Intelligent mold scheduling method and system for automatic production line

Abstract

The invention relates to the technical field of intelligent scheduling, and discloses an intelligent mold scheduling method and system for an automatic production line, and the method comprises the steps: obtaining the real-time state of a mold, production task data and equipment positioning data; carrying out use weight distribution according to the real-time state of the mold and a pre-stored historical use record to obtain a mold use weight; performing path planning according to the mold use weight, the production task data and the equipment positioning data to obtain an initial scheduling path; according to the initial scheduling path and a historical maintenance record obtained in advance, mold switching planning is carried out, and a mold switching path and mold switching time are obtained; performing path maintenance according to the mold switching path and the initial scheduling path to obtain a final scheduling path; and scheduling the mold according to the final scheduling path, and switching the mold according to the mold switching time. The method has the following effect that the service life of the mold in the automatic production line can be prolonged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent scheduling, in particular to a mold intelligent scheduling method and system for an automated production line. BACKGROUND

[0002] At present, with the continuous progress of manufacturing technology and the improvement of automation level, automated production lines are increasingly widely used in various industries. In particular, in the fields of automobile, electronic equipment and home appliance manufacturing, efficient and precise production processes are crucial to improving enterprise competitiveness. As an indispensable part of the molding process, molds play a key role in ensuring product quality and production efficiency.

[0003] In one prior art, a mold intelligent scheduling method operates using the following steps: first, using sensors installed on production equipment to collect key parameters such as mold operating status, temperature, pressure, etc.; second, transmitting these data to a central control system, evaluating the working condition of the mold through a pre-set data analysis model, and predicting potential failure risks; third, based on the evaluation results, the system automatically generates an optimal mold scheduling plan, including mold selection, replacement timing and specific position adjustment; finally, the system also continuously monitors the use of the mold, and immediately issues an alarm and takes measures to prevent production interruption once an anomaly is detected.

[0004] However, since the system focuses mainly on maximizing short-term production efficiency, it tends to use the best-performing molds frequently, which can cause some molds to bear excessive workloads and accelerate wear and tear. In this case, molds can quickly deteriorate due to overuse, thereby shortening their service life. Further, even if the system has some failure warning mechanisms, these mechanisms are mainly designed to respond to sudden conditions rather than preventive maintenance, which also exacerbates the risk of mold wear and tear. In summary, the prior art has the problem of short mold service life. SUMMARY

[0005] The present application provides a mold intelligent scheduling method and system for an automated production line to improve the service life of molds in the automated production line.

[0006] In a first aspect, to solve the above technical problems, the present application provides a mold intelligent scheduling method for an automated production line, comprising: obtaining mold surface images, mold vibration data, production task data and equipment positioning data; extracting features from the mold surface images and the mold vibration data respectively to obtain crack growth rates and abnormal harmonic data, and inputting them into a pre-trained life prediction model to obtain mold predicted life, and then classifying life according to the mold predicted life and a pre-set life threshold to obtain a mold real-time state; determining a mold usage weight according to the mold real-time state and the pre-stored historical usage record; planning an initial scheduling path according to the mold usage weight, the production task data, and the equipment positioning data; planning mold switching according to the initial scheduling path and the pre-acquired historical maintenance record, to obtain a mold switching path and a mold switching time; maintaining the path according to the mold switching path and the initial scheduling path, to obtain a final scheduling path; scheduling the mold according to the final scheduling path, and switching the mold according to the mold switching time.

[0007] In an optional implementation, the determining of the mold usage weight according to the mold real-time state and the pre-stored historical usage record comprises: calculating a mold usage frequency according to the historical usage record; classifying the mold usage frequency according to a pre-set frequency threshold, to obtain a mold frequency label; calculating a mold usage weight according to the mold real-time state and the mold frequency label.

[0008] In an optional implementation, the planning of the initial scheduling path according to the mold usage weight, the production task data, and the equipment positioning data comprises: analyzing an urgency degree according to the production task data, to obtain a task priority queue; descendingly sorting according to the mold usage weight, to obtain a candidate mold queue; simulating a time consumption according to the equipment positioning data and a pre-stored workshop layout, to obtain a device moving time; multi-objective optimizing according to the task priority queue, the candidate mold queue, and the device moving time, to obtain the initial scheduling path.

[0009] In an optional implementation, the planning of the mold switching according to the initial scheduling path and the pre-acquired historical maintenance record comprises: evaluating a time according to the initial scheduling path, to obtain a predicted usage time; correcting the time according to the predicted usage time and the historical maintenance record, to obtain the mold switching time; ascendingly sorting according to the mold switching time, to obtain a list of molds to be switched; acquiring workshop mold inventory data; generating a mold switching path according to the mold list to be switched and the workshop mold inventory data.

[0010] In an optional implementation, the path maintenance according to the mold switching path and the initial scheduling path to obtain a final scheduling path comprises: spatial detection according to the mold switching path and the initial scheduling path to obtain a coincidence time; adding a time delay operation to the mold switching path according to the coincidence time to obtain a corrected mold switching path; time detection according to the initial scheduling path and the corrected mold switching path to obtain a coincidence space; path re-planning according to the coincidence space and the initial scheduling path to obtain a final scheduling path.

[0011] In an optional implementation, the feature extraction from the mold surface image and the mold vibration data respectively to obtain crack growth rate and abnormal harmonic data, and input into a pre-trained life prediction model to obtain a mold predicted life comprises: obtaining a mold surface image and mold vibration data; crack extraction according to the mold surface image to obtain a crack growth rate; harmonic filtering according to the mold vibration data to obtain abnormal harmonic data; inputting the crack growth rate and the abnormal harmonic data into a pre-trained life prediction model to obtain a mold predicted life; life classification according to the mold predicted life and a preset life threshold to obtain a mold real-time state; The training process of the life prediction model comprises: training the life prediction model based on historical crack features, historical vibration features and historical residual life, and obtaining a trained model when the number of training reaches a preset upper limit or the loss function of the model meets the requirements.

