Search

CN-121613857-B - Multi-target cooperative control method for production system

CN121613857BCN 121613857 BCN121613857 BCN 121613857BCN-121613857-B

Abstract

The invention discloses a multi-target cooperative control method for a production system, which comprises the steps of collecting multi-source data of wire harness production and preprocessing, constructing a multi-dimensional structure containing procedures, materials and carrier time phases, establishing a time phase mapping and dynamic evolution mechanism based on the multi-dimensional structure, generating a target relative set, extracting beat, material in-place and carrier circulating signals, constructing a multi-frequency phase set and forming comprehensive deviation signals, constructing a multi-frequency phase-locked group control module, generating three multi-target control instruction sets of beat, materials and carrier, inputting the target relative set and the control instruction set into an improved Longformer model, generating a next period prediction control vector, dynamically updating a plan and path configuration according to the output parameters of the prediction control vector, and rolling and correcting phase-locked parameters. According to the invention, coordinated control optimization of wire harness production beats, materials and carriers is realized through multidimensional time phase modeling and an improved Longformer model.

Inventors

  • SONG QIANG
  • ZHANG GUOQIANG

Assignees

  • 河北农业大学
  • 保定网城软件股份有限公司

Dates

Publication Date
20260508
Application Date
20251210

Claims (9)

