Inkjet printing work order production method and system based on internet of things

Abstract

The present application relates to the field of data processing, especially to a kind of inkjet printing work order production scheduling method and system based on Internet of Things, method includes: obtaining the complexity multidimensional data of standard printing image on work order;Standard printing image and actual printing image are carried out feature matching, the result of feature matching is weighted summation according to complexity multidimensional data, obtain work order total score;Neural network model is constructed and trained;The complexity multidimensional data of new work order and equipment are input into neural network model, and the prediction value of production quality score is output;Based on all new work order, equipment and production quality score, matching is carried out using KM algorithm, and the optimal selection of each work order is assigned to equipment.The present application realizes the accurate matching and load balancing of work order and equipment by "printing complexity quantification-equipment quality evaluation-quality prediction modeling-dynamic production scheduling optimization" four-step closed-loop logic, the whole production scheduling process is intelligent, and the production scheduling efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing. In particular, it relates to a method and system for scheduling inkjet printing work orders based on the Internet of Things (IoT). Background Technology

[0002] Inkjet printing, as an important process in the textile printing and dyeing industry, has advantages such as strong customization capabilities and flexible small-batch production, making it a core technology for meeting diverse market demands. However, with the increasing complexity of orders and the diversification of equipment types, the traditional work order scheduling method has gradually revealed significant shortcomings.

[0003] For example, Chinese patent application CN119991145A discloses an automatic planning digital printing control system, including an order upload analysis and planning system and multiple digital code recognition and printing units; each digital code recognition and printing unit has a unique identification code, and the order upload analysis and planning system manages the digital code recognition and printing units based on the unique identification code; the unique identification code is bound to the size of the textile product with the printed pattern; the server identifies the unique identification code in the printing code and sends the printing code to the corresponding digital code recognition and printing unit with the unique identification code; the digital code recognition and printing unit performs digital printing production of textile products based on the printing code, and each printing code is bound to one textile product.

[0004] While the above-mentioned technical solution has achieved automated planning and printing of textile product patterns, it cannot reasonably schedule production based on the specific complexity of the order, the actual production quality of the equipment, and the load of the equipment during the automatic planning of orders, which directly affects product quality, production efficiency, and equipment utilization. Summary of the Invention

[0005] To address the aforementioned technical problem of being unable to rationally schedule production based on the specific complexity of orders, the actual production quality of equipment, and the load capacity of equipment, this invention provides solutions in the following aspects.

[0006] In the first aspect, the IoT-based inkjet printing work order scheduling method includes:

[0007] Obtain complex multidimensional data of standard printed images from historical production work orders of the equipment;

[0008] Feature matching is performed on the standard printing image and the actual printing image produced by a single device. The result of feature matching is weighted and summed according to the complexity multidimensional data to obtain the total work order score of a single device on the corresponding historical work order.

[0009] A neural network model is constructed and trained, wherein the complexity multidimensional data and the corresponding equipment are used as inputs, the total score of the work order is used as a label, and the predicted value of the production quality score of each equipment work order is output; for a single new work order, its complexity multidimensional data is calculated, the complexity multidimensional data of the new work order and all equipment are input into the trained neural network model, and the predicted value of all production quality scores is output.

[0010] Based on all new work orders, equipment, and the predicted values ​​of their corresponding production quality scores, the KM algorithm is used for matching to obtain the optimal equipment for each work order.

[0011] By adopting the above technical solutions, the complexity of the printing image is quantified, accurately capturing the technical requirements of the printing image for the equipment, avoiding the bias of traditional manual experience judgment, and ensuring that complex printing is assigned to the appropriate equipment, thereby significantly improving the product quality compliance rate. Through the registration and feature matching of actual printing images and standard printing images, the actual production quality score of the work order on specific equipment is quantified. At the same time, a predictive model is built using neural networks to replace the traditional equipment evaluation method that relies on physical parameters, further improving the stability and consistency of production quality. Intelligent scheduling using the KM algorithm reduces manual intervention and improves production efficiency and equipment utilization.

[0012] Preferably, the complexity multidimensional data includes edge density data, curvature variation data, and color contrast data.

[0013] By adopting the above technical solution, the key visual features of the printed image on the work order are decomposed, and the complexity of the printed image is transformed into a three-dimensional vector of edge density, curvature change and color contrast, which can accurately capture the technical requirements of the printed image for the equipment.