[0012] In an optional implementation, the weight calculation according to the mold real-time state and the mold frequency label to obtain a mold usage weight comprises: obtaining a mold environment temperature and performing temperature score calculation to obtain a temperature score; health degree score calculation according to the mold real-time state to obtain a health degree score; numerical mapping according to the mold frequency label to obtain a frequency weight; calculating a mold weight input value by the following formula:

[0013] wherein, represents a mold weight input value, represents a state coefficient, represents a frequency coefficient, represents a health score, represents a frequency weight, represents a temperature coefficient, represents a temperature score. nonlinear activation is performed according to the mold weight input value, to obtain a mold use weight.

[0014] In a second aspect, the present application provides a mold intelligent scheduling system of an automatic production line, comprising: a data acquisition module, configured to acquire mold surface images, mold vibration data, production task data and equipment positioning data; a life grading module, configured to perform feature extraction from the mold surface images and the mold vibration data respectively, to obtain crack growth rates and abnormal harmonic data, and input the crack growth rates and the abnormal harmonic data into a pre-trained life prediction model to obtain a mold predicted life, so as to perform life grading according to the mold predicted life and a preset life threshold, to obtain a mold real-time state; a weight distribution module, configured to perform use weight distribution according to the mold real-time state and pre-stored historical use records, to obtain a mold use weight; a path planning module, configured to perform path planning according to the mold use weight, the production task data and the equipment positioning data, to obtain an initial scheduling path; a mold switching module, configured to perform mold switching planning according to the initial scheduling path and pre-acquired historical maintenance records, to obtain a mold switching path and a mold switching time; a path maintenance module, configured to perform path maintenance according to the mold switching path and the initial scheduling path, to obtain a final scheduling path; a final implementation module, configured to schedule molds according to the final scheduling path, and to switch molds according to the mold switching time.

[0015] In a third aspect, the present application further provides an electronic device, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the mold intelligent scheduling method of the automatic production line according to any one of the above aspects when executing the computer program.

[0016] In a fourth aspect, the present application also provides a computer readable storage medium comprising a stored computer program, wherein the computer program, when executed, controls a device in which the computer readable storage medium is located to perform the mold intelligent scheduling method of any one of the above.

[0017] Compared with the prior art, the present application has the following beneficial effects: (1) The process of obtaining mold surface images, mold vibration data, production task data and equipment positioning data ensures comprehensive monitoring of the mold health status and operating environment. By capturing subtle changes in the mold surface with a high-resolution camera, monitoring the dynamic response of the mold during operation with a vibration sensor, and combining the production line task arrangement and equipment location information, detailed data support is provided for subsequent maintenance decisions. This step not only improves the accuracy of data collection, but also lays the foundation for preventive maintenance.

[0018] (2) Feature extraction is performed from the mold surface images and the mold vibration data to obtain crack growth rate and abnormal harmonic data, which are input into a pre-trained life prediction model to obtain mold predicted life, and life classification is performed according to the mold predicted life and a preset life threshold to obtain the mold real-time state. Advanced image processing techniques and signal analysis algorithms are used to identify the development trend of mold surface cracks and abnormal patterns in vibration, and a life prediction model is used to evaluate the remaining service life of the mold. This method can detect potential failures in advance and take timely measures to significantly extend the service life of the mold.

[0019] (3) The mold usage weight is obtained according to the mold real-time state and the pre-stored historical usage record. Based on the current state evaluation result of the mold and its historical usage record, the system assigns a usage weight, and preferentially selects molds with good health status and suitable for current production task requirements. By reasonably allocating the work load of the mold, overuse of some molds is avoided, wear is reduced, and the service life of the mold is further extended.

[0020] (4) The initial scheduling path is obtained according to the mold usage weight, the production task data and the equipment positioning data. Considering the mold usage weight, the urgency of the production task and the actual position of the equipment, an intelligent algorithm is used to generate the optimal initial scheduling path. This process not only optimizes the movement route of the mold on the production line, but also reduces the additional wear caused by unnecessary handling, indirectly extending the service life of the mold.

[0021] (5) According to the initial scheduling path and the pre-acquired historical maintenance record, the mold switching planning is carried out, and the mold switching path and the mold switching time are obtained. Based on the initial scheduling path and referring to the historical maintenance record, the system can plan the time and path of mold switching in advance, avoiding unnecessary downtime and delay. Reasonable switching plan can reduce the exposure of the mold to the harsh environment in unnecessary cases, reduce the damage risk, and help maintain the good state of the mold.

[0022] (6) According to the mold switching path and the initial scheduling path, path maintenance is carried out, and the final scheduling path is obtained; the mold is scheduled according to the final scheduling path, and the mold is switched according to the mold switching time. The mold switching path is integrated into the initial scheduling path for comprehensive optimization to form an efficient and reasonable final scheduling path. At the same time, strict implementation of the predetermined mold switching time and path in specific operations maximizes downtime and damage risk during the switching process. These measures work together to ensure that the mold can run in the best state, effectively delay the aging process, and thus achieve the goal of extending the service life of the mold. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is a mold intelligent scheduling method flow diagram of an automatic production line provided by the first embodiment of the present application; Figure 2 is a structure diagram of a mold intelligent scheduling system of an automatic production line provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] Referring to Figure 1 , the first embodiment of the present application provides a mold intelligent scheduling method of an automatic production line, including the following steps: S11, obtaining mold surface image, mold vibration data, production task data and equipment positioning data; S12, extracting features from the mold surface image and the mold vibration data respectively to obtain crack growth rate and abnormal harmonic data, and inputting the data into a pre-trained life prediction model to obtain mold predicted life, so as to perform life classification according to the mold predicted life and a preset life threshold to obtain mold real-time state; S13, performing usage weight distribution according to the mold real-time state and the pre-stored historical usage record, to obtain a mold usage weight; S14, performing path planning according to the mold usage weight, the production task data and the equipment positioning data, to obtain an initial scheduling path; S15, performing mold switching planning according to the initial scheduling path and the pre-acquired historical maintenance record, to obtain a mold switching path and a mold switching time; S16, performing path maintenance according to the mold switching path and the initial scheduling path, to obtain a final scheduling path; S17, scheduling the mold according to the final scheduling path, and switching the mold according to the mold switching time.

