Control method of wire harness production equipment, equipment and medium

Abstract

The invention discloses a control method of wire harness production equipment, and relates to the technical field of intelligent manufacturing. The method comprises the following steps: acquiring preset process parameter curves of all N procedures in a complete wire harness production process; inputting all the process parameter curves and the estimated environment parameters into a virtual production system integrated with a defect-process correlation model; executing continuous production simulation from the first process to the last process in the virtual system, and generating virtual product quality evaluation based on the output of the last process; and according to an evaluation result, if the evaluation result is qualified, issuing all the process parameter curves to physical equipment to execute production, and if the evaluation result is not qualified, adjusting the process parameter curves and re-simulating until the evaluation result is qualified. According to the invention, through full-process virtual simulation and beforehand quality prediction, conversion from post-process detection to beforehand prevention is realized, the quality and efficiency of wire harness production are effectively improved, and meanwhile, the production trial and error cost is remarkably reduced through a closed-loop optimization mechanism.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and industrial process control, and in particular to a control method, equipment and medium for wire harness production equipment. Background Technology

[0002] As the electrical connection hub and power / signal transmission carrier in products such as automobiles and electrical appliances, the quality of the wire harness directly determines the safety and reliability of the entire product. The wire harness manufacturing process involves multiple precision processes such as crimping, welding, and injection molding. Deviations in any process parameters can lead to defects such as cable damage, terminal deformation, and insulation burrs. Traditional production quality control heavily relies on two stages: first, manual or rule-based machine vision inspection at the end of production, which is inefficient, prone to missed inspections, and unable to trace the root cause of defects; second, after defects are discovered, experienced engineers adjust equipment parameters, a passive, reactive control strategy that cannot provide early warning or prevention before defects occur.

[0003] With the development of intelligent manufacturing technology, digital twin and simulation technologies have been introduced into the industrial production field. However, current applications in wire harness production are mostly limited to product design verification or static process flow simulation. These traditional simulation systems have two major limitations: First, they are usually open-loop simulations based on ideal environment assumptions, failing to fully consider the real-time impact of dynamic environmental factors such as temperature, humidity, and equipment vibration in the actual production workshop on the process results. This leads to deviations between simulation results and actual production results, limiting their reference value. Second, even if the simulation results are ideal, once the process parameters are sent to the equipment and production begins, there is a lack of effective real-time monitoring and dynamic adjustment mechanisms. When equipment deviates significantly from the preset process curve due to aging, malfunction, or sudden environmental interference, the system cannot autonomously detect and correct this deviation in time, resulting in the continued production of defective products until the next quality inspection stage, causing waste of materials and time.

[0004] Therefore, there is a significant gap in the existing technology: a lack of a control scheme for wire harness production equipment that can make high-confidence predictions of the process plan before production execution and achieve closed-loop adaptive control during production. This is precisely the core technical problem that this invention aims to solve. Summary of the Invention

[0005] To achieve the objectives of this invention, the technical solution adopted is as follows: According to a first aspect of the present invention, a control method for a wire harness production equipment is provided, the method comprising the following steps: S100, sequentially obtain the preset process parameter curves for all N processes involved in a complete wire harness production process, where N is an integer greater than 1.

[0006] S200, input the preset process parameter curves and estimated environmental parameters of all N processes into the virtual production system; wherein, the virtual production system is constructed in virtual space to map the complete physical production line and simulate the continuous production process from the first process to the last process, and integrates a defect-process correlation model established based on historical production data.

[0007] S300, Execute continuous production simulation from the first process to the last process in the virtual production system, and generate a virtual product quality assessment based on the simulation results output by the last process.

[0008] S400, in response to the virtual product quality assessment result being the first result, sends the process parameter curves of all N processes to the corresponding physical wire harness production equipment.

[0009] In step S500, in response to the virtual product quality assessment result being the second result, one or more of the process parameter curves for all N processes are adjusted, and steps S200 to S300 are repeated until the virtual product quality assessment result is the first result. According to a second aspect of the invention, an electronic device is provided, including a processor and a memory; the processor executes the steps of the method described in the first aspect of the invention by calling programs or instructions stored in the memory.

[0010] According to a third aspect of the present invention, a computer-readable storage medium is provided that stores a program or instructions that cause a computer to perform the steps of the method described in the first aspect of the present invention.

