A metal sheet production scheduling method and system based on artificial intelligence

Abstract

The invention discloses a metal sheet production scheduling method and system based on artificial intelligence, which belongs to the field of intelligent production scheduling technology. The invention updates the real-time production order information and real-time production line information by analyzing the number of scrapped metal sheets and the production lines where mechanical failures occur, and predicts the delivery status based on the updated information analysis, and sends a regeneration instruction to the plan generation module based on the predicted delivery status. After receiving the regeneration instruction, the plan generation module generates multiple production scheduling plans that can ensure that all orders are delivered on time. Finally, the plan selection module calculates the priority coefficient of each production scheduling plan based on the estimated production time and the estimated total energy consumption of the number of production lines to select the optimal production scheduling plan, which can improve the overall production efficiency and reduce production energy consumption on the basis of completing the delivery of each order within the delivery period of each order.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent production scheduling, and in particular to a metal plate production scheduling method and system based on artificial intelligence. Background Art

[0002] Sheet metal refers to sheet materials made from various metal materials. These materials are widely used in a variety of industries, including construction, manufacturing, aerospace, and the automotive industry. Consequently, sheet metal production is in high demand, making it crucial for production shops to plan their production schedules and meet production orders on time.

[0003] However, during the production process, some of the produced boards may not meet the quality standards for delivery and need to be re-produced. There may also be problems with a device in a production line, causing the production line to stop operating. These situations greatly affect production efficiency, and production may not be completed on time according to the original production plan. Therefore, scheduling is required according to the situation. Therefore, how to quickly and intelligently generate a production scheduling plan based on the situation has become an urgent problem that needs to be solved. Summary of the Invention

[0004] In order to solve the technical problems existing in the prior art, the present invention provides a metal sheet production scheduling system based on artificial intelligence, comprising:

[0005] A data acquisition module is used to obtain real-time production order information of metal sheets and real-time production line information of the production workshop. The real-time production order information includes the real-time number of sheets to be produced and the delivery deadline of each order. The real-time production line information includes the real-time production line number, production efficiency, energy consumption information and real-time status of each production line, and the real-time status includes normal status and maintenance status.

[0006] A sheet quality assessment module, configured to identify defect information of produced metal sheets and assess the quality type of the corresponding metal sheets based on the defect information, wherein the quality types include acceptable, repairable, and scrapped;

[0007] The fault analysis module is used to locate the faulty equipment in the corresponding production line based on the defect information of the metal sheet and analyze the fault type of the faulty equipment, including parameter fault and mechanical fault;

[0008] An information updating module, configured to update the real-time production order information and the real-time production line information;

[0009] The delivery analysis module is used to analyze and predict the delivery status based on real-time production order information and real-time production line information, and send regeneration instructions to the plan generation module based on the predicted delivery status;

[0010] A plan generation module is configured to, upon receiving a regeneration instruction, generate multiple production scheduling plans based on real-time production order information and real-time production line information using a trained scheduling algorithm model to ensure that all orders are delivered on time. The production scheduling plans record corresponding scheduling information, including the number of production lines activated S, the estimated production time T, and the estimated total energy consumption V. The production lines activated in the production scheduling plans are all in normal status;

[0011] The plan selection module is used to calculate the priority coefficient P of each production scheduling plan based on the scheduling information, and finally select the production scheduling plan with the largest priority coefficient as the new production scheduling plan.

[0012] Furthermore, the plate quality assessment module includes:

[0013] An image acquisition module, used for acquiring images of metal sheets;

[0014] An image preprocessing module, configured to perform image preprocessing on the metal plate image to obtain a preprocessed image;

[0015] A defect recognition module, configured to obtain defect information of the corresponding metal sheet according to the pre-processed image recognition;

[0016] The quality classification module is used to evaluate the quality type of the corresponding metal sheet according to the defect information and a preset classification strategy.

[0017] Furthermore, the defect recognition module specifically obtains defect information of the corresponding metal sheet through a trained first deep learning model, wherein the first deep learning model includes:

[0018] A ResNet50 backbone network is used to extract feature information of the preprocessed image and obtain first feature maps at multiple different levels;

[0019] A feature pyramid network is used to fuse the first feature maps at different levels to generate multiple second feature maps at different scales.

[0020] A region proposal network is used to generate defect feature candidate regions in the second feature map;

[0021] The RoI Align pooling layer is used to extract the region of interest from the defect feature candidate region and then perform alignment using the RoI Align technology. Finally, a third feature map is obtained through the pooling operation.