[0014] Preferably, the process for obtaining the edge density data of the standard printed image on the work order is as follows:

[0015] The edge of a standard printed image is detected using the Canny algorithm to obtain an edge image; the number of edge pixels and the total number of pixels in the edge image are counted; the ratio of the number of edge pixels to the total number of pixels is calculated to obtain edge density data.

[0016] By adopting the above technical solution, the edge density data reflects the complexity of the printed image outline. The denser the edges, the higher the edge density data, the more outline points the equipment needs to locate, and the higher the positioning accuracy requirement of the equipment.

[0017] Preferably, the process for obtaining the curvature change data of the standard printed image on the work order is as follows:

[0018] An 8-connected component tracing algorithm is used to extract continuous edge contours from the edge image, resulting in multiple continuous edge contours. Each continuous edge contour is a set of pixels that make up the continuous edge contour. For each set of pixels in a continuous edge contour, the curvature of each pixel is calculated using the central difference method, which utilizes the left and right adjacent points of each pixel. The ratio of the number of points in a single continuous edge contour to the number of edge pixels is calculated to obtain the weight of each continuous edge contour. The weighted average curvature of each continuous edge contour is calculated based on the curvature of each pixel and the weight of the continuous edge contour to which each pixel belongs. The curvature variance is calculated based on the curvature of each pixel and the weighted average curvature to obtain curvature variation data.

[0019] By adopting the above technical solution, the curvature change data reflects the degree of curvature change at the edge of the printed image. The larger the curvature change data, the more drastic the curvature change at the edge of the printed image, requiring the equipment to quickly adjust the direction, speed, and ink droplet size in a short time. The smaller the curvature change data, the smoother the edge of the printed image, and the equipment only needs to output stably without frequent adjustments.

[0020] Preferably, the process for obtaining the color contrast data of the standard printed image on the work order is as follows:

[0021] On the edge image, for each edge pixel, take the pixels on both sides of its normal direction; convert the RGB color to the LAB color space, calculate the color difference between the pixels on both sides of all edge pixels and take the average value to obtain the color contrast data.

[0022] By adopting the above technical solution, color contrast data represents the degree of color difference between the two sides of the edge pixel. The greater the difference, the stronger the visual contrast of the printed image, and the higher the requirements for the control of ink droplet concentration of the equipment.

[0023] Preferably, the process of feature matching between the standard printed image and the actual printed image produced by a single device is as follows:

[0024] The actual printed image and the standard printed image are registered to obtain the registered image;

[0025] Edge feature matching is performed on the standard printed image and the registration image. In the edge image of the standard printed image, a standard block is taken with each edge pixel as the center. A registration block of the same size is taken at the same position in the registration image. The similarity between the standard block and the corresponding registration block is calculated. The average of the similarity of all edge pixels is taken to obtain the edge accuracy score.

[0026] The standard printed image and the actual printed image are converted to the LAB color space. Color feature matching is performed on the standard printed image and the actual printed image. For each pixel, the color difference between the actual printed image and the standard printed image is calculated. The gradient magnitude of each pixel in the standard printed image is calculated. After normalization of the gradient magnitude, it is used as a weight. The weighted average of the color difference of all pixels is calculated to obtain the color fidelity score.

[0027] Preferably, the process for obtaining the total work order score of the single device on the corresponding historical work order is as follows:

[0028] The ratio of the sum of edge density data and curvature variation data to the sum of edge density data, curvature variation data and color contrast data is used as the weight; the sum of the product of the difference after subtracting the weight from 1 and the color fidelity score, and the product of the edge accuracy score and the weight, is used to obtain the total score of the work order.

[0029] By adopting the above technical solution, the weights are calculated in real time based on the complexity of multidimensional data (edge ​​density data, curvature change data, and color contrast data) of standard printed images, so that the overall score can adaptively switch between "edge clarity" and "color fidelity", ensuring that the print can be accurately evaluated in the most critical dimensions, thereby directly reflecting the compatibility between the printed image and the equipment capabilities.

[0030] Preferably, the process of using the KM algorithm to match and obtain the optimal allocation of each work order to the equipment includes the following steps:

[0031] The KM algorithm is used to perform a preliminary matching of new work orders and equipment, and the results of the preliminary matching are adjusted according to the current load of the equipment to obtain the adjustment plan for allocating each work order to the equipment.

[0032] After obtaining the adjustment plan, based on the current average load of all equipment, if the equipment corresponding to the new work order is a high-load equipment, then the work order is moved to a low-load equipment for production; wherein, if the predicted value of the production quality score calculated after the work order is moved to a low-load equipment is less than the predicted value of the production quality score calculated when it was originally assigned to a high-load equipment, then no movement operation is performed.