[0011] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves a fundamental shift from traditional post-production detection and remediation to pre-production prediction and optimization by inputting preset process parameters and estimated environmental parameters into a virtual production system that integrates a defect-process correlation model for pre-production simulation. This shift enables accurate prediction and avoidance of defects before physical production, while automatically optimizing process parameters through simulation feedback. This not only significantly improves product quality but also fundamentally reduces trial-and-error costs and resource waste, achieving simultaneous improvement in quality control and production efficiency. Attached Figure Description

[0012] Figure 1 A flowchart illustrating a control method for a wire harness production equipment provided in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example: This invention provides a control method for wire harness production equipment, applicable to manufacturing workshops producing various cable assemblies such as automotive wire harnesses, home wiring harnesses, and communication equipment wire harnesses. Specifically, this method is deployed in the production line control system of the workshop for precise control of specific production equipment on the line. Figure 1 As shown, the method includes the following steps: S100, sequentially obtain the preset process parameter curves for all N processes involved in a complete wire harness production process, where N is an integer greater than 1.

[0015] Specifically, this step retrieves a complete set of preset process parameter curves from the upper-level production management system or process database, following the product process route, for the specific model of wiring harness currently being produced (e.g., automotive engine wiring harness), from the first process to the last process.

[0016] In this embodiment of the invention, the complete set of preset process parameter curves constitutes a complete process scheme, wherein each curve corresponds to a specific process step and is used to describe the set trajectory of one or more key physical quantities (including but not limited to current, voltage, temperature, pressure, speed, and displacement) changing over time in that process step. This complete set of preset process parameter curves will serve as the benchmark input for subsequent global simulation in the virtual production system.

[0017] In this embodiment of the invention, the total time corresponding to the preset process parameter curves for executing all N processes is defined as a total product production cycle, that is, the total standard operating time corresponding to the complete execution of the preset process parameter curves for all N processes from the start of the first process to the end of the last process. This time serves as a common time benchmark for virtual simulation and environmental prediction.

[0018] S200, input the preset process parameter curves and estimated environmental parameters of all N processes into the virtual production system.

[0019] In this embodiment of the invention, the virtual production system is a virtual simulation environment built based on digital twin technology. The virtual production system is constructed in a virtual space to map a complete physical production line and simulate a continuous production process from the first to the last process. It also integrates a defect-process correlation model established based on historical production data. The specific construction process includes: First, 3D modeling and mechanism modeling are performed on the production line in the physical workshop to create virtual equipment models that correspond one-to-one with physical production equipment (such as crimping machines and welding machines). These models can respond to input process parameters and simulate their operating status and output.

[0020] Secondly, based on the actual material flow logic in the virtual space, these virtual equipment models are integrated into a complete virtual production line that can simulate the real production process.

[0021] Finally, the defect-process correlation model was integrated into the virtual production line as an analysis engine. This gave the virtual production system three main functions: physical mapping, process simulation, and quality prediction.

[0022] The defect-process correlation model is constructed by analyzing historical production data to establish the correlation between process parameters, environmental parameters, and wire harness defect types throughout the entire process from the first to the last stage. The historical production data includes historical process parameter data and historical defect data of wire harness products produced under those historical process parameter data. The defect types in the wire harness defect data can be obtained by automatically identifying and labeling historical product image data using image recognition methods (such as deep learning-based OCR visual inspection systems) or large model analysis methods. Specifically, the construction process of the defect-process correlation model includes: Step 1: Preparation of training data and feature engineering (1) Data preparation: Collect historical production data covering multiple production batches and a sufficiently long time span. This data constitutes a dataset, in which each record contains: Feature (input X) Full-process process parameters: Extract features from the process parameter curves of all N processes to obtain process parameter curve features. The extracted features may include, but are not limited to: the maximum, minimum, and average values ​​of key physical quantities (such as current and pressure) for each process, the curve slope, area (integral), and other statistical features, or directly use equally spaced sampled sequence data as input.

[0023] Environmental parameters: Actual environmental data of each process during the production of this wire harness (such as average temperature, humidity fluctuation range, and maximum vibration amplitude).

[0024] Label (output Y): Final product defect data: The final quality inspection results for each production record (e.g., a wiring harness serial number). These results are obtained through the following methods: Automated labeling: Using image recognition systems (such as deep learning-based OCR and visual defect detection models) deployed at quality inspection stations, historical product images are automatically analyzed to identify and label defect types (such as "terminal deformation", "insulation layer damage", "burrs") and their severity.

[0025] Large model-assisted annotation: Utilize large language models (LLM) or multimodal large models to analyze unstructured text data related to the product, such as maintenance reports and quality inspector logs, extract and summarize defect information, and associate it with production records.