[0022] The fully connected layer is used to perform a classification regression operation on the third feature map and output defect information.

[0023] Furthermore, the information update module updates the real-time production order information and real-time production line information, specifically: adding the number of metal plates with scrap quality type to the real-time number of plates to be produced, and updating the real-time status of the production line where the equipment with mechanical failure is located to the maintenance status.

[0024] Furthermore, the delivery status is predicted based on the real-time production order information and the real-time production line information, and a regeneration instruction is sent to the plan generation module based on the predicted delivery status, specifically:

[0025] Based on real-time production order information and real-time production line information, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction;

[0026] If the real-time status of the production line started in the current production scheduling plan changes to maintenance status, then based on the real-time production order information and real-time production line information, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction.

[0027] Furthermore, the priority coefficient P is specifically calculated as follows: P=1 / (K1×S+K2×T+K3×N), where K1, K2 and K3 represent the first preset weight, the second preset weight and the third preset weight, respectively.

[0028] Furthermore, the fault analysis module locates the equipment that has failed in the corresponding production line, specifically by locating the equipment that has failed in the corresponding production line according to the defect information of the metal sheet through a trained second deep learning model.

[0029] Furthermore, the fault analysis module analyzes the fault type of the faulty equipment, specifically by analyzing the similarity of the location areas of defects of the same defect type based on the defect information of multiple consecutive metal plates through a trained third deep learning model. If the location areas are similar, the fault type of the faulty equipment corresponding to the defect type is a parameter fault; otherwise, the fault type of the corresponding faulty equipment is a mechanical fault.

[0030] The present invention also provides a metal sheet production scheduling method based on artificial intelligence, comprising the following steps:

[0031] Acquire real-time production order information for metal sheets and real-time production line information of the production workshop, wherein the real-time production order information includes the real-time number of sheets to be produced and the delivery deadline of each order, and the real-time production line information includes the production line number, production efficiency, energy consumption information, and real-time status of each production line, wherein the real-time status includes normal status and maintenance status;

[0032] Identifying defect information of produced metal sheets and evaluating the quality type of the corresponding metal sheets based on the defect information, wherein the quality type includes acceptable, repairable, and scrapped;

[0033] Based on the defect information of the metal sheet, locate the faulty equipment in the corresponding production line and analyze the fault type of the faulty equipment, including parameter failure and mechanical failure;

[0034] Updating the real-time production order information and real-time production line information, specifically: adding the number of metal sheets with a quality type of scrap to the real-time number of sheets to be produced, and updating the real-time status of the production line where the equipment with mechanical failure is located to the maintenance status;

[0035] Analyze and predict delivery status based on real-time production order information and real-time production line information, and send regeneration instructions based on the predicted delivery status;

[0036] After receiving the regeneration instruction, the system generates multiple production scheduling plans based on real-time production order information and real-time production line information through a trained scheduling algorithm model to ensure that all orders are delivered on time. The production scheduling plans record the corresponding scheduling information, including the number of production lines started S, the estimated production time T, and the estimated total energy consumption V. The production lines started in the production scheduling plans are all in normal status;

[0037] The priority coefficient P of each production scheduling plan is calculated according to the scheduling information, and finally the production scheduling plan with the largest priority coefficient is selected as the new production scheduling plan.

[0038] Furthermore, the delivery status is predicted based on the real-time production order information and the real-time production line information, and a regeneration instruction is sent to the plan generation module based on the predicted delivery status, specifically:

[0039] Based on real-time production order information and real-time production line information, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction;

[0040] If the real-time status of the production line started in the current production scheduling plan changes to maintenance status, then based on the real-time production order information and real-time production line information, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The present invention updates the real-time production order information and real-time production line information by analyzing the number of scrapped metal sheets and the production lines with mechanical failure equipment, analyzes and predicts the delivery status based on the updated information, and sends a regeneration instruction to the plan generation module based on the predicted delivery status. After receiving the regeneration instruction, the plan generation module generates multiple production scheduling plans that can ensure that all orders are delivered on time. Finally, the plan selection module calculates the priority coefficient of each production scheduling plan based on the estimated production time and estimated total energy consumption of the number of production lines to select the optimal production scheduling plan, so as to ensure that the production can complete the delivery of each order within the delivery period of each order, thereby improving the overall production efficiency and reducing production energy consumption.