  1. 1. A multi-target cooperative control method for a production system, comprising: The method comprises the steps of collecting multi-source production data of a manufacturing execution system in a vehicle enterprise harness production system, preprocessing the multi-source production data, and constructing a multi-dimensional data structure comprising a working procedure dimension, a material dimension, a carrier dimension and a time phase dimension; Based on a multidimensional data structure, establishing a time phase data mapping and dynamic evolution mechanism, mapping the dependency relationship among process nodes, material nodes and carrier nodes into a multidimensional time sequence association matrix, realizing state update through time propulsion and phase gradient analysis, and generating a target time phase association set; Extracting a production takt signal, a material in-place signal and a carrier circulating signal from a production site, respectively defining a production takt phase, a material in-place phase and a carrier circulating phase, constructing a multi-frequency phase signal set, calculating relative deviation among phases, and forming a comprehensive phase deviation signal; Constructing a multi-frequency phase-locked group control module to process the comprehensive phase deviation signal, setting multi-frequency phase-locked parameters, generating a multi-target control instruction set comprising a beat correction instruction, a material release instruction and a carrier scheduling instruction according to the deviation amplitude, and synchronously adjusting the execution period of a production task, the material throwing time sequence and the carrier path; Inputting the target related set and the multi-target control instruction set into an improved Longformer model, extracting time sequence related features by utilizing a local attention and global attention mechanism, and generating a predictive control vector of the next control period; dynamically updating a process plan, a material distribution table and carrier path configuration according to beat adjustment parameters, material release parameters and carrier scheduling parameters output by the predictive control vector, and correcting the multi-frequency phase-locking parameters according to execution feedback after the control period is over.
  2. 2. The method of claim 1, wherein the multi-source production data includes process execution data, material in place data, carrier operation data, equipment operation data, energy consumption monitoring data, and environmental data.
  3. 3. The method according to claim 1, wherein the preprocessing the multi-source production data includes performing time stamp alignment, outlier rejection, missing value completion and normalization processing on the multi-source production data, and creating a unified index according to the production task number and the material lot identification.
  4. 4. The method according to claim 1, wherein the constructing a multi-dimensional data structure including a process dimension, a material dimension, a carrier dimension, and a time phase dimension means performing multi-dimensional association mapping on the preprocessed multi-source production data according to a uniform time index to form a multi-dimensional data matrix with a process number as a master index, a material lot and a carrier number as slave indexes, and a time phase as a dynamic index.
  5. 5. A multi-objective coordinated control method for a production system according to claim 1, wherein said generating an associated set of objectives comprises: On the basis of a multidimensional data structure, a unified index rule is established according to a process identifier, a material batch identifier, a carrier number and a time phase index, and the process state and the material supply under the same time phase and the carrier occupation are recorded and integrated to form a time phase data set taking a four-dimensional index as a main key; The method comprises the steps of taking a time-phase data set as input, constructing a multi-dimensional time-sequence incidence matrix, splicing three types of blocks of a process sequence relation, a material-process requirement relation and a carrier-process occupation relation according to a fixed sequence, establishing a two-channel mapping, using a structure channel for coding static constraint of the three types of relations, using a phase channel for coding time-sequence constraint of the same four-dimensional index between adjacent time phases, and combining two channel results into a single relative view through consistency alignment; Performing service-independent phase change amount extraction on the related view, identifying a change sudden increase section and a phase transition boundary by adopting time phase scanning with fixed step length, generating a candidate update queue based on an identification result, establishing a conflict disposal queue according to the priority order of material shortage conflict, carrier conflict and capability upper limit conflict, and outputting a candidate index set; taking the candidate index set as a drive, and executing dynamic evolution according to a time advancing rule: Firstly, calling boundary condition gating to judge the process capability, material alignment sleeve and carrier occupation, and executing local rearrangement or rollback on records which do not pass the gating; triggering a constraint propagation mechanism, synchronously propagating the influence of rearrangement or rollback to an associated index along a structure channel and a phase channel, and recording an evolution track; And carrying out consistency check and boundary convergence check on the result of the evolution, removing index combinations violating the constraint, and generating a target associated set.
  6. 6. A multi-target cooperative control method for a production system according to claim 1, wherein the forming the integrated phase deviation signal comprises: Setting a fixed sampling period and a reference time, respectively acquiring a production takt signal, a material in-place signal and a carrier circulating signal, and generating an ordered discrete time sequence in each control period; Respectively determining a reference period and a phase reference zero point for each working procedure, each key material and each carrier, wherein the reference period describes a task beat period, a material in-place period and a carrier turnover period, and the phase reference zero points are uniformly aligned with the phase initial positions of the three types of signals; At each moment of the discrete time sequence, converting the three types of signals into a production task beat phase, a material in-place phase and a carrier turnover phase respectively, and carrying out index binding according to a process identifier, a material batch identifier and a carrier number to form a multi-frequency phase signal set; Performing two-phase difference calculation on the multi-frequency phase signal set at the same moment, wherein the calculation comprises phase differences between tasks and materials, phase differences between materials and carriers and phase differences between the carriers and tasks, performing unified dimension processing and weighted synthesis on the three types of phase differences according to a preset weight sequence, and performing envelope folding and outlier rejection on out-of-range phase differences to obtain standard-amount comprehensive phase deviation; And performing time domain smoothing and denoising on the comprehensive phase deviation to generate a stable comprehensive phase deviation signal, and outputting a multi-frequency phase signal set, a two-phase difference result and the comprehensive phase deviation signal.