[0033] By adopting the above technical solutions, the KM algorithm can efficiently find an optimal equipment allocation scheme for each work order, thereby maximizing the total quality score of all work orders after equipment allocation. It balances the current load and average load to reduce the risk of equipment overload failure, while ensuring that production quality is not sacrificed while balancing the load.

[0034] Preferably, the process of adjusting the allocation scheme according to the current load of the equipment is as follows:

[0035] Based on each matching pair in the allocation scheme, each matching pair contains a work order and the equipment matched by that work order. The current load of the matched equipment is calculated. If the current load does not exceed the maximum load of the equipment, the matching is performed according to the matching pair. If the current load exceeds the maximum load of the equipment, the production quality score of the matching pair is penalized, and the matching of the work order in the matching pair is canceled. If the current load of all equipment exceeds the maximum load, the matching pair with the highest production quality score in the allocation scheme is selected for matching.

[0036] Secondly, an Internet of Things (IoT)-based inkjet printing work order scheduling system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the IoT-based inkjet printing work order scheduling method described in any one of the above embodiments is implemented.

[0037] The present invention has the following effects:

[0038] 1. This invention innovatively designs a complex quantification model for printed images, transforming the printed images on the work order into edge density data, curvature change data, and color contrast data, accurately capturing the technical requirements of the printed images for the equipment. This quantification method avoids the bias of traditional manual experience judgment, ensuring that complex prints are assigned to appropriate equipment, thereby significantly improving the product quality compliance rate.

[0039] 2. Based on actual printed images and standard printed images collected by the Internet of Things, the actual production quality score of the work order on a specific device is quantified through image registration, edge matching and color fidelity calculation. At the same time, a predictive model is built using neural networks, and patterns are mined from historical data to replace the traditional device evaluation method that relies on physical parameters, thereby further improving the stability and consistency of production quality.

[0040] 3. This invention introduces a dynamic penalty mechanism through an improved multi-round KM algorithm. While maximizing the total quality score, it reduces the risk of equipment overload failure and extends the service life of the equipment through overload penalty and post-processing balancing adjustment, thus achieving dual optimization of quality and efficiency.

[0041] 4. This invention constructs an intelligent scheduling system for inkjet printing work orders based on the Internet of Things. Through a four-step closed-loop logic of "quantification of printing complexity - equipment quality evaluation - quality prediction modeling - dynamic scheduling optimization", it realizes accurate matching and load balancing between work orders and equipment. The entire scheduling process is automated and intelligent, reducing manual intervention and improving scheduling efficiency. Attached Figure Description

[0042] Figure 1 This is a flowchart of steps S1-S5 in the Internet of Things-based inkjet printing work order scheduling method of the present invention.

[0043] Figure 2 This is a flowchart of steps S20-S22 in the inkjet printing work order scheduling method based on the Internet of Things in this embodiment of the invention. Detailed Implementation

[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0045] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0046] Reference Figure 1 The IoT-based inkjet printing work order scheduling method includes steps S1-S5, as follows:

[0047] S1. Obtain the complexity multidimensional data of standard printed images from historical production work orders of the equipment.

[0048] Complexity multidimensional data includes edge density data, curvature variation data, and color contrast data. By decomposing the key visual features of the printed image on the work order, the complexity of the printed image is transformed into a three-dimensional vector of edge density, curvature variation, and color contrast, which can accurately capture the technical requirements of the printed image for the equipment.

[0049] The process for obtaining the edge density data of the standard printed image on the work order is as follows:

[0050] The Canny algorithm is used to detect edges in a standard printed image, resulting in an edge image (edge ​​points are represented as 1, non-edge points as 0). The number of edge pixels and the total number of pixels in the edge image are counted. The ratio of edge pixels to total pixels is calculated to obtain edge density data. Specifically, the formula for calculating edge density data is as follows:

[0051]

[0052] In the formula, Represents edge density data. Indicates the number of pixels at the edge. This represents the total number of pixels. Edge density data reflects the complexity of the printed image outline. The denser the edges, the higher the edge density data, the more outline points the equipment needs to locate, and the higher the positioning accuracy requirement for the equipment.