[0043] In the first deep learning model of the present invention, a ResNet50 backbone network is used to replace the feature extraction network of the original Faster R-CNN learning model. By introducing residual blocks, the common gradient vanishing or exploding problems in deep network training are solved, effectively learning deeper feature representations. A feature pyramid network is added to the basic architecture of the original Faster R-CNN learning model to achieve high-level semantics while retaining details, thereby helping the model to better detect smaller features. By combining feature information at different levels, the model is made more flexible and powerful when processing irregularly shaped defects, more effectively detecting various defect types on the surface of metal sheets, and improving overall detection accuracy. The RoI Align pooling layer is used to replace the region of interest pooling layer of the original Faster R-CNN learning model to avoid positional deviation when the region of interest is divided on the second feature map, better retain information, improve detection accuracy, and thus improve the accuracy of the quality classification assessment of metal sheets, thereby ensuring the accuracy of equipment failure analysis and the accuracy of predicted delivery status analysis, and ensuring that the generated production scheduling plan can ensure that all orders are delivered on time. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 This is a structural diagram of a metal sheet production scheduling system based on artificial intelligence of the present invention;

[0047] Figure 2This is a structural block diagram of a sheet metal quality assessment module of an artificial intelligence-based metal sheet metal production scheduling system of the present invention;

[0048] Figure 3 This is a flow chart of a metal sheet production scheduling method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0051] In addition, the descriptions of "first", "second", etc. in the present invention are for descriptive purposes only and should not be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0052] Example 1

[0053] See also Figure 1 As shown, the present invention provides an artificial intelligence-based metal sheet production scheduling system, comprising:

[0054] Data acquisition module, plate quality assessment module, fault analysis module, information update module, delivery analysis module, plan generation module and plan selection module, as well as a central processing unit that communicates with the above modules.

[0055] The data acquisition module is used to obtain real-time production order information of metal plates and real-time production line information of the production workshop. The real-time production order information includes the real-time number of plates to be produced and the delivery deadline of each order. The real-time production line information includes the real-time production line number, production efficiency, energy consumption information and real-time status corresponding to each production line.

[0056] The data acquisition module obtains real-time production order information of metal sheets, which can be obtained by connecting with an ERP system. The ERP system, namely Enterprise Resource Planning (ERP), is used to generate and store production order information.

[0057] The data acquisition module obtains real-time production line information of the production workshop, which can be obtained by connecting with the MES system. The MES system, namely the Manufacturing Execution System, is used to monitor and collect data from each production line in the production workshop in real time, and can obtain production line information through data analysis. The energy consumption information of the production line reflects the total energy consumption generated by the corresponding production line on average per hour.

[0058] The plate quality assessment module is used to identify defect information of the produced metal plates and assess the quality type of the corresponding metal plates according to the defect information, wherein the quality types include qualified, repairable and scrapped.

[0059] See Figure 2 As shown, the plate quality assessment module includes: an image acquisition module, an image preprocessing module, a defect recognition module and a quality classification module.

[0060] The image acquisition module is used to acquire the image of the metal plate.

[0061] The image acquisition module can specifically acquire images through an industrial camera, which has the characteristics of high resolution, high precision, high definition, good color reproduction and low noise.

[0062] In some embodiments, the industrial camera may optionally be an area array CCD camera.

[0063] The image preprocessing module is used to perform image preprocessing operations on the metal plate image to obtain a preprocessed image.

[0064] The image preprocessing operation includes noise removal processing, image enhancement processing, image normalization processing, etc.

[0065] During the image acquisition and transmission process, images are easily disturbed by the surrounding environment and various other factors, and noise is inevitable. In order to meet the visual requirements for image clarity and subsequent image processing, it is necessary to remove noise from the image. Usually, noise is statistically approximated as additive Gaussian white noise. The noise removal process can be performed through any existing filtering method such as Gaussian filtering, median filtering, mean filtering, and bilateral filtering.

[0066] In some embodiments, the image is first edge detected before the noise removal process, and the image is divided into an edge region and a smooth region. The edge region and the smooth region are then denoised using a non-local mean filtering denoising algorithm and a bilateral filtering denoising algorithm. Bilateral filtering is computationally simple and easy to implement. Since bilateral filtering is based on adjacent pixels, the effect on preserving edge structure is not very obvious. Non-local mean filtering is a filtering method based on the similarity between image blocks, which can well preserve the structural information of the image, but has a high computational complexity. Therefore, this solution uses a non-local mean filtering denoising algorithm and a bilateral filtering denoising algorithm to denoise the edge region and the smooth region respectively, quickly achieving the goal of clearly retaining the denoising of the edge region and completely removing the noise in the smooth region.