  7. 7. The method according to claim 1, wherein generating a multi-target control instruction set including a beat correction instruction, a material release instruction, and a vehicle scheduling instruction according to the deviation amplitude comprises: Initializing a multi-frequency phase-locked group control module, wherein the multi-frequency phase-locked group control module consists of a phase diagnosis and priority shaping unit, a phase locking coordination and phase releasing execution unit and an instruction feasible region projection and unification write-back unit, phase locking parameters, dead zone threshold values, control step sizes, capability upper limits, material alignment threshold values and carrier occupation upper limits of all channels are set, and a cross-channel coupling relation table is established to describe the influencing sequence among three phases of production task beats, material in place and carrier turnover; the phase diagnosis and priority shaping unit receives the comprehensive phase deviation signals, performs layered diagnosis on three types of phase differences, generates conflict classification and severity level, constructs a phase credit rating table and remodels treatment priority, and adopts a resonance suppression window to suppress a continuous jitter interval so as to rotationally shift an index combination requiring local rearrangement of a candidate queue record; The phase-locked coordination and phase release execution unit executes fine-granularity stepping correction and controlled phase release on three types of channels in an allowable phase window according to the priority after the remodelling treatment: calculating and outputting a beat correction quantity according to the minimum step and the maximum step for the production task beat channel; calculating and outputting a material release time sequence adjustment quantity for the material feeding channel according to the alignment threshold and the window boundary; calculating and outputting a carrier path correction quantity for the carrier turnover channel according to the occupied upper limit and the path feasibility; When fine granularity stepping cannot meet phase convergence, minimum disturbance rearrangement is carried out on index combinations in a displacement candidate queue, a cross-channel coupling relation table and a phase credit rating table are updated after each adjustment, and three correction amounts are bound and shaped into atomic instruction entries according to indexes; The method comprises the steps that an instruction feasible domain projection and unification write-back unit carries out merging and conflict elimination processing on atomic instruction items according to time phases and resource constraints, sequentially checks an upper limit of capacity, a flush threshold value and an upper limit of occupation, cuts down or delays to take effect on the items which do not meet the constraints according to channel priorities, sets a rollback check point and an effective time window for the items which pass the checking, and generates a multi-target control instruction set comprising a beat correction instruction, a material release instruction and a carrier scheduling instruction in a combined mode; and writing the multi-target control instruction set back to the target associated set, and synchronously adjusting the execution period of the production task, the material throwing time sequence and the carrier running path.
  8. 8. A multi-target cooperative control method for a production system according to claim 1, wherein the generating a predictive control vector for a next control period includes: Initializing an improved Longformer model, wherein the improved Longformer model consists of a channel alignment and index coding module, a cross-channel sparse-global mixed attention module and an instruction memory and constraint coordination module, and establishes input and output interfaces of a phase correlation set, a comprehensive phase deviation signal and a multi-target control instruction set with a target; The channel alignment and index coding module performs time axis serialization processing on the historical state of the target associated set, the historical sequence of the comprehensive phase deviation signal and the multi-target control instruction set, and adds a channel identifier, a position identifier and an event identifier to a production task beat channel, a material in-place channel and a carrier turnover channel respectively, so as to generate an aligned multi-channel input sequence according to a unified index rule; the cross-channel sparse-global mixed attention module executes local dependency modeling on the multi-channel input sequence aligned in the fixed local window, and establishes cross-channel global attention connection by taking phase mutation points, material shortage events and carrier conflict events as global anchor points, and outputs a context representation sequence containing short-period fluctuation and long-time dependency; the instruction memory module receives the context expression sequence, invokes recent execution record and feedback information of a multi-target control instruction set, plays back and performs association analysis on historical instructions of three types of control channels of beats, materials and carriers, generates candidate prediction control results, and calculates a prediction adjustment value and corresponding confidence coefficient for each control channel respectively; And performing feasible domain checking and unification on the candidate predictive control vectors to generate predictive control vectors comprising beat adjustment parameters, material release parameters and carrier scheduling parameters, and forming the predictive control vectors, channel-level confidence level, constraint suggestions and attention weight summaries into a feedforward output packet.
  9. 9. The method of claim 1, wherein dynamically updating the process plan, the material distribution table, and the carrier path configuration comprises: Receiving a predictive control vector, mapping beat adjustment parameters, material release parameters and carrier scheduling parameters into beat correction amounts, material release time sequence adjustment amounts and carrier path correction amounts respectively, assembling into an execution instruction packet, performing compliance check according to an upper capacity limit, a complete threshold and an upper occupation limit, and setting an instruction effective time window and a rollback check point; dynamically updating a process plan, a material distribution table and carrier path configuration according to an execution instruction packet passing through checking, generating a state snapshot of a relevant set of targets in a current control period, marking an affected process index, a material batch index and a carrier number, and setting an execution monitoring point to acquire an actual result; the method comprises the steps of collecting an actual beat adjustment result, an actual material release result and an actual carrier scheduling result in a current control period, and comparing the actual beat adjustment result, the actual material release result and the actual carrier scheduling result with expected results when an instruction is effective respectively to form an execution deviation record divided by beat channels, material channels and carrier channels; and carrying out rolling correction on the multi-frequency phase-locked parameters and the boundary conditions according to the execution deviation record, and outputting the updated parameter snapshot and the corresponding set state snapshot of the target as the input of the next control period.