[0026] Manual review: The automatic labeling results are sampled and reviewed to ensure the accuracy of the labels.

[0027] (2) Data alignment and cleaning: Based on timestamps and unique product serial numbers, data from different systems (MES, sensors, quality inspection systems) are precisely aligned to ensure that the process, environment and defect results in each record correspond one-to-one. Then, data cleaning is performed to handle missing values ​​and outliers.

[0028] Step 2: Model Selection and Training (1) Model selection: This problem is a typical multivariate, nonlinear classification or regression problem. Therefore, machine learning models suitable for this type of task are preferred, such as: Gradient boosting decision trees (such as XGBoost and LightGBM) have the advantage of strong feature importance ranking ability, can effectively handle tabular data, and have relatively good model interpretability.

[0029] Deep learning models (such as fully connected neural networks and Transformer encoders) have the advantage of automatically learning complex high-order feature interactions, making them particularly suitable for processing high-dimensional sequence inputs (such as raw process curves).

[0030] (2) Model training: Training objective: Using prepared features X as input, train the model to predict the final defect label Y. For multiple defect types, a multi-task learning model or a multi-classification model can be constructed.

[0031] Training process: The dataset is divided into training, validation, and test sets. The model is trained using the training set data, and the model parameters are adjusted using optimization algorithms such as backpropagation. Model performance is monitored on the validation set to prevent overfitting.

[0032] Step 3: Model Integration and Deployment (1) Interpretability analysis: After training, the contribution of each input feature (such as "maximum pressure in the second process" and "ambient temperature in the fifth process") to the prediction of specific defects is analyzed using built-in model tools (such as decision trees) or post-process analysis tools (such as SHAP analysis). This step transforms the "black box" model into usable process knowledge and clarifies "which processes are related to the defects".

[0033] (2) Model Deployment: The trained and qualified model is solidified into an analysis engine and integrated into the virtual production system. When new process curves and environmental parameters are input, the defect-process correlation model can output the predicted probability of the wire harness defect type in real time.

[0034] In this embodiment of the invention, the estimated environmental parameters are the environmental parameters predicted for a future period of the aforementioned production cycle, and the acquisition method includes: S201, collect current environmental data and historical environmental data.

[0035] In this embodiment of the invention, the environmental data includes at least one of temperature, humidity, and vibration data.

[0036] The current environmental data refers to the latest or near real-time (e.g., second- or minute-level delay) environmental parameter readings collected by a sensor network (such as temperature and humidity sensors, vibration accelerometers) deployed around key equipment on the production line.

[0037] Historical environmental data refers to continuous environmental parameter records stored in a time-series database, spanning from a certain point in the past to the present. The duration of historical environmental data should cover a sufficient number of complete production cycles to ensure the capture of typical patterns of environmental fluctuations (such as diurnal temperature variations, weekday-weekend differences, and the impact of different seasons). As a non-limiting example, the time span of the historical environmental sample data is typically no less than four weeks, preferably three months or longer. This duration provides sufficient data for time-series forecasting models to learn patterns while avoiding data that is too outdated to reflect recent environmental characteristics. In practical applications, this duration can be configured according to the specific production line's rhythm and the degree of environmental stability.

[0038] Before deriving the model from the input environment, the current and historical environmental data usually need to undergo a series of preprocessing steps, including but not limited to: data cleaning (removing obvious sensor false alarms or outliers caused by communication anomalies), data alignment (unifying sensor data with different sampling frequencies to the same timestamp), and data standardization, in order to improve the convergence speed and prediction accuracy of model training.

[0039] S202, the current environmental data and historical environmental data are input into a pre-trained environmental inference model, and the estimated environmental parameters are output. Preferably, the environmental inference model is constructed using a temporal prediction algorithm (such as Long Short-Term Memory (LSTM), Temporal Convolutional Network (TCN), or Transformer). The training of this environmental inference model is based on a large amount of historical environmental sample data. The architecture of the environmental inference model includes at least an input layer, a hidden layer for capturing temporal dependencies, and an output layer. It learns the temporal dependencies in the environmental data through training to output predictions of environmental parameters for a specific future time period.

[0040] The training process of this environmental inference model is as follows: The collected, preprocessed historical environmental sample data is divided into training and validation sets in chronological order. Environmental data from a continuous historical time window is used as input, and environmental data from a subsequent production cycle is used as the prediction target. The model parameters are trained by minimizing the error (e.g., mean squared error) between the predicted and actual values. The trained model can capture the time-dependent dynamic patterns in the environmental data.