[0067] The defect recognition module is used to obtain defect information of the corresponding metal plate based on the preprocessed image recognition.

[0068] The defect information includes: defect type, defect location and defect degree.

[0069] In some embodiments, the defect recognition module obtains defect information of the corresponding metal sheet based on the image recognition of the metal sheet after image preprocessing operation, and selects to obtain the defect information through the trained Faster R-CNN learning model. The existing Faster R-CNN model includes a feature extraction network, which uses a set of basic convolutional layers, activation layers and pooling layers to extract useful feature maps from the input image; a region proposal network, which is used to generate candidate regions that potentially contain detection targets; a region of interest pooling layer, which combines the input feature map and the candidate region, and then integrates this information to extract specific feature maps of interest; a fully connected network, which classifies the image through an activation function and uses border regression to obtain a precisely positioned detection frame; the specific recognition method and algorithm of the Faster R-CNN learning model are existing technologies and will not be repeated here.

[0070] In some embodiments, the defect recognition module obtains defect information of the corresponding metal sheet based on the image recognition of the metal sheet after image preprocessing operation, and selects a trained first deep learning model for recognition and acquisition.

[0071] The first deep learning model is improved based on the existing Faster R-CNN model, and the first deep learning model includes:

[0072] A ResNet50 backbone network is used to extract feature information of the preprocessed image and obtain first feature maps at multiple different levels;

[0073] A feature pyramid network is used to fuse the first feature maps at different levels to generate multiple second feature maps at different scales.

[0074] A region proposal network is used to generate defect feature candidate regions in the second feature map;

[0075] The RoI Align pooling layer is used to extract the region of interest from the defect feature candidate region and then perform alignment using the RoI Align technology. Finally, a third feature map is obtained through the pooling operation.

[0076] The fully connected layer is used to perform a classification regression operation on the third feature map and output defect information.

[0077] In the first deep learning model, the ResNet50 backbone network is used to replace the feature extraction network of the original Faster R-CNN learning model. The feature extraction network of the original Faster R-CNN learning model basically uses convolutional neural networks (such as the VGG series). When the number of layers increases, it is prone to gradient vanishing or exploding problems, which makes training difficult and the model difficult to converge. In addition, the feature representation ability is relatively weak, especially when dealing with complex scenes or multi-scale targets. ResNet50 is a deep residual network that solves the common gradient vanishing or exploding problems in deep network training by introducing residual blocks, and can effectively learn deeper feature representations.

[0078] In the first deep learning model, a feature pyramid network was added to the infrastructure of the original Faster R-CNN learning model. The feature pyramid network (FPN) is an image feature extraction method used for object detection and semantic segmentation. The fusion process is an existing technology and will not be described here. After obtaining features through ResNet50, the image will be compressed, resulting in smaller feature candidate boxes generated subsequently, which to a certain extent reduces the model's feature discrimination function. The addition of the feature pyramid network fusion technology can achieve high-level semantics while retaining details, thereby helping the model to better detect smaller features. By combining feature information at different levels, the model is more flexible and powerful in dealing with irregular-shaped defects, more effectively detecting various defect types on the surface of metal sheets, and improving overall detection accuracy.

[0079] In the first deep learning model, the RoI Align pooling layer is used to replace the region of interest pooling layer of the original Faster R-CNN learning model. The region of interest pooling layer of the original Faster R-CNN learning model will produce two quantizations when quantizing the input feature map. These two quantization operations are prone to information loss, especially for smaller regions of interest. RoI Align uses the bilinear interpolation method to convert the two quantization processes into a continuous operation, avoiding position deviation when the region of interest is divided on the second feature map, better retaining information, and improving inspection accuracy.

[0080] The quality classification module is used to evaluate the quality type of the corresponding metal plate according to a preset classification strategy based on the defect information.

[0081] The preset classification strategy is specifically formulated by the user according to actual needs.

[0082] The metal plates with qualified quality type are plates that can be delivered immediately, the metal plates with repairable quality type are plates whose defects can be reduced to qualified or eliminated, and the metal plates with scrapped quality type are plates whose defects cannot be reduced to qualified or eliminated, that is, the plates cannot be delivered.

[0083] The fault analysis module is used to locate the faulty equipment in the corresponding production line based on the defect information of the metal sheet and analyze the fault type of the faulty equipment, which includes parameter fault and mechanical fault.