[0026] In step S11, mold surface images, mold vibration data, production task data and equipment positioning data are acquired. In an embodiment, the production task data is acquired in real time relying on an ERP system interface to acquire order parameters (such as 3000 automobile B-pillar order quantities per day), combined with MES to analyze XML process files to extract key indicators (such as 220±5℃ constant temperature standards required by die casting process), and an EPC Gen2 RFID reader deployed in the production line to capture process flow state at a frequency of 20 times per second, and a mobile terminal to support code scanning and emergency order entry, these data are managed by an InfluxDB time series database with millisecond-level time stamp dynamic queue, an Oracle relational database stores process parameter mapping table, and an HBase archives historical task data within five years, to provide a calculation basis for intelligent scheduling system to match mold performance and delivery period reverse arrangement; the equipment positioning data is acquired by a Decawave UWB base station deployed on the workshop ceiling to realize three-dimensional space positioning (positioning accuracy ±10cm), an Impinj RFID reader buried in the key passage to divide regional coordinates, an AGV-mounted SICK NAV350 laser navigation system to construct a dynamic path with an angle resolution of 0.1°, a Heidenhain absolute value encoder equipped in the crane motion mechanism to transmit ±1mm positioning data through an SSI interface, real-time coordinates are written into a Redis cluster at a frequency of 500ms, mobile trajectories are stored in a MongoDB sharded cluster, and time series positioning data are stored by TDengine, to support a crane three-dimensional obstacle avoidance algorithm to calculate collision probability (such as dynamically avoiding a 200T hydraulic machine in the D5 area maintenance), and to provide an optimal path planning for an AGV cluster scheduling system (straight-line distance calculation error <0.5m).

[0027] In a specific embodiment, a new energy vehicle battery box production line integrates ERP orders and AGV laser positioning data through OPC UA protocol, realizes millisecond-level collaboration of mold transportation and press preparation process, and shortens the mold changing time from 45 minutes to 22 minutes; an aerospace workshop adopts UWB+KVH fiber gyroscope fusion positioning technology, realizes three-dimensional positioning error <5 cm of the crown block in a 200m×80m operation area, predicts path conflicts 36 hours in advance combined with the digital twin system, and reduces the annual equipment downtime loss by 9.2 million yuan; a 3C electronic factory realizes mold transfer positioning accuracy ±3 cm in the SMT workshop through the fusion of RFID area positioning and AGV laser navigation data, and the comprehensive utilization rate of equipment is improved to 89.7%.

[0028] In step S12, feature extraction is performed from the mold surface image and the mold vibration data respectively to obtain crack growth rate and abnormal harmonic data, and the mold prediction life is obtained by inputting the crack growth rate and the abnormal harmonic data into the pre-trained life prediction model, so that life classification is performed according to the mold prediction life and a preset life threshold to obtain the real-time state of the mold.

[0029] In an embodiment, the mold surface image and the mold vibration data are obtained; crack extraction is performed according to the mold surface image to obtain the crack growth rate; harmonic filtering is performed according to the mold vibration data to obtain abnormal harmonic data; the crack growth rate and the abnormal harmonic data are input into a pre-trained life prediction model to obtain a mold prediction life; and life classification is performed according to the mold prediction life and a preset life threshold to obtain the real-time state of the mold; wherein the training process of the life prediction model comprises: training the life prediction model based on historical crack features, historical vibration features and historical residual life, and obtaining the trained model when the training times reach a preset upper limit or the loss function of the model meets the requirements.

[0030] It is worth noting that first, the mold surface image is obtained in real time by an industrial camera or an image acquisition device, and at the same time, the vibration data of the mold during operation is collected by a vibration sensor. For the mold surface image, image processing techniques such as edge detection, threshold segmentation or deep learning models are used to identify and track the crack morphology. By comparing image data at different times, the rate of crack propagation (such as monthly growth rate of crack length or area) is calculated, and thus the crack growth rate is obtained. For vibration data, signal processing techniques such as Fourier transform are used to decompose the harmonic components in the vibration signal, and filtering algorithms such as band-pass filtering or wavelet denoising are used to remove normal harmonics and retain abnormal harmonics (such as vibration components with excessively high amplitude or abnormal frequency), and thus the abnormal harmonic data is obtained.

[0031] In one specific embodiment, the training process of the life prediction model based on the random forest model includes: first, collecting historical mold data as a training set, including crack features (such as crack length, propagation rate, area ratio), vibration features (abnormal harmonic amplitude, frequency, energy ratio), and corresponding actual remaining life (such as remaining days or number of uses). After standardization (such as Z-Score) and cleaning (removing outliers, filling missing values), the input feature vector is constructed, for example, the crack growth rate is combined with the abnormal harmonic amplitude, and the derived features (such as the product of the two) are added. Subsequently, the random forest regression model is initialized, the parameters such as the number of trees, the maximum depth, etc. are set, the hyperparameters are adjusted through cross-validation (such as 5-fold) to optimize the prediction error (such as mean square error MSE). During training, when the MSE of the model on the validation set is lower than the threshold (such as 0.05) or reaches the upper limit of the preset iteration number (such as 200 times), the training is stopped, and the model performance is evaluated through the test set (such as RMSE<10 days). Finally, after the model is deployed, real-time crack growth rate and abnormal harmonic data can be input to predict the remaining life of the mold, and the mold real-time state is divided according to the preset threshold. The mold real-time state includes normal, loss, and imminent scrap.