[0053] The process of obtaining the curvature variation data of the standard printed image on the work order is as follows:

[0054] An 8-connected component tracing algorithm is used to extract continuous edge contours from the edge image, resulting in multiple continuous edge contours. Each continuous edge contour is a set of pixels that constitute that continuous edge contour. For each set of pixels in a continuous edge contour, the curvature of each pixel is calculated using the central difference method, which utilizes the left and right adjacent points of each pixel. The ratio of the number of points in a single continuous edge contour to the number of edge pixels is calculated to obtain the weight of each continuous edge contour. The sum of the products of each pixel in a single continuous edge contour with the weight of that continuous edge contour is calculated, and the ratio of this sum of products to the sum of the weights of all pixels in that continuous edge contour is calculated to obtain the weighted average curvature of the single continuous edge contour. The sum of the squares of the differences between each pixel and its corresponding weighted average curvature in a single continuous edge contour and the corresponding weights is calculated. After summing the above products for all continuous edge contours, the ratio of this sum to the sum of the weights of all continuous edge contours is calculated to obtain the curvature change data. Specifically, the calculation of the weighted average curvature satisfies the following formula:

[0055]

[0056] In the formula, Indicates the first The weighted average curvature of a continuous edge profile Indicates the first The length of the nth consecutive edge contour, i.e. the nth The number of pixels contained in a continuous edge contour. Indicates the first 1 pixel Indicates the first The weights of continuous edge contours, , Indicates the number of pixels at the edge. Indicates the first The curvature of each pixel. The weighted average curvature can be used as a shape feature of each continuous edge contour, reflecting the overall curvature of the continuous edge contour.

[0057] The calculation of curvature variation data satisfies the following formula:

[0058]

[0059] In the formula, This represents the data on curvature variation, specifically the weighted standard deviation (curvature variance). This represents the total number of continuous edge contours. Indicates the first The length of the nth consecutive edge contour, i.e. the nth The number of pixels contained in a continuous edge contour. Indicates the first The weights of continuous edge contours, Indicates the first Curvature of each pixel Indicates the first The weighted average curvature of a continuous edge profile.

[0060] Curvature change data reflects the degree of curvature change at the edge of the printed image. The larger the curvature change data, the more drastic the curvature change at the edge of the printed image, requiring the equipment to quickly adjust the direction, speed, and ink droplet size in a short period of time. The smaller the curvature change data, the smoother the edge of the printed image, and the equipment only needs to output stably without frequent adjustments.

[0061] The process for obtaining the color contrast data of the standard printed image on the work order is as follows:

[0062] In the edge image, for each edge pixel, take the pixels on either side of its normal direction (or approximately use adjacent pixels); convert the RGB color to the LAB color space, calculate the color difference between the pixels on either side of all edge pixels and take the average to obtain the color contrast data, denoted as . The process of calculating the color difference between the pixels on both sides is existing technology and will not be described in detail here.

[0063] Color contrast data indicates the degree of color difference between the two sides of an edge pixel. The greater the difference, the stronger the visual contrast of the printed image, and the higher the requirements for ink droplet concentration control of the equipment.

[0064] Through the above three steps, the complexity vector of the standard printed image is obtained. The technical requirements of equipment for mass production.

[0065] S2. Perform feature matching on the standard printing image and the actual printing image produced by a single machine. Based on the complexity of the multidimensional data, perform a weighted summation of the feature matching results to obtain the total work order score of the single machine on the corresponding historical work order.

[0066] For the actual production capacity of each piece of equipment, traditional evaluation methods based on equipment physical data are not only difficult to collect data, but also difficult to construct an evaluation model that conforms to reality. This invention proposes a work order quality score calculation method based on matching the actual production results of the equipment with a standard template, which is used to evaluate the actual production capacity of the current equipment.

[0067] like Figure 2 As shown, the process of feature matching between the standard printed image and the actual printed image produced by a single device includes steps S20-S22:

[0068] S20. Register the actual printed image with the standard printed image to obtain the registered image.

[0069] To eliminate interference from image displacement and rotation caused by equipment positioning errors, the actual printed image and the standard printed image must first be registered to ensure pixel-level matching accuracy. The specific process is as follows: Key feature points (such as edge corners and texture abrupt changes) are extracted from the standard and actual printed images using the SIFT algorithm; the Euclidean distance between feature points is calculated using the FLANN matcher, and the top 100 pairs of feature points with the highest matching degree are selected; the coordinates of the matching feature points are extracted, and the transformation matrix (affine transformation) from the actual printed image to the standard printed image is calculated using the RANSAC method; the calculated transformation matrix is ​​used to perform rotation and translation correction on the actual printed image to obtain the aligned registered image, ensuring that the pixel coordinates of the aligned registered image correspond one-to-one with the standard printed image, with an error not exceeding 1 pixel.