Description

Multi-target cooperative control method for production system Technical Field The invention relates to the technical field of industrial automation and intelligent manufacturing control, in particular to a multi-target cooperative control method for a production system. Background At present, a vehicle enterprise harness production system generally adopts a production scheduling and process control mode based on a Manufacturing Execution System (MES), and the visualization and semi-automation of a production process are realized by carrying out informatization management on a process plan, material distribution and carrier operation. However, most of the existing MES systems use single-target control logic as a core, mainly focus on optimization of production beats or process sequences, and lack dynamic cooperative control capability on material flow and carrier turnover. In the actual production environment with multiple parallel processes, multiple material batch crossing and frequent carrier circulation, the single control mode is difficult to cope with the problems of beat fluctuation and resource conflict, local bottleneck and beat drift are easy to cause, and the overall stability and response speed of the production line are reduced. In the prior art, partial improvement schemes attempt to introduce a time sequence prediction or rule optimization method to identify potential risks such as production delay and material shortage in advance, but most of the improvement schemes are based on static characteristics or linear prediction models and lack dynamic modeling capability of a multidimensional time sequence relation. The control instructions are issued according to fixed priority, and self-adaptive coordination among beat control, material release and carrier scheduling cannot be realized, so that time dislocation exists between a predicted result and an execution action, and multi-target synchronous control and rolling optimization in a real sense cannot be realized. The existing production control system mostly adopts centralized scheduling logic, and has larger delay in information feedback and parameter correction. When the production state is suddenly changed (such as material delay, carrier congestion and beat abnormality), the system often needs manual intervention or waits for the next scheduling period, and real-time self-adaptive adjustment cannot be realized. The beat instability, material retention and carrier conflict caused by the method not only affect the production efficiency, but also make the system energy consumption and the resource utilization rate difficult to optimize. Therefore, how to provide a multi-objective cooperative control method for a production system is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention provides a multi-target cooperative control method for a production system, which combines time phase mapping modeling, multi-frequency phase-locked group control and improved Longformer predictive control technologies, constructs a dynamic cooperative control mechanism oriented to process beats, material delivery and carrier scheduling, realizes self-adaptive coordination and rolling optimization of the production beats, material flows and carrier paths, realizes real-time sensing, prediction and closed-loop adjustment of a multi-target control process of the production system through multi-dimensional time phase data modeling and intelligent prediction, and has the advantages of high beat stability, high response speed and strong self-learning capability of the system. According to an embodiment of the invention, a multi-target cooperative control method for a production system comprises the following steps: The method comprises the steps of collecting multi-source production data of a manufacturing execution system in a vehicle enterprise harness production system, preprocessing the multi-source production data, and constructing a multi-dimensional data structure comprising a working procedure dimension, a material dimension, a carrier dimension and a time phase dimension; Based on a multidimensional data structure, establishing a time phase data mapping and dynamic evolution mechanism, mapping the dependency relationship among process nodes, material nodes and carrier nodes into a multidimensional time sequence association matrix, realizing state update through time propulsion and phase gradient analysis, and generating a target time phase association set; Extracting a production takt signal, a material in-place signal and a carrier circulating signal from a production site, respectively defining a production takt phase, a material in-place phase and a carrier circulating phase, constructing a multi-frequency phase signal set, calculating relative deviation among phases, and forming a comprehensive phase deviation signal; Constructing a multi-frequency phase-locked group control module to process the comprehensi