[0084] The fault analysis module locates the equipment that has failed in the corresponding production line, specifically by using a trained second deep learning model to locate the equipment that has failed in the corresponding production line based on the defect information of the metal sheet.

[0085] Generally speaking, in a metal sheet production line, each major device corresponds to a different production process, and failures of different major devices result in different major defect types in the metal sheet.

[0086] The fault analysis module analyzes the fault type of the faulty equipment. Specifically, it uses a trained third deep learning model to analyze the similarity of the location areas of defects of the same defect type based on the defect information of multiple consecutive metal plates. If the location areas are similar, the fault type of the faulty equipment corresponding to the defect type is a parameter fault; otherwise, the fault type of the corresponding faulty equipment is a mechanical fault.

[0087] Generally speaking, when there is a parameter failure, the location areas where defects of the same defect type appear in continuous metal sheets are basically similar. However, when there is a mechanical failure, due to the instability of the equipment machinery, the location areas where defects of the same defect type appear in continuous metal sheets are also unstable and irregular.

[0088] Parameter failure means that there are inaccuracies in the settings or programs of the equipment parameters, and the parameters can be corrected; while mechanical failure means that there are failures in the equipment components or other parts, and the operation needs to be stopped and the staff needs to be notified for maintenance. During this process, the production of the production line needs to be stopped.

[0089] The information update module is used to update the real-time production order information and real-time production line information, specifically: adding the number of metal plates with scrap quality type to the real-time number of plates to be produced, and updating the real-time status of the production line where the equipment with mechanical failure is located to the maintenance status.

[0090] The delivery analysis module is used to analyze and predict delivery status based on real-time production order information and real-time production line information, and send regeneration instructions to the plan generation module based on the predicted delivery status. Specifically:

[0091] Based on real-time production order information and real-time production line information, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction;

[0092] If the real-time status of the production line started in the current production scheduling plan changes to maintenance status, then based on the real-time production order information and real-time production line information, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction.

[0093] The plan generation module is used to generate multiple production scheduling plans that ensure that all orders are delivered on time based on real-time production order information and real-time production line information after receiving the regeneration instruction through a trained scheduling algorithm model. The production scheduling plan records the corresponding scheduling information, including the number of production lines started S, the estimated production time T and the estimated total energy consumption V. The production lines started in the production scheduling plan are all in normal status.

[0094] It should be noted that the production scheduling plan at the beginning of production is also generated and formulated by the plan generation module.

[0095] The plan selection module is used to calculate the priority coefficient P of each production scheduling plan based on the scheduling information, and finally select the production scheduling plan with the largest priority coefficient as the new production scheduling plan, P=1 / (K1×S+K2×T+K3×N), K1, K2 and K3 represent the first preset weight, the second preset weight and the third preset weight respectively.

[0096] If it is known that all orders can be delivered on time, the production scheduling plan needs to take into account energy consumption, the number of production lines used, and working hours to reduce equipment wear, K3>K1>K2, and K1+K2+K3=1.

[0097] Example 2

[0098] See Figure 3 As shown, the present invention also provides an artificial intelligence-based metal sheet production scheduling method, which is applied to the above-mentioned artificial intelligence-based metal sheet production scheduling system, and specifically includes the following steps:

[0099] S1. Obtaining real-time production order information of metal sheets and real-time production line information of production workshops;

[0100] S2. identifying defect information of the produced metal sheet and evaluating the quality type of the corresponding metal sheet based on the defect information;

[0101] S3. Locate the faulty equipment in the corresponding production line based on the defect information of the metal sheet and analyze the fault type of the faulty equipment;

[0102] S4. Update the real-time production order information and real-time production line information;

[0103] S5. Analyze and predict delivery status based on real-time production order information and real-time production line information, and send a regeneration instruction based on the predicted delivery status;

[0104] S6. After receiving the regeneration instruction, generate multiple production scheduling plans that ensure that all orders are delivered on time based on the real-time production order information and the real-time production line information using a trained scheduling algorithm model, wherein the production scheduling plans record the corresponding scheduling information;

[0105] S7. Calculate the priority coefficient P of each production scheduling plan based on the scheduling information, and finally select the production scheduling plan with the largest priority coefficient as the new production scheduling plan.

[0106] For step S1, the real-time production order information includes the real-time number of plates to be produced and the delivery deadline of each order, and the real-time production line information includes the production line number, production efficiency, energy consumption information and real-time status corresponding to each production line.