[0070] S21. Perform edge feature matching on the standard printed image and the registered image. The specific matching process is as follows:

[0071] Extract the edge image from the standard printing image. For each edge pixel in the edge image, take a W×W block centered on that pixel as the standard block. For example, a 5×5 block can be taken as the standard block, and a block of the same size can be taken at the same position in the registered image as the registration block. ;Normalized cross-correlation (NCC) is used to calculate each standard block. and corresponding registration block The similarity is calculated by obtaining the NCC value of each edge pixel, averaging the NCC values ​​of all edge pixels to obtain the edge matching similarity; the average of the NCC values ​​is then mapped to a score range of 0 to 100 to obtain the edge accuracy score. .

[0072] Edge accuracy scores are used to evaluate the accuracy of pattern edges in actual printed images, i.e., their consistency with standard printed images. Edge accuracy directly reflects the positioning accuracy of the equipment; the more accurate the edges, the more accurate the positioning.

[0073] S22. Perform color feature matching between the standard printed image and the actual printed image. The specific matching process is as follows:

[0074] The standard printed image and the actual printed image are converted to the LAB color space. The Sobel operator is used to calculate the gradient magnitude of each pixel in the standard printed image, resulting in a gradient magnitude map. Regions with large gradients represent large color changes, and these regions require high color accuracy. The gradient magnitude map is normalized to the range [0, 1] and used as a weight map. For each pixel, the color difference between the actual printed image and the standard printed image is calculated. The color difference of each pixel is multiplied by its corresponding weight. All products are summed and then divided by the sum of the weights of all pixels to obtain the weighted average color difference. The weighted average color difference is mapped to a score range of 0 to 100 to obtain the color fidelity score. .

[0075] By converting the printed image to the LAB color space and introducing the Sobel gradient magnitude as a weight, high-sensitivity detection of color difference in the "human eye sensitive area" of the printed image is achieved. The color difference in the area with a large gradient (i.e. the area with large color change) has a greater impact on the score and can better reflect the color fidelity.

[0076] The process of obtaining the total work order score for a single device on the corresponding historical work order is as follows: The ratio of the sum of edge density data and curvature variation data to the sum of edge density data, curvature variation data, and color contrast data is used as the weight; the sum of the product of 1 minus the weight and the color fidelity score, and the product of the edge accuracy score and the weight, is calculated to obtain the total work order score. Specifically, the calculation of the total work order score satisfies the following formula:

[0077]

[0078] In the formula, This indicates the total score for the work order. Represents weight, , Represents edge density data. Represents data on curvature variation. Represents color contrast data. This indicates the edge accuracy score. This indicates the color fidelity score.

[0079] By calculating weights in real time using multidimensional data on the complexity of standard printed images (edge ​​density data, curvature variation data, and color contrast data), the overall score can adaptively switch between "edge clarity" and "color fidelity": prints with high edge density and curvature variation automatically increase edge weights, and prints with high color contrast automatically increase color weights, thereby ensuring that the print can be accurately evaluated in the most critical dimensions.

[0080] S3. Construct and train a neural network model, in which the complex multidimensional data and the corresponding equipment are used as inputs, the total score of the work order is used as the label, and the predicted value of the production quality score of each equipment work order is output.

[0081] Construct a neural network model, which uses multidimensional data on the complexity of historical production work orders of the equipment. The system takes the corresponding device ID as input and the total work order score as the label, outputting the predicted production quality score for each device's work order. A neural network model is trained using historical production work order data from the devices. The loss function is the mean squared error, used to calculate the difference between the pre-set production quality score and the actual total work order score. The training process of the neural network model includes multiple training rounds, with each round updating the weights through backpropagation to minimize the loss function.

[0082] S4. For a single new work order, calculate its complexity multidimensional data, input the complexity multidimensional data of the new work order and all equipment into the trained neural network model, and output the predicted values ​​of all production quality scores.

[0083] For a single new work order, calculate its complexity using multidimensional data. Multidimensional data on the complexity of new work orders The system inputs all device IDs into a trained neural network model, outputting a predicted production quality score for a single new work order across all devices. This yields the predicted production quality scores for all new work orders on each device. A score matrix M is then constructed from all new work orders, devices, and their corresponding predicted production quality scores. ), Indicates the number of new work orders. Indicates the number of devices. Indicates work order In the equipment The predicted value of the production quality score.