[0041] To ensure the long-term accuracy of the model, a model update mechanism is also set up, which is triggered under two modes: Periodic updates: Performed at preset time intervals (e.g., weekly).

[0042] Triggered update: Starts when the predictive performance of the environment-derived model systematically degrades.

[0043] Specifically, the criteria for triggering updates are as follows: when the average prediction error of the environmental inference model exceeds the set dynamic error threshold in K consecutive prediction periods (e.g., K=3), it is determined that the environmental inference model can no longer accurately reflect the current environmental patterns, and the update process is then triggered.

[0044] The dynamic error threshold is set as follows: After the environmental derivation model completes its last deployment or update, the prediction errors generated by the model during the subsequent stable operation phase (e.g., continuous normal operation for at least 24 hours or completion of at least 10 full production cycles) are recorded. The mean (μ) and standard deviation (σ) of all errors during this phase are calculated, and the dynamic error threshold is set based on this (e.g., set as μ + nσ, where n = 2). This dynamic error threshold characterizes the upper limit of the prediction error fluctuation of the environmental derivation model under normal conditions.

[0045] Once the update mechanism is triggered, the model parameters will be fine-tuned immediately using the newly collected environmental data, enabling it to adapt to the slow time-varying characteristics of the workshop environment and restore and maintain prediction accuracy.

[0046] It should be noted that, in this embodiment of the invention, the prediction duration of the estimated environmental parameters output by the environmental derivation model can be flexibly configured according to specific simulation requirements. The prediction duration can cover a complete product production cycle, thus providing a consistent environmental background for global simulation; or it can be specifically set to the production cycle of a particular process to achieve more refined environmental simulation of key processes. This flexibility allows the virtual production system to select the most suitable environmental prediction granularity based on the accuracy and efficiency requirements of different scenarios, thereby optimizing the utilization of computing resources while ensuring simulation confidence.

[0047] S300, a production simulation is performed in the virtual production system, and a virtual product quality assessment is generated based on the simulation results. The virtual product quality assessment includes a prediction of the type of wire harness defect.

[0048] Specifically, this step is the core execution process of the virtual production system. The virtual production system loads the complete set of process parameter curves and estimated environmental parameters input from S200, and initiates continuous dynamic simulation from the first process to the last process. Its specific process includes: Process-by-process simulation: The virtual production system drives the corresponding virtual equipment models to run according to the process sequence. Each virtual equipment model simulates the physical processing effect of the process based on its input process parameter curves (such as current and pressure curves) and the current virtual environment parameters (such as temperature and humidity). For example, in the crimping process, the virtual equipment model calculates the deformation and stress of the terminals under specific pressure and current; in the injection molding process, the virtual equipment model simulates the flow and curing of plastic at specific temperatures.

[0049] State Transfer and Accumulation: After each process simulation is completed, the virtual production system updates the state attributes of the processed wire harness components. These states include not only geometric morphology (such as size and shape) but also physical states (such as residual stress and changes in material properties). These states serve as the initial conditions for the next process simulation, thereby realizing the accumulation and transfer of processing effects and accurately simulating the streamlined operations in real production.

[0050] Defect Prediction and Assessment Generation: After the entire simulation process is completed, the virtual simulation system calls the integrated defect-process correlation model to comprehensively evaluate the final virtual wire harness product. This defect-process correlation model takes the entire process execution data (such as key features of parameter curves for each process step), environmental interference data, and cumulative state data of the wire harness as input. Through the correlation rules embedded in the defect-process correlation model, it calculates and outputs a virtual product quality assessment. This assessment not only includes an overall pass / fail determination (first result or second result), but also specifically includes predictions of possible defect types (e.g., "Insufficient tensile strength at terminal crimping: 85%", "Risk of burrs in insulation layer: High"), as well as key process traceability information leading to the defect.

[0051] S400, in response to the virtual product quality assessment result being the first result, sends the preset process parameter curve to the corresponding physical wire harness production equipment to control the physical wire harness production equipment to operate according to the preset process parameter curve.

[0052] In this embodiment of the invention, the first result represents a qualified virtual product quality assessment conclusion. Specifically, this means that the simulation results in the virtual production system show that, under the combined effect of the currently set complete set of process parameter curves and the estimated environmental parameters, the produced virtual wire harness product can meet the preset quality standards and technical specifications in terms of geometric dimensions, physical properties, and electrical characteristics, and there is no predicted defect risk.

