A workshop order allocation method and system based on multi-feature evaluation, an electronic device, and a storage medium

Abstract

The application provides a workshop order distribution method and system based on multi-feature evaluation, an electronic device and a storage medium, relates to the technical field of production distribution, and comprises the following steps: collecting original data of all workshop orders and calculating corresponding characteristic values; quantitatively processing the characteristic values of the workshop orders; obtaining the ranking positions and delay probability values of the workshop orders and determining corresponding early warning levels; based on a preset workshop resource configuration information table, distributing the workshop orders to corresponding employees to obtain exclusive to-do lists of the employees; determining the push channel and the push frequency corresponding to the early warning level of each workshop order based on the early warning level; and pushing the exclusive to-do list to the corresponding employee through the push channel and the push frequency. The application solves the technical problem that the traditional manual dispatching mode is difficult to dynamically determine the optimal processing sequence by comprehensively considering multiple factors, leading to a high order delay rate, and has the effect of improving the timeliness of workshop order delivery.

Description

Technical Field

[0001] This invention relates to the field of production allocation technology, and in particular to a workshop order allocation method, system, electronic device and storage medium based on multi-feature evaluation. Background Technology

[0002] As the manufacturing industry transforms towards intelligent manufacturing, workshop production allocation, as a core link connecting the planning and execution layers, directly impacts order delivery capabilities due to the quality of its decision-making. In discrete manufacturing scenarios such as machining, workshops need to process dozens of work-in-process orders at different stages of processing daily, each involving multi-dimensional constraints such as delivery dates, equipment, materials, and personnel. In recent years, systems such as ERP, MES, and WMS have been widely adopted in manufacturing enterprises, providing a rich data foundation for production management. However, how to transform this data into precise allocation instructions remains a bottleneck restricting the intelligent upgrading of enterprises.

[0003] In current engineering practice, most machining workshops primarily employ two scheduling methods: one is for planners to manually schedule based on experience and verbally assign processing sequences during morning meetings; the other is to use a single fixed field such as customer level or order amount in the ERP system for sorting. However, static priority fields cannot respond to dynamic changes on-site, such as equipment failures and material delays, leading to a disconnect between scheduling plans and actual production. Furthermore, the lack of a mechanism to comprehensively prioritize multiple work-in-process orders under the same quantitative scale means that when multiple urgent orders coexist, planners often rely solely on intuition to determine the order, resulting in chaotic scheduling and frequent misjudgments. These problems ultimately lead to persistently high order delay rates in the workshop. Summary of the Invention

[0004] In view of the above-mentioned shortcomings in the existing technology, the purpose of this invention is to provide a workshop order allocation method based on multi-feature evaluation, which has the characteristic of improving the timeliness of workshop order delivery.

[0005] The above-mentioned objective of this invention is achieved through the following technical solution:

[0006] A workshop order allocation method based on multi-feature evaluation includes:

[0007] Collect raw data of all workshop orders and calculate the corresponding feature values ​​based on the raw data;

[0008] The feature values ​​of each workshop order are quantified to obtain the ranking position and delay probability value of each workshop order;

[0009] Based on the ranking position and delay probability value of each workshop order, the corresponding warning level is determined;

[0010] Based on the preset workshop resource configuration information table, each workshop order is assigned to the corresponding employee, and a unique to-do list is generated for each employee.

[0011] Based on the warning level of each workshop order, determine the push channel and push frequency corresponding to the warning level, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

[0012] By adopting the above technical solution, firstly, multi-dimensional feature extraction and comprehensive quantitative evaluation are performed on all orders in production to generate a unique global ranking position and delay probability for each order, solving the problem of lack of a unified comparison scale in manual scheduling. Then, based on the ranking results and warning levels, orders are accurately assigned to corresponding employees to form a personal to-do list, avoiding the ambiguity of responsibilities caused by mass notifications. Finally, the personalized to-do list is pushed out by matching differentiated push channels and frequencies according to the warning level, effectively avoiding information omissions and improving the on-time delivery rate of orders.

[0013] Preferably, the specific method for calculating the corresponding feature value based on the original data is as follows:

[0014] The raw data is parsed to obtain the promised delivery date, standard working hours, reported working hours, current inventory quantity, and required quantity for each workshop order;

[0015] Obtain the current time, calculate the time urgency based on the promised delivery date, the current time, and the standard working hours, and obtain the time pressure index;

[0016] The processing progress is calculated based on the reported working hours and the standard working hours to obtain the progress completion rate.