[0107] Regarding step S2, the quality types include qualified, repairable, and scrapped. The specific defect identification and quality type assessment steps are as follows:

[0108] T1. Acquire metal sheet images;

[0109] T2. performing an image preprocessing operation on the metal plate image to obtain a preprocessed image;

[0110] T3. Obtaining defect information of the corresponding metal plate according to the preprocessed image recognition, wherein the defect information includes: defect type, defect location, and defect degree;

[0111] T4. Evaluate the quality type of the corresponding metal sheet according to the defect information and a preset classification strategy.

[0112] In step T3, defect information is specifically obtained by identifying using a trained first deep learning model. The first deep learning model is improved based on the existing Faster R-CNN model. The first deep learning model includes:

[0113] A ResNet50 backbone network is used to extract feature information of the preprocessed image and obtain first feature maps at multiple different levels;

[0114] A feature pyramid network is used to fuse the first feature maps at different levels to generate multiple second feature maps at different scales.

[0115] A region proposal network is used to generate defect feature candidate regions in the second feature map;

[0116] The RoI Align pooling layer is used to extract the region of interest from the defect feature candidate region and then perform alignment using the RoI Align technology. Finally, a third feature map is obtained through the pooling operation.

[0117] The fully connected layer is used to perform a classification regression operation on the third feature map and output defect information.

[0118] Corresponding to step S3, the fault types include parameter fault and mechanical fault;

[0119] Locating the faulty equipment in the corresponding production line, specifically by using the trained second deep learning model to locate the faulty equipment in the corresponding production line based on the defect information of the metal sheet;

[0120] Analyze the fault type of the faulty equipment. Specifically, the trained third deep learning model is used to analyze the similarity of the defect location areas of the same defect type based on the defect information of multiple consecutive metal plates. If the location areas are similar, the fault type of the faulty equipment corresponding to the defect type is a parameter fault. Otherwise, the fault type of the corresponding faulty equipment is a mechanical fault.

[0121] In step S4, specifically, the number of metal sheets of scrap quality type is added to the real-time number of sheets to be produced, and the real-time status of the production line where the equipment with mechanical failure is located is updated to the maintenance status.

[0122] Regarding step S5, the delivery status is predicted based on the real-time production order information and the real-time production line information, and a regeneration instruction is sent to the plan generation module based on the predicted delivery status, specifically:

[0123] Based on real-time production order information and real-time production line information, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction;

[0124] If the real-time status of the production line started in the current production scheduling plan changes to maintenance status, then based on the real-time production order information and real-time production line information, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction.

[0125] For step S6, the scheduling information includes the number of production lines started S, the estimated production time T and the estimated total energy consumption V. The production lines started in the production scheduling plan are all in normal status.

[0126] For step S7, P=1 / (K1×S+K2×T+K3×N), where K1, K2, and K3 represent the first preset weight, the second preset weight, and the third preset weight, respectively.

[0127] If it is known that all orders can be delivered on time, the production scheduling plan needs to take into account energy consumption, the number of production lines used, and working hours to reduce equipment wear, K3>K1>K2, and K1+K2+K3=1.

[0128] Example 3

[0129] The present invention also provides an electronic device, comprising: a processor, a sending device, an input device, an output device and a memory. The processor can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, and is used to execute relevant programs to implement the technical solution provided in the embodiments of the present application. The memory can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device or a random access memory (RAM), and is used to store computer program code. The computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes a method as described in any of the above possible implementation methods.

[0130] Example 4

[0131] The present invention also provides a computer-readable storage medium, in which a computer program is stored. The computer program includes program instructions. When the program instructions are executed by a processor of an electronic device, the processor executes a method as described in any one of the possible implementation methods described above.

[0132] The beneficial effects of the present invention are:

[0133] The present invention updates the real-time production order information and real-time production line information by analyzing the number of scrapped metal sheets and the production lines with mechanical failure equipment, analyzes and predicts the delivery status based on the updated information, and sends a regeneration instruction to the plan generation module based on the predicted delivery status. After receiving the regeneration instruction, the plan generation module generates multiple production scheduling plans that can ensure that all orders are delivered on time. Finally, the plan selection module calculates the priority coefficient of each production scheduling plan based on the estimated production time and estimated total energy consumption of the number of production lines to select the optimal production scheduling plan, so as to ensure that the production can complete the delivery of each order within the delivery period of each order, thereby improving the overall production efficiency and reducing production energy consumption.