[0032] In one embodiment, the model outputs the predicted value of the remaining life of the mold as 800 hours. Then, according to the preset life classification standard (such as normal: >1000 hours; loss: 500-1000 hours; imminent scrap: <500 hours), the predicted life of 800 hours is classified as loss, and the final determination of the current real-time state of the mold is "loss state, need to strengthen monitoring".

[0033] It is worth noting that the crack growth rate and abnormal harmonic data are input into the pre-trained life prediction model (such as a random forest or neural network model), which has learned the relationship between crack propagation, vibration anomaly and mold life based on historical data, thereby predicting the remaining service life of the current mold (such as remaining days or number of uses). According to the preset life threshold, the mold state is classified: if the predicted remaining life exceeds the threshold (such as more than 180 days), it is determined to be in the "normal" state; if the remaining life is in the middle interval (such as 60 to 180 days), it is determined to be in the "loss" state; if the remaining life is below the minimum threshold (such as less than 60 days), it is determined to be in the "imminent scrap" state. This process updates the real-time state of the mold dynamically through real-time data input and model calculation, providing a basis for subsequent scheduling.

[0034] In a specific embodiment, the training data are as follows: sample 1: crack growth rate 0.15 mm / month, abnormal harmonic amplitude 9.2 dB, remaining life 180 days; sample 2: crack growth rate 0.32 mm / month, abnormal harmonic amplitude 14.5 dB, remaining life 85 days; sample 3: crack growth rate 0.08 mm / month, abnormal harmonic amplitude 6.7 dB, remaining life 250 days; sample 4: crack growth rate 0.45 mm / month, abnormal harmonic amplitude 18.3 dB, remaining life 40 days; sample 5: crack growth rate 0.21 mm / month, abnormal harmonic amplitude 11.0 dB, remaining life 120 days.

[0035] In step S13, the mold usage weight is obtained according to the mold real-time state and the pre-stored historical usage record.

[0036] In an implementation, the mold usage frequency is obtained according to the historical usage record; the mold frequency label is obtained according to the mold usage frequency and the preset frequency threshold; and the mold usage weight is obtained according to the mold real-time state and the mold frequency label.

[0037] It is worth noting that the weight allocation mechanism in production task scheduling realizes decision optimization through multi-stage data processing. The core process first calculates the average usage frequency per unit time based on the historical work order records stored in the device management system, forms a quantitative frequency benchmark value, and then compares the calculated frequency value with the pre-set classification threshold interval, for example, divides the monthly average usage of more than 50 times into a high-frequency level, 20-50 times into a medium-frequency level, and less than 20 times into a low-frequency level, generates a corresponding frequency label through a three-level classification system, and finally determines the specific weight value through a dynamic weighting algorithm by comprehensively considering the real-time running state parameters of the mold (including the current working temperature monitored by the temperature sensor, the bearing wear index collected by the vibration sensor, the surface scratch level identified by the visual system, etc.) and the established frequency label, for example, the high-frequency level mold can obtain a basic weight coefficient of 0.7 under normal working conditions, but if the temperature is detected to be out of limit, a weight reduction factor of 0.3 is triggered, forming a final usage priority of 0.21.

[0038] It is worth mentioning that the original data of mold frequency comes from the equipment start-stop log recorded by the manufacturing execution system, the mold retrieval events captured by the radio frequency identification reader, and the process parameters associated with the production work order. The generation of the frequency label depends on the grading standard system set by the enterprise according to the equipment maintenance cycle and production load characteristics, and the main goal is to provide the basis for the equipment scheduling priority decision of the intelligent production scheduling system. Taking a certain automobile covering stamping workshop as an example, it marks the aluminum alloy mold with a monthly average use of more than 80 times as a red high-frequency label by analyzing the production data of the past 12 months, and prioritizes matching the Tesla Model Y roof order when scheduling; The galvanized steel plate mold with a monthly average use of 30-80 times is marked as a yellow medium-frequency label, which is used for regular order production; And the carbon fiber mold with a monthly average use of less than 30 times is marked as a blue low-frequency label, which triggers equipment maintenance reminders and restricts its participation in emergency order scheduling. Through this mechanism, the failure rate of high-value molds is reduced by 37%, and the maintenance timeliness rate of special material molds is guaranteed to reach 100%.

[0039] In an embodiment, the mold environment temperature is obtained, and a temperature score calculation is performed to obtain a temperature score; According to the real-time state of the mold, a health score calculation is performed to obtain a health score; According to the mold frequency label, a numerical mapping is performed to obtain a frequency weight; The mold weight input value is calculated by the following formula:

[0040] wherein, represents the mold weight input value, represents the state coefficient, represents the frequency coefficient, represents the health score, represents the frequency weight, represents the temperature coefficient, represents the temperature score; According to the mold weight input value, a nonlinear activation is performed to obtain a mold use weight.