[0017] By comparing the current inventory quantity with the quantity required by the order, the material shortage index is obtained;

[0018] The time pressure index, the progress completion rate, and the material shortage index are used as characteristic values ​​corresponding to each workshop order.

[0019] By adopting the above technical solution, raw data is automatically extracted from multi-source systems and uniformly transformed into three standardized feature values: time pressure index, progress completion rate, and material tension index. This maps heterogeneous information such as delivery date, progress, and materials onto the same quantitative scale, avoiding deviations caused by manual input and subjective judgment.

[0020] Preferably, the step of quantifying the feature values ​​of each workshop order to obtain the ranking position and delay probability value of each workshop order includes:

[0021] Based on the feature values ​​corresponding to each workshop order, an order feature set is constructed;

[0022] The order feature set is input into a preset score prediction model to obtain the emergency score corresponding to each workshop order;

[0023] Based on the urgency score of each workshop order, all workshop orders are sorted in descending order to construct a global priority sequence, and the ranking position of each workshop order in the global priority sequence is determined.

[0024] The urgency score of each workshop order is converted based on a preset mapping function to obtain the delay probability value.

[0025] By adopting the above technical solution, a pre-trained score prediction model is used to perform unified reasoning on the feature set of all orders in production, outputting an emergency score that can be compared horizontally for each order, thereby generating a global priority sequence containing a unique ranking position, and simultaneously converting it into an intuitive delay probability value, thus comprehensively quantifying multi-dimensional features into a unified ranking basis.

[0026] Preferably, the training method for the preset score prediction model is as follows:

[0027] Obtain the historical feature set of historical workshop orders, along with the corresponding actual delay tags and actual priority tags;

[0028] Construct an initial score prediction network, input the historical feature set into the initial score prediction network, and obtain the emergency score corresponding to each historical workshop order;

[0029] Based on the emergency score, the actual priority label, and the actual delay label, a total loss function is constructed that includes a ranking loss term and a delay prediction loss term, wherein the ranking loss term is used to constrain the ranking of the predicted emergency score to be consistent with the ranking of the actual priority label;

[0030] Based on the total loss function, the parameters of the initial score prediction network are iteratively updated using the backpropagation algorithm until the total loss function converges, thus obtaining the preset score prediction model.

[0031] By adopting the above technical solution, the actual processing order and delay results of orders in historical data are used as supervision signals to construct a joint loss function that includes both sorting loss and delay prediction loss for model training. This enables the model to not only learn the relative urgency of orders, but also to simultaneously perceive the risk of delay.

[0032] Preferably, determining the corresponding early warning level based on the ranking position and delay probability value of each workshop order includes:

[0033] Based on the preset ranking threshold range and the preset delay probability threshold range, the basic early warning level of each workshop order is determined;

[0034] Obtain the order attribute data of each workshop order, and correct the basic warning level based on the order attribute data to obtain the final warning level corresponding to each workshop order.

[0035] By adopting the above technical solution, the basic warning level is first determined based on a combination of ranking position and delay probability. Compared with warning based on a single dimension, the judgment result is more comprehensive and accurate. Then, order attribute data is introduced to differentiate the basic level, so that special factors such as customer level and material shortage can be reasonably reflected in the warning. This avoids misjudgment or omission of special orders by general rules, and effectively improves the accuracy of warning classification and scenario adaptability.

[0036] Preferably, based on a preset workshop resource allocation information table, each workshop order is assigned to a corresponding employee, generating a dedicated to-do list for each employee, including:

[0037] Obtain the workshop resource configuration information table, which records the association between the current process and the corresponding operator for each workshop order;

[0038] Based on the workshop resource allocation information table, find the employee corresponding to the current process in each workshop order, and assign each workshop order to the found employee.

[0039] For each employee, all workshop orders assigned to that employee are sorted according to their ranking position, generating a personalized to-do list for that employee.

[0040] By adopting the above technical solution, the workshop resource allocation information table is used to accurately match orders in the global sorting sequence to the corresponding operators according to the current process. This ensures that each order can be accurately assigned to the employee actually responsible for that process. Furthermore, the orders under each employee's name are re-sorted according to the global ranking to generate a personalized to-do list. This allows each operator to clearly know the order of all their pending orders without having to manually sift through factory-wide mass notifications. This effectively avoids information overload and ambiguity of responsibilities, and improves on-site execution efficiency.