[0053] When the evaluation result is determined to be the first result, it signifies that the currently set complete set of process parameter curves has passed virtual verification and is ready for actual production. Subsequently, the preset process parameter curves for all N processes are distributed sequentially and individually to the local controllers of the corresponding physical production equipment (such as pressing machines, welding machines, and injection molding machines) via the workshop network. Upon receiving the instructions, each physical production device will start according to the preset timing sequence and strictly follow the distributed process curves, thereby reproducing the manufacturing process that has been verified and qualified in the virtual environment in the physical world, achieving a precise mapping of the digital twin from virtual to real.

[0054] S500, in response to the virtual product quality assessment result being the second result, initiate the process parameter optimization process, adjust the preset process parameter curves for all N processes, and repeat S200 to S300 until the virtual product quality assessment result is the first result, and then issue the finally determined adjusted process parameter curves.

[0055] In this embodiment of the invention, the second result represents a virtual product quality assessment conclusion of non-compliance, i.e., a prediction that the product has one or more defects. At this point, an automated, feedback-based iterative optimization loop will be initiated.

[0056] Furthermore, adjusting the preset process parameter curve specifically includes: S501, based on the wire harness defect type in the virtual product quality assessment, locate one or more key processes that have the greatest impact on the generation of the defect type based on the defect-process correlation model.

[0057] First, the virtual product quality assessment results are analyzed to identify specific defect types (e.g., "insufficient crimping tensile strength"). Then, a reverse tracing analysis is performed based on the defect-process correlation model. This model outputs the contribution or sensitivity of process parameters and environmental parameters of different processes to the current defect type. Based on this, one or more key processes with the greatest impact on the generation of this defect type are identified.

[0058] S502, according to the preset optimization algorithm, the process parameter curves of the key process are iteratively corrected first.

[0059] A pre-defined optimization algorithm is used to prioritize and iteratively correct the process parameter curves of the identified key processes. The optimization algorithm includes gradient descent, genetic algorithms, and Bayesian optimization. The optimization objective is to transform the virtual product quality assessment result from a secondary result to a primary result, while simultaneously satisfying all process constraints. The specific steps of each iterative correction include: Parameter adjustment: Based on the scheme generated by the optimization algorithm, modify specific parameters in the key process curves (e.g., increase the peak current of the crimping process or extend the duration of the welding process).

[0060] Resimulation: Input the adjusted complete set of process curves and the original estimated environmental parameters back into the virtual production system, and repeat S200 to S300 to conduct a new round of simulation and evaluation.

[0061] Result Judgment: Determine whether the new evaluation result is the first result. If yes, exit the iteration loop; otherwise, the optimization algorithm will generate the next adjustment scheme based on the new simulation result and continue iterating.

[0062] This cycle continues until the virtual product quality assessment results meet the passing standard, or until the preset maximum number of iterations is reached.

[0063] Finally, the optimized and adjusted process parameter curves are sent to the physical production line.

[0064] In this embodiment of the invention, during the iterative optimization process described in S500, the estimated environmental parameters remain fixed. The technical consideration is as follows: First, the goal of this optimization process is to find the optimal combination of process parameters under the expected future environmental conditions. Fixing the environmental parameters is equivalent to setting a clear and invariant boundary condition for the optimization algorithm, thereby ensuring that the optimization process can converge stably and efficiently to an optimal or satisfactory solution for the expected environment.

[0065] Secondly, from an engineering practice perspective, performing multiple process iterations based on the same environmental prediction yields optimization results with better consistency and comparability. Dynamically updating the environmental prediction in each iteration can cause the optimization algorithm to oscillate due to the constantly changing baseline of the objective function, making convergence difficult and incurring unnecessary computational overhead.

[0066] Ultimately, the process scheme obtained through prediction using fixed environmental parameters is essentially a verification and optimization of its robustness under the expected environmental conditions. This ensures that when the proposed process scheme is executed, as long as the actual environment does not significantly deviate from the prediction range, the scheme can stably produce qualified products.

[0067] Furthermore, following S400, the method further includes: The S410 collects the actual process parameter curves of the physical production equipment during the operation of each process in real time.

[0068] Sensors and data acquisition units deployed on each physical production device are used to continuously and in real time collect the actual process parameter curves for each step of the production process. These actual process parameter curves and preset process parameter curves have the same data structure and sampling frequency to ensure comparability. The collected raw data undergoes preprocessing such as noise reduction and filtering to form standard time-series data for comparison.