[0041] It is worth noting that, according to the preset lifespan grading standards, the mold health score is calculated through the following steps: First, the predicted lifespan of the mold is compared with the lifespan threshold. If the remaining lifespan exceeds 1000 hours (e.g., 1200 hours), the health score is a full score of 1.0; if the remaining lifespan is between 500 and 1000 hours (e.g., 800 hours), the health score is the remaining lifespan divided by 1000, resulting in 0.8; if the remaining lifespan is less than 500 hours (e.g., 300 hours), the health score is the remaining lifespan divided by 1000, resulting in 0.3. For example, if a mold has a predicted lifespan of 600 hours, belonging to the "wear and tear" level, its health score is 600 / 1000 = 0.6; if the predicted lifespan is 400 hours, belonging to the "about to be scrapped" level, the health score is 0.4. This method ensures that the score reflects both the absolute value of the remaining lifespan and conforms to the grading standards by linearly mapping the lifespan threshold to the 0-1 range, intuitively quantifying the health status of the mold, and the calculation process is clear and executable.

[0042] It is worth noting that when mapping values ​​based on the mold frequency labels, the average monthly usage frequency is first divided into three intervals: high frequency (≥80 times per month), medium frequency (30-80 times per month), and low frequency (<30 times per month). Then, a standardized mapping is used to convert these intervals into weighted values ​​of 0-1. For example, a mold used an average of 85 times per month falls under the high-frequency label and is mapped to a weight of 0.8; a mold used an average of 50 times per month falls under the medium-frequency label and is mapped to a weight of 0.5; while a low-frequency label of 20 times per month is mapped to a weight of 0.2. This mapping rule assigns higher weights (0.8) to high-frequency, followed by medium (0.5), and lowest weights (0.2) to low-frequency. This reflects the difference in the impact of usage frequency on mold wear and tear, and eliminates the dimensional interference of the original usage frequency through 0-1 standardization, ensuring compatibility with the weighted calculation of health scores.

[0043] It is worth noting that the temperature score is calculated using the following formula:

[0044] in, This represents a temperature score (dimensionless). Value Used for dimensional normalization This indicates the ambient temperature of the mold. This indicates the maximum value of the preset temperature range (value is 323K). This indicates the minimum value of the preset temperature range (value is 353K).

[0045] It is worth noting that the nonlinear activation function uses the sigmoid function. The mold weight input values ​​are input into the nonlinear activation function, and the output value is used as the mold usage weight. The range of both the mold weight input value and the mold usage weight is (0, 1).

[0046] In step S14, path planning is performed according to the mold usage weight, the production task data, and the equipment positioning data to obtain an initial scheduling path.

[0047] In an embodiment, urgency analysis is performed according to the production task data to obtain a task priority queue; descending order sorting is performed according to the mold usage weight to obtain a candidate mold queue; time consumption simulation is performed according to the equipment positioning data and a pre-stored workshop layout map to obtain equipment moving time; and multi-objective optimization is performed according to the task priority queue, the candidate mold queue, and the equipment moving time to obtain an initial scheduling path.

[0048] It is worth noting that the system extracts the original data of the task to be processed, including the deadline, the customer's urgency requirement level (such as "high", "medium", and "low"), the order value, the product type (such as high-value products or regular products), and the current production stage, and then converts the non-numerical parameters into score items through preset rules - for example, the remaining deadline is scored according to the inverse rule (10 points for 1 day, 7 points for 3 days, 5 points for 5 days, and 1 point for the lowest), the urgency requirement level is mapped to a fixed score (8 points for "high", 5 points for "medium", and 2 points for "low"), the order value is divided into grades according to the amount (6 points for ≥1 million, 4 points for 50-100 million, and 2 points for ≤50 million), and 3 points are added for the key production stage. All parameters are weighted and summed according to the weight proportion (such as 40% for the deadline, 30% for the urgency requirement, 20% for the value, and 10% for the stage), for example, a high-value order with 2 days remaining and an urgency requirement of "high" has a total score of (8 points x 40%) + (8 points x 30%) + (6 points x 20%) + 3 points x 10% = 7.1 points. Finally, the system sorts the tasks according to the urgency score from high to low to form a dynamically updated "task priority queue": tasks with an urgency score of ≥8 points trigger priority scheduling, tasks with a score of 7-8 points enter a secondary priority pool, and tasks with a score of <7 points are scheduled normally, ensuring accurate resource allocation and zero delay for critical tasks.

[0049] It is worth noting that the system sorts all available molds in descending order according to the "usage weight" of the mold (which combines the current state of the mold, the historical usage frequency, and the health level, and the higher the weight value, the more the mold should be prioritized for use) to generate a "candidate mold queue". For example, molds in good condition and used less frequently will be placed in the front row, while molds in poor condition or used frequently will be placed at the back, ensuring that high-priority tasks use high-quality molds and avoiding excessive consumption of high-load molds.

[0050] It is worth mentioning that the system combines the current position data of the equipment (such as the real-time coordinates of the AGV trolley or the mechanical arm) with the pre-stored workshop layout diagram (including the path between equipment, obstacle information, etc.), simulates the time required for different equipment to move from the current position to the target mold or task station through path planning algorithm (such as calculating the shortest path or obstacle avoidance path between two points), and generates a "equipment moving schedule". For example, if a certain equipment is far away from the target mold and the path is congested, the system will estimate a longer moving time, and vice versa.

[0051] It is worth mentioning that the system first calculates the shortest path of the equipment from the current position to the target mold based on the "equipment moving schedule" through Dijkstra algorithm, and generates multiple candidate paths and calculates the moving time consumption of each path in combination with the time window constraints such as the end time of the current task of the equipment (such as the equipment is expected to be idle at 14:00); then enters the multi-objective balance decision stage, and scores the candidate paths by weight - for high-priority tasks with an urgency of ≥8 points, the optimal path that meets the deadline is forced to be selected; for non-urgent tasks, the path with the shortest moving time is preferred (such as path A takes 15 minutes, better than path B which takes 20 minutes); at the same time, through the mold usage balancing mechanism, low-frequency molds (such as molds with a frequency weight ≤0.3 and a daily usage frequency ≤20 times) are preferentially scheduled, and high-frequency molds (such as molds that have been used 75 times a day and have a frequency weight ≥0.8) are automatically avoided; finally, the task execution order, mold allocation details, equipment moving route and time period are integrated into the "initial scheduling path".