[0041] Preferably, after determining the push channel and push frequency corresponding to the warning level based on the warning level of each workshop order, and pushing the exclusive to-do list to the corresponding employee based on the push channel and push frequency, the process includes:

[0042] The corresponding confirmation time limit is determined based on the aforementioned warning level;

[0043] If a confirmation instruction for the personalized to-do list is received from the employee within the confirmation time limit, the repeated pushing of the personalized to-do list to the employee will be stopped.

[0044] If no confirmation instruction is received from the employee within the specified confirmation period, an escalation warning mechanism will be triggered, and the exclusive to-do list will be pushed to the employee's corresponding supervisor.

[0045] By adopting the above technical solution, different confirmation time limits are set for different warning levels, linking timeliness control with risk level. Once confirmed, repeated pushes are stopped immediately to avoid excessive interference with operators. If confirmation is not made within the time limit, a hierarchical reporting mechanism is automatically triggered. This effectively solves the defects of traditional verbal or mass notification methods, such as easy omission of information and difficulty in tracing responsibility, and ensures that urgent orders are responded to and processed in a timely manner.

[0046] The second objective of this invention is to provide a workshop order allocation system based on multi-feature evaluation, which has the characteristic of improving the timeliness of workshop order delivery.

[0047] The second objective of this invention is achieved through the following technical solution:

[0048] A workshop order allocation system based on multi-feature evaluation includes: a data acquisition module, used to collect raw data of all workshop orders and calculate corresponding feature values ​​based on the raw data;

[0049] The quantization processing module is used to quantify the feature values ​​of each workshop order to obtain the ranking position and delay probability value of each workshop order.

[0050] The level determination module is used to determine the corresponding warning level based on the ranking position and delay probability value of each workshop order;

[0051] The task allocation module is used to allocate each workshop order to the corresponding employee based on a preset workshop resource configuration information table, and generate an exclusive to-do list for each employee.

[0052] The list push module is used to determine the push channel and push frequency corresponding to the warning level of each workshop order, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

[0053] By adopting the above technical solutions, the data acquisition module, quantitative processing module, level determination module, task allocation module, and list push module complement each other, effectively avoiding the problems of broken information transmission, ambiguous division of responsibilities, and delayed allocation response in traditional manual scheduling, and realizing the automation, standardization, and precision of workshop production allocation.

[0054] The third objective of this invention is to provide an electronic device that improves the timeliness of order delivery in the workshop.

[0055] The above-mentioned objective three of this invention is achieved through the following technical solution:

[0056] An electronic device includes a memory and a processor, wherein the memory stores a computer program capable of being loaded by the processor and executing the workshop order allocation method based on multi-feature evaluation as described above.

[0057] The fourth objective of this invention is to provide a computer-readable storage medium capable of storing corresponding programs, which facilitates the improvement of on-time delivery of workshop orders.

[0058] The fourth objective of this invention is achieved through the following technical solution:

[0059] A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the workshop order allocation method based on multi-feature evaluation as described above.

[0060] In summary, the present invention has at least one of the following beneficial technical effects:

[0061] 1. This invention solves the problems of lack of objective comparison criteria and ambiguous responsibilities in manual scheduling by uniformly quantifying and sorting all in-process orders and pushing them to the corresponding employees in a differentiated manner based on the warning level, thereby effectively improving the on-time delivery rate of orders;

[0062] 2. This invention determines the basic warning level by jointly using ranking position and delay probability as two dimensions, and then introduces order attributes for differentiated correction, so that special factors such as customer level and material shortage can be reasonably reflected, avoiding misjudgment or omission, and improving the accuracy of warning classification and scenario adaptability. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the steps of a workshop order allocation method based on multi-feature evaluation provided in Embodiment 1 of the present invention.

[0064] Figure 2 This is a structural block diagram of a workshop order allocation system based on multi-feature evaluation provided in Embodiment 2 of the present invention.

[0065] Figure 3 This is a schematic block diagram of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0066] This invention provides a workshop order allocation method, system, electronic device, and storage medium based on multi-feature evaluation. It addresses the technical problem of high order delay rates caused by the difficulty of dynamically determining the optimal processing sequence based on multiple factors in traditional manual dispatching methods. This results in improved on-time delivery of workshop orders.