[0084] S5. Based on all new work orders, equipment, and the predicted values ​​of their corresponding production quality scores, the KM algorithm is used for matching to obtain the optimal equipment for each work order.

[0085] The purpose of this step is to allocate work orders based on the predicted values ​​of production quality scores. Generally, the KM algorithm can be used to allocate work orders. However, because the traditional KM algorithm can only perform one-to-one allocation, in real-world scenarios, the number of work orders is often greater than the number of devices, and one device can handle multiple work orders.

[0086] Therefore, this invention proposes an improved KM algorithm that can handle situations where the number of work orders exceeds the number of devices (i.e., one device is assigned multiple work orders), while also considering load balancing. The core idea is to introduce a load penalty mechanism within the framework of the KM algorithm. When a device is assigned work orders exceeding its load capacity, the revenue subsequently allocated to that device is reduced, thereby maximizing the total matching score while achieving load balancing.

[0087] Initialization is performed before matching. All new work orders are listed and sorted in descending order of complexity to form a work order list. To reduce computational load, edge density data is used. Represents the complexity of the work order; lists all devices, each with an initial load of 0; sets a penalty factor. Punishment factor The initial value is 0.1, and it increases with each processing round; a round counter is set, and the number of rounds... Initial value is 0, maximum number of rounds To prevent an infinite loop.

[0088] The top label is initialized based on the classic KM algorithm. For the left-hand work order node, the work order top label is: For each work order Top mark Initialize to the maximum production quality score achievable on all devices for this work order; for the device nodes on the right, the device top label is... For each device Top mark Initialize to 0 for each edge The weight is Work order In the equipment The dynamic value of the production quality score is initialized to the work order. In the equipment Predicted values ​​of production quality scores .

[0089] The process of using the KM algorithm to perform preliminary matching of new work orders and equipment is as follows:

[0090] The number of work orders in the list is not empty, and the number of rounds is... The maximum number of rounds has not been reached. At that time, the number of rounds Increment by 1, and repeat the following steps:

[0091] The work orders in the work order list are selected in descending order of complexity, and the number of work orders that are the same as the number of devices are selected for matching. If the number of work orders in the work order list is less than the number of devices, all of them are selected (in the classic KM algorithm, it is usually required that the number of nodes on the left and right sides are equal, so the number of work orders is selected based on the number of devices in each round of matching).

[0092] Construct a corresponding temporary score matrix based on the selected work orders and equipment. The size of the matrix is ​​the number of work orders selected in the current round. Number of devices, temporary score matrix It records the dynamic values ​​of the production quality score of the selected work order in the current round on each device;

[0093] Based on the temporary score matrix The standard KM algorithm is run to obtain the preliminary matching results (i.e., the matching pairs formed by all work orders and specific equipment in the current round of selected work orders); the preliminary matching results are adjusted according to the current load of the equipment to obtain the adjustment plan for each work order to be assigned to the equipment.

[0094] The process of adjusting the initial matching results based on the current load of the equipment is as follows:

[0095] For each matching pair in the initial matching results , This indicates a specific work order selected in the current round. Represents the device, calculates the current load of the device after allocation, and the current load of the device after allocation is equal to the sum of the complexities of all work orders allocated to that device;

[0096] When the current load does not exceed the maximum load of the device (the maximum load is calculated from the complexity of the device's historical production work orders), the matching pair is used. Assign (indicating a successful match), update the device load, and remove the successfully matched work order from the work order list;

[0097] When the current load exceeds the device's maximum load, the product of the difference between the current load and the maximum load, the penalty factor, and the current round number is used as the penalty value. The difference between 1 and the penalty value is used as the weight to multiply the matching pair. The dynamic value of the production quality score is used to lower the production quality score of the matched pair and thus reduce the production quality score of the matched pair. The work orders in the list remain in the work order list (meaning that the match was unsuccessful) and participate in the subsequent rounds of allocation;

[0098] When all devices are currently overloaded, for any new work order selected in the current round, choose the matching pair with the highest production quality score. Distribute the work orders (because although all devices are overloaded, “overload” is a relative concept; not all devices are completely unable to continue working, and some devices may only be slightly overloaded and can still handle a small amount of extra tasks), update the device load, remove successfully matched work orders from the work order list, and issue an overload warning.

[0099] In the number of rounds The maximum number of rounds has been reached. However, if the work order list is not empty, for each remaining work order in the work order list, select the device with the lowest current load for allocation, thereby obtaining the adjustment plan for all work orders to be allocated to the device.