[0052] In step S15, mold switching planning is performed according to the initial scheduling path and the pre-acquired historical maintenance record to obtain a mold switching path and a mold switching time.

[0053] In an embodiment, the initial scheduling path is calculated to perform time evaluation to obtain a predicted usage time; the predicted usage time and the historical maintenance record are used to perform time correction to obtain a mold switching time; the mold switching time is sorted in ascending order to obtain a list of molds to be switched; workshop mold inventory data is acquired; the list of molds to be switched and the workshop mold inventory data are used to generate a path to obtain a mold switching path.

[0054] In one embodiment, the implementation process of the mold switching plan is first based on the initial scheduling scheme generated by the intelligent scheduling system, through analyzing the predicted operation time of each station mold (including production batch, equipment rated speed, process standard time consumption and other factors) to simulate the whole process time, to form a prediction model containing the start and end time of each process node; Then call the historical maintenance records in the equipment management database (such as the average failure interval time of the same type of mold in the past half year, the cumulative use time after the last preventive maintenance, and the repair time statistics of typical fault types), use the time correction algorithm based on Bayesian inference, Add maintenance risk buffer value in the predicted use time, for example, add 15% of the standby time for old molds with more than 500,000 cumulative stamping times; After completing the time correction, sort the mold switching requirements according to the emergency degree, and process the devices with remaining life below the safety threshold first, to generate a list of molds to be switched including mold code, switching period and priority code; The system synchronously accesses the real-time inventory data of the warehouse management system (covering the storage location, available quantity and supporting fixture status of each type of mold in the three-dimensional warehouse), combined with the real-time position information of AGV and the congestion index of workshop logistics channel, use ant colony optimization algorithm to plan the material handling path considering time efficiency and resource utilization, for example, allocate exclusive transportation channel for high-priority molds in workshop C area and dispatch nearby AGV to execute pick and delivery work, to ensure that the mold switching operation and production rhythm are accurately matched.

[0055] It is worth noting that the following is the complete steps of mold switching path planning by ant colony optimization algorithm: first, the system constructs a digital map containing all nodes and path segments according to the warehouse management system data and real-time logistics status, and labels the basic attributes of each path segment. Then initialize the algorithm parameters, including the number of ants (such as 100), the pheromone evaporation coefficient (such as 0.1), the maximum number of iterations (such as 50), and set the initial pheromone value for each path segment (such as 1), and set the heuristic factor weight according to the task urgency (such as high priority path focuses on the shortest time, weight 0.7; low priority path focuses on resource utilization, weight 0.5). After the algorithm starts, each ant starts from the current position of AGV, and selects the path segment by probability: the transition probability is calculated by the weighted formula of pheromone concentration (α=1.5) and real-time loss (the higher the congestion index, the higher the loss, β=1), and the path with high pheromone and low real-time loss is preferred (such as low congestion path in workshop C area). After the path search is completed, the system evaluates the path quality according to the total loss (transportation time + congestion penalty value, such as congestion increases by 5% time per level), and updates the pheromone in reverse - high-quality path (low loss) increases Δτ=1 / (1+loss), and low-quality path decreases Δτ=evaporation coefficient×current pheromone, while synchronously adjusting the heuristic factor with real-time data (such as AGV new position, channel status). Iterate to the set number of times or meet the convergence condition (such as no change in the optimal path for 5 consecutive times), select the global optimal path (such as the shortest time and the least conflict), generate execution instructions (such as "AGV-03 starts from C area storage C3-12, along path C→D→E, estimated time 12 minutes, avoid B area channel congestion"), and synchronize to AGV control system for execution. For example, when a high-priority mold needs to be taken out from the C area storage, the algorithm determines the optimal path as C area→main channel→workshop east door→workstation 5 by simulating the path selection of 100 ants, and the path is selected because of its low congestion index (0 level) and high historical pheromone concentration. The system triggers AGV-07 to execute the task and reserves a dedicated channel, ensuring that the switching operation is accurately matched with the production rhythm. This process takes into account path efficiency and dynamic environmental adaptability through real-time data fusion, dynamic pheromone update, and multiple iterations.

[0056] In step S16, path maintenance is performed according to the mold switching path and the initial scheduling path to obtain a final scheduling path.

[0057] In an embodiment, spatial detection is performed according to the mold switching path and the initial scheduling path to obtain a coincidence time; a time delay operation is added to the mold switching path according to the coincidence time to obtain a corrected mold switching path; time detection is performed according to the initial scheduling path and the corrected mold switching path to obtain a coincidence space; and path re-planning is performed according to the coincidence space and the initial scheduling path to obtain a final scheduling path.