[0067] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0068] It should be noted that in this embodiment of the invention, all content involving object data must be obtained with the object's authorization and consent, and must comply with current laws and standards. If the embodiment involves personal information, it must ensure that the individual's consent has been obtained; if it involves sensitive information, the separate consent of the information subject must be obtained. The implementation of the entire embodiment should also be based on the object's authorization and consent.

[0069] It should be noted that the terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this disclosure.

[0070] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0071] Example 1:

[0072] Please see Figure 1 The present invention provides a workshop order allocation method based on multi-feature evaluation, comprising:

[0073] Step 101: Collect the raw data of all workshop orders and calculate the corresponding feature values ​​based on the raw data.

[0074] Preferably, step 101 may include the following sub-steps:

[0075] S11. Parse the raw data to obtain the promised delivery date, standard working hours, reported working hours, current inventory quantity, and required quantity for each workshop order.

[0076] Raw data refers to basic production site data directly obtained from multiple heterogeneous business systems such as ERP and MES through the data acquisition module.

[0077] Workshop orders refer to work-in-process orders.

[0078] The promised delivery date refers to the latest delivery date agreed upon between the order and the customer.

[0079] Standard working hours refer to the rated operating time required to complete all processing content of an order under normal technological conditions.

[0080] Reported working hours refer to the total processing hours that the operator has reported for the order in the MES system.

[0081] Current inventory quantity refers to the real-time available stock of the key materials required for this order in the warehouse.

[0082] The quantity required for an order refers to the total rated consumption of the material required to complete the production of that order.

[0083] Specifically, data is collected and parsed from different data sources at a preset collection frequency (e.g., 30 seconds), and the five fields obtained from the parsing are then linked and integrated according to the order number.

[0084] It is worth mentioning that the order number can also be used as the primary key to match records returned by different data sources. When a field value is missing in a data source, it is filled with a default value or an empty value and an exception log is generated to ensure the robustness of the data parsing process and avoid the interruption of the entire allocation process due to the lack of data from a single source.

[0085] S12. Obtain the current time, calculate the time urgency based on the promised delivery date, current time, and standard working hours, and obtain the time pressure index.

[0086] The time pressure index refers to the relative urgency of the order in terms of time.

[0087] Specifically, the current time is obtained from the server clock, and the current time, along with the promised delivery date and standard working hours for each workshop order, are entered into the calculation formula to calculate: Time Pressure Index = (Promised Delivery Date - Current System Time) / Standard Working Hours, where the promised delivery date and the current time are both converted to values ​​in hours.

[0088] S13. Calculate the processing progress based on the reported working hours and standard working hours to obtain the progress completion rate.

[0089] Progress completion rate refers to the proportion of the current completed working hours of an order to its rated total working hours, which is used to quantify the degree of completion of the order in terms of processing progress.

[0090] Understandably, the reported working hours and standard working hours of the order are entered into the calculation formula: Progress completion rate = Reported working hours / Standard working hours.

[0091] S14. Compare the current inventory quantity with the quantity required by the order to obtain the material shortage index.

[0092] The material shortage index is a binary indicator used to determine whether the key materials required for an order are sufficient, and it is used to quantify the degree of shortage of materials for the order in terms of material supply.

[0093] Specifically, the current inventory quantity of the order is compared with the quantity required by the order. If the current inventory quantity is greater than or equal to the quantity required by the order, the material shortage index of the order is 0; if the current inventory quantity is less than the quantity required by the order, the material shortage index of the order is 1. In this embodiment, a material shortage index of 1 indicates that there is a risk of material shortage, and 0 indicates that the inventory is sufficient.

[0094] S15. Use the time pressure index, progress completion rate, and material shortage index as characteristic values ​​for each workshop order.

[0095] In this embodiment of the invention, for any work-in-process order, its time pressure index is a positive floating-point number, reflecting the urgency of the order in the time dimension; its progress completion is a non-negative floating-point number, reflecting the processing progress status of the order at the current moment; and its material shortage index is a binary identifier, taking the value of 0 or 1, reflecting whether there is a risk of material shortage for the order. The above three feature values ​​are bound to the order number to form a complete feature record, and the above aggregation process is performed on all work-in-process orders one by one.

