock in a discrete scenario of claim 1, wherein the closed loop execution comprises: Step 1, analyzing a production scheme into equipment instructions through an OPC-UA protocol and transmitting the equipment instructions to a PLC, wherein the instructions comprise residual material calling coordinates, cutting length, welding parameters and heating time; And 2, collecting cutting precision, welding parameters and equipment states in real time, and ensuring that production is advanced according to a plan.
9. The method for adaptively scheduling multi-specification tube stock in a discrete scenario of claim 1 wherein the model update process is: Step 1, detecting a weld junction defect through DR flaw detection, and generating a feedback sample containing welding parameters and defect labels; Step 2, updating a prediction model coefficient of the welding defect by adopting an incremental SVD algorithm, so as to avoid full retraining; the gene library updating process comprises the following steps: step 1, re-measuring parameters of new remainder produced in production and judging whether the remainder can be regenerated or not; And 2, recoding the renewable clout, adding a clout real-time gene library, and deleting the consumed clout genes.
10. A self-adaptive scheduling system for preparing materials of pipes with multiple specifications in discrete scenes is characterized in that: the adaptive scheduling system is developed based on the adaptive scheduling method of any one of claims 1 to 9; The adaptive scheduling system has: The system comprises a residual material data acquisition module, a RFID tag matching relation and a RFID tag matching relation, wherein the residual material data acquisition module is used for acquiring parameters of corresponding residual materials; The order data acquisition module is used for executing order reading and acquiring parameters of a corresponding order; the equipment state acquisition module is used for acquiring real-time signals of equipment, and inputting historical data and industry knowledge; the quality feedback acquisition module is used for acquiring defect data in the closed loop execution process; The device health compensation module is used for executing the construction process of the device health compensation model; The residue gene coding module is used for executing standardized coding of residue information and constructing a residue real-time gene library; an order demand gene expression module, the order demand gene expression module being used to perform a construction process of order demand gene expression; a welding decay field modeling module for performing a construction process of a welding defect probability prediction model; 6-KDFE evolution and production module, wherein the 6-KDFE evolution and production module is used for executing the construction and optimization process of a production scheme; the scheme issuing execution module is used for executing a closed loop execution process; and the data feedback updating module is used for acquiring new residual data in the closed loop execution process.

Description

Self-adaptive scheduling method and system for multi-specification pipe stock in discrete scene Technical Field The invention relates to the technical field of enterprise production intelligent management, in particular to a self-adaptive scheduling method and system for multi-specification pipe stock in discrete scenes. Background In the technical field of general mechanical manufacturing, the preparation of multi-specification stainless steel pipes is a key link for supporting the production of chemical equipment, food machinery, medical equipment and the like, and orders under the production scene show the characteristics of three more than one short, namely, the types of specifications are more (usually, the outer diameter range is 3-325 mm, the wall thickness range is 0.5-25 mm), the types of materials are more (usually 304/316L and the like), the orders are more in small batches (usually, the single-specification requirement range is 1-50 pieces), and the delivery cycle is short (usually, 3-10 days). Therefore, how to optimize the production scheduling scheme has a direct impact on the production management and delivery of the enterprise. Aiming at the phenomenon, the current production scheduling is mainly finished by means of manual experience, and adopts a production scheduling mode of 'first to first scheduling' or 'batch scheduling according to specifications', which has the following three main technical problems: 1. the management of the residual materials is coarsely put; In the process, only the length of the residual material is recorded, and key properties such as the material quality, the wall thickness and the like of the residual material are not related, so that a large number of reusable residual materials are idle, and the raw material utilization rate is generally lower than 80%; 2. Incomplete consideration of constraints in production scheduling; The risk of uncombined welding defects and the health state of equipment in the production process lead to the re-working rate of a welded junction exceeding 5 percent and the non-planned downtime period of the equipment exceeding 1 hour; 3. the production scheduling is static and stiff, and cannot cope with dynamic changes such as order insertion, equipment faults and the like, and the order delay rate exceeds 15%. Therefore, aiming at the phenomenon of multi-specification pipe stock in the discrete scene, the current production scheduling mode mainly depends on manual experience, and cannot form fine, standardized and intelligent scientific management. Disclosure of Invention Aiming at the particularity of the multi-specification pipe stock in the discrete scene and the defects of the existing production scheduling technology, the invention provides a self-adaptive production scheduling method based on a genetic modeling-multi-constraint fusion-dynamic evolution optimization-closed loop feedback technology path and a production scheduling system based on the self-adaptive production scheduling method, so as to improve the technical capability of refinement, standardization and intellectualization of enterprise production management. The technical aim of the invention is achieved by the following technical scheme, namely a self-adaptive production scheduling method for preparing materials for pipes with multiple specifications in discrete scenes, which comprises the following technical processes: collecting parameters of the residual materials, constructing a complete attribute file in an RFID mode, and obtaining a real-time gene library of the residual materials; Reading order demand parameters, aligning the standardized demand vector with the dimension of the residue genes, and constructing order demand gene expression by taking priority and delivery date as references to obtain an order queue to be expressed; Based on historical data and real-time equipment data, constructing a welding defect probability prediction model and an equipment health supplement model; Performing scheduling by using the real-time genetic data of the residual materials, the order queue to be expressed, the welding defect probability prediction model and the equipment health supplement model to obtain a scheduling scheme; Performing closed-loop execution in a production scheduling scheme; and updating the model and the gene library according to the closed-loop execution result. Further, the process for obtaining the residue real-time gene bank comprises the following steps: Step 1, acquiring parameters including outer diameter, wall thickness, length, material, furnace batch number, equivalent depth of surface defects, storage coordinates and material taking time of corresponding excess materials in a mode of combining laser measurement with RFID tracing; Step 2, mapping the acquired parameters into a 42bit binary gene sequence to realize standardized coding of the residue information, wherein coding logic discretizes each parameter according to a value range, a