[0069] S420 compares the actual process parameter curve of each process with the preset process parameter curve of the issued process in real time to obtain the corresponding comprehensive deviation value.

[0070] The actual process parameter curves collected in real time for each process step are compared with the preset process parameter curves for that process step at a high frequency. The high frequency refers to a frequency that matches the control cycle and data acquisition capability of the production equipment. As a non-limiting example, this frequency is no less than 10Hz, that is, data sampling and comparison are performed every 100 milliseconds. This frequency is sufficient to capture the dynamic changes in process parameters caused by actuator drift or sudden interference in most industrial equipment (such as crimping machines and welding machines), while avoiding excessive computational load.

[0071] The comparison is based on a multi-dimensional comprehensive calculation of the similarity of curve shapes, the consistency of key feature points (such as peaks, valleys, and integral areas), and the overall deviation, so as to obtain a quantitative comprehensive deviation value.

[0072] In this embodiment of the invention, the overall deviation value satisfies the following condition: Dc = α·DTW norm +β·Df norm +γ·IED norm .

[0073] Among them, DTW norm To normalize the morphological similarity deviation, the morphological similarity deviation can be obtained by calculating the DTW distance between the actual process parameter curve and the preset process parameter curve using the dynamic time warping algorithm. The larger the DTW distance, the greater the morphological difference.

[0074] Df norm The key feature deviation (Df) is obtained by extracting key feature points from the preset process parameter curves, calculating the deviations of the actual process parameter curves at the corresponding points, and then performing a weighted sum. The formula for calculating Df can be expressed as: Df=w1·|A peak -P peak |+w2·|A valley -P valley |+w3·|S A -S P |+w4·|Area A -Area P |, where A peak P represents the peak value of the actual process parameter curve. peak A represents the peak value of the preset process parameter curve. valley P represents the valley value of the actual process parameter curve. valley S represents the valley value of the preset process parameter curve. A S represents the average slope of the actual process parameter curve. P The average slope of the preset process parameter curve, Area A Area is the integral area of ​​the actual process parameter curve. P The integral area of ​​the preset process parameter curve is w1 to w4, which are the preset weighting coefficients of the characteristic deviations.

[0075] IED norm The normalized overall deviation is obtained by calculating the overall degree of deviation between the actual process parameter curve and the preset process parameter curve at each sampling point. Specifically, the entire production cycle is divided into n consecutive sampling times at fixed time intervals, and the actual process parameter value A is recorded simultaneously at each sampling time i. iand preset process parameter value P i The overall deviation is calculated as follows: First, the difference between the actual process parameter value and the preset process parameter value at each sampling time is calculated and squared. Then, the squared differences at all sampling times are summed. Finally, the summation result is squared. This calculation method can comprehensively reflect the overall deviation of the two curves over the entire time dimension, effectively complementing other deviation measurement methods that only focus on local feature points.

[0076] In this embodiment of the invention, the normalized morphological similarity deviation, the normalized key feature deviation, and the normalized overall deviation are equal to the morphological similarity deviation, the key feature deviation, and the overall deviation divided by their respective preset maximum allowable values.

[0077] α, β, and γ are the fusion weight coefficients for normalized morphological similarity deviation, normalized key feature deviation, and overall deviation, respectively, and satisfy α+β+γ=1. Each fusion weight coefficient and the preset maximum allowable value can be configured according to the accuracy requirements of the specific process. More stringent parameter settings can be used for critical processes or sensitive process sections.

[0078] S430: When the combined deviation between the actual process parameter curve and the preset process parameter curve of any process exceeds the preset threshold corresponding to that process, an equipment abnormality signal is generated and the subsequent process adjustment process is triggered.

[0079] The calculated real-time comprehensive deviation value for each process step is compared with the preset threshold corresponding to that process step. The preset threshold takes into account the process accuracy requirements of different processes. For critical processes (such as terminal crimping) or sensitive sections in the process curve (such as the heating stage), more stringent thresholds are set. When the comprehensive deviation value between the actual curve and the preset curve of any process step continuously exceeds the preset threshold corresponding to that process step for a preset stable duration (e.g., three consecutive production cycles), it is determined that the equipment operation has experienced abnormal drift. An equipment abnormality signal containing the location of the abnormal process, the magnitude of the deviation, and a timestamp is immediately generated, and subsequent process adjustment procedures are automatically triggered.