[0096] Step 102: Quantify the feature values ​​of each workshop's orders to obtain the ranking position and delay probability value of each workshop's orders.

[0097] Preferably, step 102 may include the following sub-steps:

[0098] S21. Construct an order feature set based on the feature values ​​corresponding to orders in each workshop.

[0099] Specifically, each in-process order corresponds to a feature record, which includes four fields: order number, time pressure index, progress completion rate, and material shortage index. All feature records of in-process orders are integrated into an N-row, four-column data table, where N is the total number of current in-process orders, the first three columns are the three feature values, and the fourth column is the order number as a unique identifier.

[0100] S22. Input the order feature set into the preset score prediction model to obtain the emergency score corresponding to each workshop order.

[0101] The pre-set score prediction model refers to a neural network model pre-trained based on historical order data. This model maps the multi-dimensional feature vectors of orders to a single quantized score. The model employs a three-layer neural network architecture, including a feature encoding layer and a ranking scoring layer. The feature encoding layer is used to increase the dimensionality of the input three-dimensional feature vector and extract non-linear combined features. The ranking scoring layer maps the encoded feature vectors to a floating-point score. The model's training process is supervised by the actual processing order and delay results of historical orders, ensuring that the model's output score reflects the relative urgency of orders.

[0102] The urgency score, output by a pre-defined scoring prediction model, is a quantitative value measuring the relative urgency of an order among all currently in-process orders. This score is a floating-point number with no fixed upper or lower bound. A higher score indicates a higher relative urgency for the order among all in-process orders, and it should be prioritized for processing; a lower score indicates a relatively lower urgency for the order, and it can be postponed. The urgency score is only used for horizontal comparison between different orders at the same time; scores output at different times are not comparable.

[0103] Specifically, the model receives the time pressure index, progress completion rate, and material shortage index of each order as input. After being calculated sequentially by the feature encoding layer and the ranking and scoring layer inside the model, a floating-point number is output as the urgency score of the order. The above inference calculation is performed in parallel on all orders to obtain the urgency score corresponding to each order.

[0104] S23. Based on the urgency score of each workshop order, sort all workshop orders in descending order, construct a global priority sequence, and determine the ranking position of each workshop order in the global priority sequence.

[0105] The global priority sequence refers to the ordered queue formed by sorting all orders in production according to their urgency scores from highest to lowest, and is used to reflect the relative order of all orders in processing at the current moment.

[0106] Ranking position refers to the order number of each workshop order in the global priority sequence.

[0107] Specifically, based on the urgency score of each order, all orders in production are sorted in descending order of score, with the highest score at the top of the sequence and the lowest score at the bottom, thus constructing a global priority sequence. Then, each order is determined by its unique ranking position in the sequence, which is a consecutive integer starting from 1. The higher the ranking, the higher the processing priority of the order.

[0108] S24. Based on the preset mapping function, the urgency score of each workshop order is converted to obtain the delay probability value.

[0109] The delay probability value refers to the percentage probability obtained by mapping the urgency score of an order to the range of 0% to 100% through a normalization function. It is used to quantify the risk of delivery delay when the order continues to proceed at the current pace under the current state. The higher the delay probability value, the greater the risk of delay.

[0110] Specifically, the urgency score of each order is used as input to the Sigmoid function, through the formula... The function calculates the percentage probability of delay for each order, where s is the urgency score and e is a natural constant. This function compresses the unbounded urgency score to a range of 0 to 1, multiplying it by 100% to obtain a delay probability value from 0% to 100%. For example, an order with an urgency score of 0.5 has a delay probability of approximately 62% after mapping with the Sigmoid function; if the score is 2.0, the delay probability is approximately 88%. This calculation is performed on all orders in progress to obtain the delay probability value for each order, which is then output along with its ranking position.

[0111] In this embodiment of the invention, the three-dimensional feature vectors of each order are first aggregated into a structured order feature set according to the order number. This feature set contains a unified quantitative representation of all current in-process orders. Then, this feature set is input into a pre-trained score prediction model. Based on the ranking rules learned from historical data, the model outputs a horizontally comparable urgency score for each order. Next, all in-process orders are sorted in descending order according to the urgency score, a global priority sequence is constructed, and a unique ranking position for each order is determined, realizing a comprehensive ranking of all orders on the same quantitative scale. Finally, the urgency score of each order is normalized to a delay probability value from 0% to 100% using the Sigmoid function, giving each order an intuitive risk metric. After the above processing, each workshop order obtains two quantitative results: a ranking position and a delay probability value. The ranking position reflects the relative priority of the order among all in-process orders, and the delay probability value reflects the risk level of timely delivery of the order.