[0134] In the first deep learning model of the present invention, a ResNet50 backbone network is used to replace the feature extraction network of the original Faster R-CNN learning model. By introducing residual blocks, the common gradient vanishing or exploding problems in deep network training are solved, effectively learning deeper feature representations. A feature pyramid network is added to the basic architecture of the original Faster R-CNN learning model to achieve high-level semantics while retaining details, thereby helping the model to better detect smaller features. By combining feature information at different levels, the model is made more flexible and powerful when processing irregularly shaped defects, more effectively detecting various defect types on the surface of metal sheets, and improving overall detection accuracy. The RoI Align pooling layer is used to replace the region of interest pooling layer of the original Faster R-CNN learning model to avoid positional deviation when the region of interest is divided on the second feature map, better retain information, improve detection accuracy, and thus improve the accuracy of the quality classification assessment of metal sheets, thereby ensuring the accuracy of equipment failure analysis and the accuracy of predicted delivery status analysis, and ensuring that the generated production scheduling plan can ensure that all orders are delivered on time.

[0135] Throughout the specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0136] In addition, the functional units in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The aforementioned storage medium includes various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0137] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is intended to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A metal sheet production scheduling system based on artificial intelligence, characterized in that: include: A data acquisition module is used to obtain real-time production order information of metal sheets and real-time production line information of the production workshop. The real-time production order information includes the real-time number of sheets to be produced and the delivery deadline of each order. The real-time production line information includes the real-time production line number, production efficiency, energy consumption information and real-time status of each production line, and the real-time status includes normal status and maintenance status. A sheet quality assessment module, configured to identify defect information of produced metal sheets and assess the quality type of the corresponding metal sheets based on the defect information, wherein the quality types include acceptable, repairable, and scrapped; A fault analysis module is used to locate faulty equipment in the corresponding production line based on the defect information of the metal sheet and analyze the fault type of the faulty equipment, which includes parameter faults and mechanical faults. Specifically, a trained third deep learning model is used to analyze the similarity of the location areas of defects of the same defect type based on the defect information of multiple consecutive metal sheets. If the location areas are similar, the fault type of the faulty equipment corresponding to the defect type is a parameter fault; otherwise, the fault type of the faulty equipment is a mechanical fault. An information update module is used to update the real-time production order information and real-time production line information, specifically by adding the number of metal sheets with a quality type of scrap to the real-time number of sheets to be produced, and updating the real-time status of the production line where the equipment with mechanical failure is located to the maintenance status; The delivery analysis module is used to analyze and predict delivery status based on real-time production order information and real-time production line information, and send regeneration instructions to the plan generation module based on the predicted delivery status. Specifically: Based on the real-time production order information and real-time production line information, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction. If the real-time status of the production line started in the current production scheduling plan changes to the maintenance state, then based on the real-time production order information and real-time production line information, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order. If not, send a regeneration instruction. A plan generation module is configured to, upon receiving a regeneration instruction, generate multiple production scheduling plans based on real-time production order information and real-time production line information using a trained scheduling algorithm model to ensure that all orders are delivered on time. The production scheduling plans record corresponding scheduling information, including the number of production lines activated S, the estimated production time T, and the estimated total energy consumption V. The production lines activated in the production scheduling plans are all in normal status; The plan selection module is used to calculate the priority coefficient P of each production scheduling plan based on the scheduling information, and finally select the production scheduling plan with the largest priority coefficient as the new production scheduling plan; The plate quality assessment module includes: an image acquisition module for acquiring an image of a metal plate; an image preprocessing module for performing image preprocessing operations on the metal plate image to obtain a preprocessed image; a defect recognition module for obtaining defect information of the corresponding metal plate based on the preprocessed image; and a quality classification module for evaluating the quality type of the corresponding metal plate based on the defect information according to a preset classification strategy. The defect recognition module specifically obtains defect information of the corresponding metal sheet through a trained first deep learning model. The first deep learning model includes: a ResNet50 backbone network for extracting feature information of the preprocessed image to obtain multiple first feature maps at different levels; a feature pyramid network for fusing the first feature maps at different levels to generate multiple second feature maps at different scales; a region proposal network for generating defect feature candidate regions in the second feature map; a RoI Align pooling layer for extracting a region of interest from the defect feature candidate region and performing an alignment operation using the RoI Align technology, and finally obtaining a third feature map through a pooling operation; and a fully connected layer for performing a classification and regression operation on the third feature map to output defect information. The priority coefficient P is specifically calculated as follows: P=1 / (K1×S+K2×T+K3×N), where K1, K2, and K3 represent the first preset weight, the second preset weight, and the third preset weight, respectively. The fault analysis module locates the equipment that has failed in the corresponding production line, specifically by locating the equipment that has failed in the corresponding production line according to the defect information of the metal sheet through a trained second deep learning model.