[0080] In one illustrative embodiment, the preset threshold T of the j-th process... j The following conditions can be met: T j =T jbase ×K j ×S j , among which, T jbase K is the baseline threshold for the j-th process, K is the 95th quantile of the historical comprehensive deviation set for the j-th process, and K is the threshold value for the j-th process. j The criticality coefficient for the j-th process is set according to the importance of the process: if it is a critical process (such as terminal crimping, laser welding), take K.j =0.8, if it is an important process (such as wire stripping, insulation treatment), take K. j =0.9, if it is a general process (such as label printing, bundling and packaging), take K. j =1.0. S j The sensitivity correction factor for the j-th process step is set according to the characteristics of the process segment: if it is a core process segment (such as the heating stage or pressure holding stage), S j =0.7, if it is a transitional process section (such as a temperature rise / fall transition section), S j =0.7, if it is a stable process section, S j =1.0, where j ranges from 1 to m, and m is the number of processes.

[0081] It should be noted that the specific values ​​of the criticality coefficient and sensitivity correction factor described in this embodiment are merely illustrative examples and can be adjusted according to the accuracy requirements of different production scenarios in actual applications. This hierarchical dynamic configuration method ensures both the scientific validity and adaptability of the threshold setting, and guarantees the monitoring accuracy requirements of key processes and core technology segments.

[0082] This monitoring mechanism ensures that equipment performance degradation or process execution deviations can be detected early in production, providing accurate and timely decision-making basis for subsequent precise rolling optimization.

[0083] Furthermore, after triggering the subsequent process adjustment procedure, it also includes: S440 acquires the actual environmental data and the operating status of all physical production equipment at the trigger time.

[0084] When an abnormal signal is triggered by the device, comprehensive real-time data at the moment of triggering is immediately collected and recorded. Specifically, this includes: Actual environmental data: The measured values ​​of environmental parameters such as temperature, humidity, and vibration intensity are obtained through the workshop environmental sensor network; Equipment operating status: Collect real-time operating parameters of all physical production equipment, including but not limited to key indicators that can reflect the instantaneous status of the equipment, such as the current working mode of each piece of equipment, the position of the actuator, the motor torque, and the temperature controller reading. Work-in-process status: Obtain the actual status data of the semi-finished wire harness currently being processed in the Mth process, such as the completed processing dimensions and accumulated thermal history.

[0085] S450, lock the process number M where the deviation occurred.

[0086] Based on the abnormal process information contained in the equipment anomaly signal, the process number M where the parameter deviation occurred is clearly identified. Here, process number M refers to the first process in the entire N-process production flow where a process parameter deviation was actually detected. M is used as the dividing point for process adjustment to ensure that subsequent optimizations only target processes that have not yet been executed.

[0087] S460: Using the current actual state as the initial condition, only for the process parameter curves of process M+1 to process N, repeat the simulation optimization steps of S200 to S500 to generate and issue a new set of process parameter curves for subsequent processes.

[0088] Using the real-time production status snapshot captured by S440 as the initial condition, the virtual simulation and optimization steps of S200 to S500 are re-executed only for the subsequent process parameter curves from process M+1 to process N. This rolling optimization process is characterized by: Optimization scope limitation: Keep the processes that have been actually executed in the first M steps unchanged, and only re-plan the subsequent processes; State inheritance: The current actual state of the work-in-process is used as the initial input condition for the (M+1)th process in the virtual simulation; High-efficiency iteration: Rapidly simulate multiple adjustment schemes in the virtual production system until a combination of subsequent process parameters that can make the final product quality assessment meet the qualified standard is found; Dynamic distribution: The optimized new set of process parameter curves for subsequent processes is distributed to the corresponding physical production equipment in real time, replacing the corresponding parts in the original process plan.

[0089] Through this mechanism, the present invention achieves online, dynamic, and localized adjustment of process parameters, enabling rapid response to production anomalies without downtime, thus forming a complete adaptive production closed-loop control. This scheme, by dynamically adjusting the process parameters of subsequent processes, achieves technical compensation for production anomalies and optimization of manufacturing resources. Specific technical effects are as follows: When an abnormal process parameter is detected in a certain process step during production, the processing state of the preceding process is inherited, and the parameters of the subsequent process are re-optimized. This achieves the technical effect of processing state inheritance and adaptive parameter correction. Based on the simulation and deduction capabilities of the virtual production system, subsequent process schemes that can compensate for the deviations of the preceding process can be accurately calculated, thus constructing a complete closed loop of adaptive parameter correction technology.