[0058] In one specific embodiment, a three-dimensional dynamic model of the workshop is first constructed based on digital twinning technology, and the mold switching path and the initial scheduling path are imported into the simulation system for space-time dual conflict detection: by scanning the space-time coordinates of the equipment movement track (such as AGV needs to pass through A3 channel at 10:15, while the hoisting equipment is operating in the same period in the area), the path overlap period is identified and automatically marked as a red conflict area; then a dynamic priority insertion mechanism is used to retain the time window of the conflict section involved in high emergency tasks (such as the mold transportation of special urgent order X needs to use A3 channel at 10:20-10:25) in priority, while applying a delay strategy to the path of regular task Y (such as adjusting the start time of its mold switching path from 10:18 to 10:26), generating the corrected mold switching path; then an improved ant colony algorithm is used for global re-planning, by merging transportation batches to optimize equipment load (for example, two sets of molds originally planned to be transported to D area separately are integrated into a multi-target transportation, making the AGV utilization rate increase by 35%), while implanting elastic buffer time (such as reserving 5 minutes of maneuvering period at 11:00-11:15) to deal with unexpected tasks; finally, the feasibility of the new path is verified through Monte Carlo simulation, when it is detected that the B area punching machine receives two sets of molds at 11:30, the system automatically triggers the hierarchical unloading protocol, and the molds with precision level ±0.02mm are temporarily stored in the transition buffer rack in priority, and a path execution list with three-dimensional coordinate time sequence (such as AGV-07 needs to move at a speed of 2m / s along the grid coordinates X35-Y78 to X42-Y83 at 10:22:30) is generated, and the equipment load balancing index is updated to the central scheduling board.

[0059] In step S17, the molds are scheduled according to the final scheduling path, and the mold switching is performed according to the mold switching time.

[0060] In a specific embodiment, first, based on the final scheduling path, an AGV control instruction set containing a sequence of three-dimensional coordinates and timestamps is generated (for example, AGV-12 needs to move at a speed of 1.5 m / s along the grid coordinates X20-Y45 to X28-Y52 at 10:15:30), and at the same time, through the digital twin platform, a mold preheating instruction is synchronously issued to the target station. For example, when it is detected that the No. 5 punch press in area D needs to be switched to the M-207 mold at 11:00, the constant temperature heating program of the mold is triggered 30 minutes in advance to make it stable at 85℃±3℃; during the transportation stage, the system verifies the binding relationship between the AGV carrying the mold through the laser positioning device and the RFID scanner (for example, whether the mold code on the AGV-09 rack matches the M-109 mold of task sheet T-335), and if the offset error exceeds 5 cm or the ID does not match, the sound and light alarm is triggered immediately and the transportation chain is paused; when the AGV arrives at the target device, the automatic hoisting system performs a hierarchical unloading operation according to the preset priority, for example, the mold of special urgent order X adopts a direct installation mode, and the mold of regular task Y is temporarily stored in the device side buffer zone to shorten the switching interval; during the physical switching process of the mold, the system synchronously collects pressure sensor data (for example, whether the clamping force of the installed M-207 mold reaches the standard value of 1200N±50N), and if an abnormality is detected, the standby mold rapid replacement mechanism is started, for example, the preheated M-207B mold is called and reinstalled within 3 minutes; at the same time, the dynamic load balancing module updates the mold usage frequency map in real time, and when the daily usage time of a certain mold accumulates more than 8 hours, it is automatically removed from the candidate queue and a maintenance work order is triggered to push to the patrol terminal, ensuring the health of the device and the continuity of production.

[0061] In summary, the present application discloses a mold intelligent scheduling method of an automatic production line, which aims to predict the remaining life of the mold and balance the planning based on the mold usage frequency, so as to improve the life of the mold in the intelligent scheduling process of the automatic production line.

[0062] With reference to Figure 2 , the second embodiment of the present application provides a mold intelligent scheduling system of an automatic production line, comprising: a data acquisition module for acquiring mold surface images, mold vibration data, production task data and device positioning data; a life grading module for extracting features from the mold surface images and the mold vibration data respectively to obtain crack growth rate and abnormal harmonic data, and inputting the data into a pre-trained life prediction model to obtain mold predicted life, so as to grade the life according to the mold predicted life and a preset life threshold to obtain a mold real-time state; a weight allocation module for allocating usage weights according to the mold real-time state and pre-stored historical usage records to obtain mold usage weights; a path planning module, configured to perform path planning according to the mold usage weight, the production task data and the equipment positioning data to obtain an initial scheduling path; a mold switching module, configured to perform mold switching planning according to the initial scheduling path and a pre-acquired historical maintenance record to obtain a mold switching path and a mold switching time; a path maintenance module, configured to perform path maintenance according to the mold switching path and the initial scheduling path to obtain a final scheduling path; a final implementation module, configured to schedule molds according to the final scheduling path and switch molds according to the mold switching time.

[0063] It should be noted that the mold intelligent scheduling system of the automatic production line provided by the embodiments of the present application is used to perform all process steps of the mold intelligent scheduling method of the automatic production line provided by the embodiments, and the working principles and beneficial effects of the two are one-to-one correspondence, thus not being described in detail.

[0064] The embodiments of the present application further provide an electronic device. The electronic device comprises a processor, a memory and a computer program, such as a data acquisition program, stored in the memory and executable on the processor. The processor implements the steps in the above various mold intelligent scheduling method embodiments of the automatic production line when executing the computer program, such as the step S11 shown. Figure 1 Alternatively, the processor implements the functions of each module / unit in the above various device embodiments when executing the computer program, such as a data acquisition module.

[0065] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0066] The electronic device can be a desktop computer, a notebook, a palm computer and a smart tablet, etc. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device, and can include more or fewer components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0067] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the electronic device, and connects various parts of the electronic device through various interfaces and lines.