2. A metal sheet production scheduling method based on artificial intelligence, applied to the metal sheet production scheduling system based on artificial intelligence according to claim 1, characterized in that: The following steps are involved: Acquire real-time production order information for metal sheets and real-time production line information of the production workshop, wherein the real-time production order information includes the real-time number of sheets to be produced and the delivery deadline of each order, and the real-time production line information includes the production line number, production efficiency, energy consumption information, and real-time status of each production line, wherein the real-time status includes normal status and maintenance status; Identifying defect information of produced metal sheets and evaluating the quality type of the corresponding metal sheets based on the defect information, wherein the quality type includes acceptable, repairable, and scrapped; Based on the defect information of the metal sheets, the faulty equipment in the corresponding production line is located and the fault type of the faulty equipment is analyzed. The fault types include parameter faults and mechanical faults. Specifically, the trained third deep learning model is used to analyze the similarity of the location areas of defects of the same defect type based on the defect information of multiple consecutive metal sheets. If the location areas are similar, the fault type of the faulty equipment corresponding to the defect type is parameter fault; otherwise, the fault type of the faulty equipment is mechanical fault. Updating the real-time production order information and real-time production line information, specifically: adding the number of metal sheets with a quality type of scrap to the real-time number of sheets to be produced, and updating the real-time status of the production line where the equipment with mechanical failure is located to the maintenance status; Analyze the predicted delivery situation based on the real-time production order information and the real-time production line information, and send a regeneration instruction based on the predicted delivery situation. Specifically, analyze whether the current production scheduling plan can complete all orders within the latest delivery deadline in the order based on the real-time production order information and the real-time production line information. If not, send a regeneration instruction. If the real-time status of the production line started in the current production scheduling plan changes to the maintenance status, analyze whether the production of the remaining production lines in the current production scheduling plan with normal real-time status can complete all orders within the latest delivery deadline in the order based on the real-time production order information and the real-time production line information. If not, send a regeneration instruction. After receiving the regeneration instruction, the system generates multiple production scheduling plans based on real-time production order information and real-time production line information through a trained scheduling algorithm model to ensure that all orders are delivered on time. The production scheduling plans record the corresponding scheduling information, including the number of production lines started S, the estimated production time T, and the estimated total energy consumption V. The production lines started in the production scheduling plans are all in normal status; Calculate the priority coefficient P of each production scheduling plan based on the scheduling information, and finally select the production scheduling plan with the largest priority coefficient as the new production scheduling plan; The method of identifying defect information of the produced metal sheet and evaluating the quality type of the corresponding metal sheet based on the defect information specifically comprises: obtaining an image of the metal sheet; performing an image preprocessing operation on the metal sheet image to obtain a preprocessed image; obtaining defect information of the corresponding metal sheet based on the preprocessed image; and evaluating the quality type of the corresponding metal sheet based on the defect information and a preset classification strategy; The defect information of the corresponding metal sheet is obtained by identifying the defect information of the corresponding metal sheet according to the preprocessed image, specifically by identifying the defect information of the corresponding metal sheet through a trained first deep learning model, wherein the first deep learning model includes: a ResNet50 backbone network for extracting feature information of the preprocessed image to obtain multiple first feature maps at different levels; a feature pyramid network for fusing the first feature maps at different levels to generate multiple second feature maps at different scales; a region proposal network for generating defect feature candidate regions in the second feature map; a RoI Align pooling layer for extracting a region of interest from the defect feature candidate region and performing an alignment operation using the RoI Align technology, and finally obtaining a third feature map through a pooling operation; and a fully connected layer for performing a classification regression operation on the third feature map to output defect information. The priority coefficient P is specifically calculated as follows: P=1 / (K1×S+K2×T+K3×N), where K1, K2, and K3 represent the first preset weight, the second preset weight, and the third preset weight, respectively. The positioning corresponds to the equipment that has failed in the production line, specifically positioning the equipment that has failed in the production line according to the defect information of the metal sheet through a trained second deep learning model.

Images