[0090] This anomaly localization and subsequent compensation adjustment mechanism significantly improves the fault tolerance and process adaptability of the manufacturing system while ensuring that the final product meets quality standards. By dynamically updating parameters only in subsequent processes, it maintains the continuity of the production flow and optimizes the energy efficiency of the manufacturing process, demonstrating the technological advantages of modern intelligent manufacturing systems in dealing with process fluctuations. This solution effectively enhances the robustness and intelligence level of the production system by constructing a complete technology chain from anomaly detection to parameter adaptation.

[0091] This invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being configured to perform the method described in this invention.

[0092] This invention also provides a computer-readable storage medium storing computer-executable instructions for performing the methods described in this invention.

[0093] The embodiments disclosed herein are preferred embodiments, but are not limited thereto. Those skilled in the art can readily grasp the spirit of the present invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of the present invention, they are all within the protection scope of the present invention.

Claims

1. A control method of a wire harness production apparatus, characterized by, The method comprises the following steps: S100, sequentially obtaining preset process parameter curves of all N processes involved in a complete wire harness production process, wherein N is an integer greater than 1; S200, inputting the preset process parameter curves of all N processes and estimated environment parameters into a virtual production system; wherein the virtual production system is constructed in a virtual space, is used for mapping a complete physical production line and simulating a continuous production process from the first process to the last process, and integrates a defect-process correlation model established according to historical production data; S300, performing continuous production simulation from the first process to the last process in the virtual production system, and generating a virtual product quality evaluation based on the simulation result output by the last process; S400, in response to the virtual product quality evaluation result being a first result, issuing the process parameter curves of all N processes to corresponding physical wire harness production equipment; S500, in response to the virtual product quality evaluation result being a second result, adjusting one or more of the process parameter curves of all N processes, and repeatedly performing S200 to S300 until the virtual product quality evaluation result is the first result.

2. The method of claim 1, wherein, After S400, the method further comprises: real-time collection of actual process parameter curves of each process during the operation of the physical production equipment; real-time comparison of the actual process parameter curves of each process with the preset process parameter curves of the process to obtain corresponding comprehensive deviation values; when the comprehensive deviation value between the actual process parameter curve and the preset process parameter curve of any process exceeds the preset threshold value corresponding to the process, generating an equipment abnormality signal and triggering a subsequent process adjustment process.

3. The method of claim 2, wherein, After triggering the subsequent process adjustment process, further comprising: acquiring actual environment data at the triggering time and the running state of all physical production equipment; locking the process sequence number M where the deviation occurs; taking the current actual state as an initial condition, only adjusting the process parameter curves of the M+1th process to the Nth process, re-executing the simulation optimization steps of S200 to S500 to generate and issue a new set of subsequent process parameter curves.

4. The method of claim 1, wherein, The acquisition method of the estimated environment parameters comprises: collecting current environment data and historical environment data, the environment data including at least one of temperature, humidity and vibration data; inputting the current environment data and the historical environment data into a pre-trained environment derivation model to output the estimated environment parameters; wherein the environment derivation model is trained based on historical environment sample data and is used to output predicted values of environment parameters in future periods.

5. The method of claim 1, wherein, The defect-process correlation model is constructed by analyzing historical production data, and is used to establish the correlation between the process parameters, environment parameters and wire harness defect types from the first process to the last process; wherein the wire harness defect types are automatically identified and labeled from historical production data by an image recognition method or a large model analysis method.

6. The method of claim 1, wherein, The process parameter curve is used to describe the set trajectory of one or more key physical quantities changing with time in the production process; the key physical quantities include at least one of current, voltage, temperature, pressure, speed and displacement.

7. The method of claim 1, wherein, The adjusting the preset process parameter curve in S500 specifically includes: According to the wire harness defect type in the virtual product quality evaluation, one or more key processes that have the greatest impact on generating the defect type are located based on the defect-process association model; According to a preset optimization algorithm, process parameter curves of the key processes are preferentially iteratively corrected.

8. The method of claim 1, wherein, The environment derivation model is constructed by using a time series prediction algorithm, and is provided with a model updating mechanism. The model updating mechanism is used to periodically or triggeringly update parameters of the environment derivation model according to newly collected environment data.

9. An electronic device, comprising: comprise a processor and a memory; The processor is used to execute the steps of the method according to any one of claims 1 to 8 by calling programs or instructions stored in the memory.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium is used to store programs or instructions, which make the computer execute the steps of the method according to any one of claims 1 to 8.

Images