[0100] To facilitate understanding, an example is provided below. Suppose there are 100 new work orders and 10 devices. The new work orders are sorted in descending order of complexity (edge ​​density data) to form a work order list. In the first round, starting with the work orders at the beginning of the list, 10 work orders are selected for matching. These 10 work orders, 10 devices, and the production quality scores of the 10 work orders on the 10 devices constitute a temporary score matrix. Based on the production quality scores, the KM algorithm is used to match the 10 work orders and 10 devices, resulting in 10 matching pairs. Calculate the current load of 10 devices. For devices whose current load does not exceed the maximum load, the matching pair between the device and the corresponding work order is considered successful. The load of the device is updated, and the successfully matched work order is removed from the work order list. For devices whose current load exceeds the maximum load, the matching pair between the device and the corresponding work order is considered unsuccessful. The production quality score of the matching pair is penalized, and the unsuccessful work order is put back into the work order list to wait for the next round of allocation. If the current load of all devices exceeds the maximum load, for each work order, calculate the production quality score of the work order on the 10 devices, select the device with the highest production quality score to form a matching pair with the work order for allocation, and then proceed to the second round of work order selection and allocation until all work orders are allocated to the corresponding devices.

[0101] The logic for adjusting the allocation scheme based on the current load of the equipment is as follows: when the current load of an allocated piece of equipment exceeds its maximum load, a penalty factor (positively correlated with the round and overload) is used to reduce the production quality score of the matching pair, decreasing the probability of the pair being matched again in subsequent rounds and guiding work orders to equipment with lower loads. Each time, work orders are selected from the remaining work orders in descending order of complexity, matching the number of equipment. A standard KM algorithm is used for initial matching, prioritizing the allocation of high-complexity work orders to more suitable equipment. The purpose of multiple iterations is to gradually process the work orders, ensuring that each piece of equipment can handle as many work orders as possible within its load capacity, avoiding load imbalance caused by a single allocation.

[0102] After obtaining the adjustment plan, the load balancing adjustment is performed again to obtain the optimal allocation of each work order to the equipment. The adjustment process is as follows:

[0103] Calculate the current average load of all devices;

[0104] Sort all devices in descending order of their current load, from high load to low load;

[0105] For devices with a current load higher than the average load, sort the work orders on that device in ascending order of complexity (edge ​​density data), that is, sort them from low complexity to high complexity.

[0106] For each work order on the device in the sorted order, iterate through other devices whose current load is less than the average load, compare the predicted production quality score of the work order on the original device with the predicted production quality score of the work order when it is moved to another device (the device whose current load is less than the average load). If the predicted production quality score does not decrease, move the work order to another device, update the allocation relationship and the load of the two devices, and return to the above step of "sorting all devices in descending order of current load" to repeat the loop.

[0107] If no work orders are moved on any of the devices currently under load above the average load, the loop stops, and the final allocation scheme, i.e., the optimal allocation of each work order to a device, is output.

[0108] Based on the average load benchmark, low-complexity work orders on high-load equipment (equipment with current load higher than average load) are migrated to low-load equipment (equipment with current load lower than average load). Only work orders that do not reduce production quality scores are migrated. Without sacrificing overall quality, the equipment load is further balanced, ultimately achieving the dual goals of optimal quality and load balancing.

[0109] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the Internet of Things-based inkjet printing work order scheduling method according to the first aspect of the present invention.