[0068] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), etc.; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0069] The modules / units integrated in the electronic device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. that can carry the computer program code. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0070] It should be noted that the above-described device embodiments are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. In addition, the connection relationship between the modules in the device embodiment provided by the present application indicates that there is a communication connection between them, which can be realized as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0071] The above-described specific embodiments further illustrate the purpose, technical solutions, and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only examples of the present application and are not intended to limit the protection scope of the present application. In particular, any modifications, equivalent replacements, improvements, etc. made by those skilled in the art within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligent scheduling of molds in an automated production line, characterized in that, include: Acquire mold surface images, mold vibration data, production task data, and equipment positioning data; Feature extraction is performed on the mold surface image and the mold vibration data to obtain crack growth rate and abnormal harmonic data, which are then input into a pre-trained life prediction model to obtain the mold prediction life. Based on the mold prediction life and the preset life threshold, life classification is performed to obtain the real-time state of the mold. The usage weight is obtained by allocating usage weights based on the real-time status of the mold and the pre-stored historical usage records; Based on the mold usage weight, the production task data, and the equipment positioning data, path planning is performed to obtain the initial scheduling path; Based on the initial scheduling path and the pre-acquired historical maintenance records, mold switching planning is performed to obtain the mold switching path and mold switching time. The final scheduling path is obtained by performing path maintenance based on the mold switching path and the initial scheduling path. The molds are scheduled according to the final scheduling path, and the molds are switched according to the mold switching time.

2. The intelligent mold scheduling method for automated production lines according to claim 1, characterized in that, The step of allocating usage weights based on the real-time status of the mold and pre-stored historical usage records to obtain mold usage weights includes: The frequency of mold usage is calculated based on the historical usage records. Based on the frequency of mold use and a preset frequency threshold, frequency classification is performed to obtain mold frequency labels; The mold usage weight is obtained by calculating the weight based on the real-time status of the mold and the mold frequency label.

3. The intelligent mold scheduling method for automated production lines according to claim 1, characterized in that, The step of performing path planning based on the mold usage weight, the production task data, and the equipment positioning data to obtain the initial scheduling path includes: Based on the production task data, an urgency analysis is performed to obtain a task priority queue; The candidate molds are sorted in descending order according to their weights to obtain a queue of candidate molds. The equipment movement time is obtained by simulating the time consumption based on the equipment positioning data and the pre-stored workshop layout diagram; The initial scheduling path is obtained by performing multi-objective optimization based on the task priority queue, the candidate mold queue, and the device movement time.

4. The intelligent mold scheduling method for automated production lines according to claim 1, characterized in that, The step of planning the mold switching based on the initial scheduling path and pre-acquired historical maintenance records to obtain the mold switching path and mold switching time includes: The initial scheduling path is calculated and time-evaluated to obtain the predicted usage time; The mold switching time is obtained by adjusting the time based on the predicted usage time and the historical maintenance records. Sort the molds in ascending order according to their switching times to obtain a list of molds to be switched. Obtain workshop mold inventory data; The mold switching path is generated based on the list of molds to be switched and the workshop mold inventory data.

5. The intelligent mold scheduling method for an automated production line according to claim 1, characterized in that, The step of maintaining the path based on the mold switching path and the initial scheduling path to obtain the final scheduling path includes: Spatial detection is performed based on the mold switching path and the initial scheduling path to obtain the overlap time; Based on the overlap time, a delay operation is added to the mold switching path to obtain a corrected mold switching path; Time detection is performed based on the initial scheduling path and the corrected mold switching path to obtain the overlapping space; Based on the overlapping space and the initial scheduling path, path replanning is performed to obtain the final scheduling path.

6. The intelligent mold scheduling method for an automated production line according to claim 1, characterized in that, The step of extracting features from the mold surface image and the mold vibration data to obtain crack growth rate and abnormal harmonic data, and then inputting them into a pre-trained life prediction model to obtain the mold's predicted life, includes: Cracks are extracted from the mold surface image to obtain the crack growth rate. Harmonic filtering is performed on the mold vibration data to obtain abnormal harmonic data; The crack growth rate and the abnormal harmonic data are input into a pre-trained life prediction model to obtain the mold's predicted life. The training process of the lifespan prediction model includes: The lifetime prediction model is trained based on historical crack characteristics, historical vibration characteristics, and historical remaining lifetime. When the number of training iterations reaches a preset upper limit or the loss function of the model is detected to meet the requirements, the trained model is obtained.

7. The intelligent mold scheduling method for an automated production line according to claim 2, characterized in that, The step of calculating the mold usage weight based on the real-time status of the mold and the mold frequency tag includes: The ambient temperature of the mold is obtained, and a temperature score is calculated to obtain the temperature score. A health score is calculated based on the real-time status of the mold. The frequency weights are obtained by numerical mapping based on the mold frequency labels; The mold weight input value is calculated using the following formula: ； in, This represents the mold weight input value. Represents the state coefficient. Represents the frequency coefficient. Indicates health score, Represents frequency weights. Indicates the temperature coefficient. Indicates temperature rating; The mold usage weight is obtained by performing nonlinear activation based on the mold weight input value.

8. A mold intelligent scheduling system for an automated production line, characterized in that, include: The data acquisition module is used to acquire mold surface images, mold vibration data, production task data, and equipment positioning data; The life grading module is used to extract features from the mold surface image and the mold vibration data to obtain crack growth rate and abnormal harmonic data, and input them into a pre-trained life prediction model to obtain the mold prediction life. Then, the life is graded according to the mold prediction life and the preset life threshold to obtain the real-time status of the mold. The weight allocation module is used to allocate usage weights based on the real-time status of the mold and the pre-stored historical usage records to obtain the mold usage weights. The path planning module is used to perform path planning based on the mold usage weight, the production task data, and the equipment positioning data to obtain an initial scheduling path; The mold switching module is used to plan the mold switching based on the initial scheduling path and the pre-acquired historical maintenance records, and to obtain the mold switching path and mold switching time. The path maintenance module is used to maintain the path based on the mold switching path and the initial scheduling path to obtain the final scheduling path. The final implementation module is used to schedule the molds according to the final scheduling path and to switch the molds according to the mold switching time.

9. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the intelligent mold scheduling method for an automated production line as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the intelligent mold scheduling method for an automated production line as described in any one of claims 1 to 7.

Images