[0110] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0111] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An inkjet printing work order scheduling method based on the Internet of Things, characterized in that, The method comprises the following steps: obtaining complexity multi-dimensional data of a standard printing image on a historical production work order of a device, the complexity multi-dimensional data comprising edge density data, curvature variation data and color contrast data; performing feature matching on the standard printing image and an actual printing image produced by a single device, and performing weighted summation on the result of the feature matching according to the complexity multi-dimensional data to obtain a work order total score of the single device on the corresponding historical work order; constructing and training a neural network model, wherein the complexity multi-dimensional data and the corresponding device are taken as inputs, the work order total score is taken as a label, and a predicted value of a work order production quality score of each device is output; for a single new work order, complexity multi-dimensional data thereof is calculated, the complexity multi-dimensional data of the new work order and all devices are input into the trained neural network model, and predicted values of production quality scores of all devices are output; based on all new work orders, devices and corresponding predicted values of production quality scores, performing matching by using a KM algorithm to obtain an optimal selection of assignment of each work order to a device. 2.The IoT-based job order production method for inkjet printing according to claim 1, wherein, The edge density data of the standard printing image on the work order is obtained by the following process: using a Canny algorithm to detect edges of the standard printing image to obtain an edge image; counting the number of edge pixels and the total number of pixels in the edge image; and calculating the ratio of the number of edge pixels to the total number of pixels to obtain the edge density data. 3.The IoT-based job order production method for inkjet printing according to claim 2, wherein, The curvature variation data of the standard printing image on the work order is obtained by the following process: using an 8-connected domain tracking algorithm to extract continuous edge contours from the edge image to obtain a plurality of continuous edge contours, each continuous edge contour being a point set comprising pixel points constituting the continuous edge contour; for the point set of each continuous edge contour, using a central difference method to calculate the curvature of each pixel point, wherein the central difference method uses left and right adjacent points of each pixel point for calculation; calculating the ratio of the number of points of a single continuous edge contour to the number of edge pixels to obtain the weight of each continuous edge contour; and calculating a weighted average curvature of each continuous edge contour according to the curvature of each pixel point and the weight of the continuous edge contour in which each pixel point is located. The curvature variation data is obtained by calculating the curvature variance according to the curvature of each pixel point and the weighted average curvature. 4.The IoT-based job order production method for inkjet printing according to claim 2, wherein, The color contrast data of the standard printing image on the work order is obtained by the following process: on the edge image, for each edge pixel point, taking two pixel points on the normal direction thereof; converting the RGB color into the LAB color space, calculating the color difference of the two pixel points on both sides of each edge pixel point and taking the average value to obtain the color contrast data. 5.The IoT-based job order production method for inkjet printing according to claim 1, wherein, The feature matching on the standard printing image and the actual printing image produced by the single device is performed by the following process: aligning the actual printing image and the standard printing image to obtain an aligned image; performing edge feature matching on the standard printing image and the aligned image, taking a standard block centered at each edge pixel point in the edge image of the standard printing image, taking a registration block of the same size at the same position in the aligned image, calculating the similarity of the standard block and the corresponding registration block, and taking the average value of the similarity of all edge pixel points to obtain an edge accuracy score; The standard printed image and the actual printed image are converted to the LAB color space, the color features of the standard printed image and the actual printed image are matched, for each pixel point, the color difference between the actual printed image and the standard printed image is calculated, the gradient amplitude of each pixel point of the standard printed image is calculated, the gradient amplitude is normalized to obtain a weight, the weighted average of the color difference of all pixel points is calculated to obtain a color fidelity score. 6.The IoT-based job order production method for inkjet printing according to claim 5, wherein, The total score of the work order on the corresponding historical work order of the single device is obtained as follows: The ratio of the sum of the edge density data and the curvature change data to the sum of the edge density data, the curvature change data and the color contrast data is taken as a weight; the product of the difference between 1 minus the weight and the color fidelity score and the sum of the edge accuracy score and the weight is calculated to obtain a total work order score. 7.The IoT-based job order production method for inkjet printing according to claim 1, wherein, The matching by the KM algorithm to obtain the optimal selection of each work order to the device includes the following steps: The KM algorithm is used to preliminarily match the new work order and the device, and the preliminary matching result is adjusted according to the current load of the device to obtain an adjustment scheme of the distribution of each work order to the device; After obtaining the adjustment scheme, the current average load of all devices is taken as a reference, if the device to which the new work order is distributed is a high-load device, the work order is moved to a low-load device for production; wherein, if the predicted value of the production quality score calculated after the work order is moved to the low-load device is less than the predicted value of the production quality score calculated when the work order is originally distributed to the high-load device, the moving operation is not performed. 8.The IoT-based job order production method for inkjet printing according to claim 7, wherein, The adjustment of the distribution scheme according to the current load of the device is as follows: Based on each matching pair in the distribution scheme, each matching pair contains a work order and the device to which the work order is matched, the current load of the matched device is calculated, when the current load does not exceed the maximum load of the device, the matching pair is matched; when the current load exceeds the maximum load of the device, the production quality score of the matching pair is penalized, and the matching of the work order in the matching pair is cancelled; when the current load of all devices exceeds the maximum load, the matching pair with the highest production quality score in the distribution scheme is selected for matching.

9. An inkjet printing work order production system based on the Internet of Things, characterized in that, It comprises: A processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the method for printing work order production based on Internet of Things is realized according to any one of claims 1-8.

Images