[0112] The preferred training method for the preset score prediction model is as follows:

[0113] S1. Obtain the historical feature set of historical workshop orders, as well as the corresponding actual delay tags and actual priority tags.

[0114] The historical feature set refers to a multi-dimensional feature vector composed of three feature values ​​of each order in production at a historical moment. Each feature vector is associated with the moment when the feature vector was collected, the set of orders in production at that moment, and the order in which the orders were actually processed after that moment.

[0115] The actual delay label indicates whether a delay occurred when each order was finally delivered. If the actual completion date is later than the promised delivery date, it is marked as 1 to indicate a delay; otherwise, it is marked as 0 to indicate no delay.

[0116] Actual priority tags refer to the relative ordering of orders determined based on the historical actual processing sequence.

[0117] Specifically, order completion data from the past two years are extracted from the historical database to construct a sample set for training the score prediction model. At each collection, the order number of all orders in production at that moment and the corresponding time pressure index, progress completion rate and material shortage index are recorded. At the same time, the order in which each order is executed by the operator in actual production after that moment is recorded, and whether each order is ultimately delayed is also recorded.

[0118] For any two orders i and j within the same snapshot, if order i starts processing before order j in actual production, then the actual priority of order i is higher than that of order j; if order i and order j are in the same snapshot but their actual processing order is not different (such as parallel processing on different devices), then they do not constitute a sorting constraint pair and do not participate in the sorting loss calculation for that snapshot.

[0119] The snapshot data from all collection times are aggregated to form a training dataset containing a large number of order pairs. Each training sample contains the feature vectors of the two orders at the same time and the actual priority relationship between them. Each order is also independently accompanied by its actual delay label.

[0120] S2. Construct an initial score prediction network. Input the historical feature set into the initial score prediction network to obtain the emergency score corresponding to each historical workshop order.

[0121] The initial score prediction network refers to a three-layer neural network structure with randomly initialized parameters that has not been trained. It maps the input three-dimensional feature vector to a floating-point score value. This network includes a feature encoding layer and a ranking score layer. The feature encoding layer consists of a first hidden layer and a second hidden layer connected in series. The first hidden layer contains 64 neurons using the ReLU activation function, and the second hidden layer contains 32 neurons using the ReLU activation function. The ranking score layer is the output layer, containing one neuron and not using an activation function. The weights and biases of each layer are randomly assigned during initialization (e.g., using a Xavier uniform distribution) and are gradually adjusted and optimized during subsequent training using the backpropagation algorithm.

[0122] Specifically, for a single order, the initial score prediction network calculates as follows: Each neuron in the first hidden layer multiplies the input 3D feature vector by its three corresponding weights, sums the results, adds a bias value, and outputs a value using the ReLU function (formula: ReLU(x) = max(0, x)). Each of the 64 neurons calculates independently and outputs 64 values. The second hidden layer multiplies these 64 values ​​by the 64 weights of each of its 32 neurons, sums the results, adds a bias value, and outputs 32 values ​​using the ReLU function. The output layer multiplies these 32 values ​​by the 32 weights of one neuron in its layer, sums the results, adds a bias value, and directly outputs a floating-point number as the emergency score for the order. This forward propagation calculation is performed in parallel for all orders within the same batch to obtain the emergency scores for each historical workshop order.

[0123] S3. Based on the urgency score, actual priority label, and actual delay label, construct a total loss function that includes a ranking loss term and a delay prediction loss term. The ranking loss term is used to constrain the ranking of the predicted urgency score to be consistent with the ranking of the actual priority label.

[0124] Specifically, among all orders in production within the same snapshot, randomly select two orders i and j. If order i was executed before order j in the historical actual processing sequence, then the network's predicted urgency score for order i should be greater than the urgency score for order j. The formula for calculating the ranking loss term is:

[0125]

[0126] In the formula, For sorting loss terms, and These are the emergency scores of the two orders output by the network. The threshold is a preset normal number (e.g., 1.0). This is used to require that the score difference of the correctly sorted orders must reach at least this threshold for the loss to drop to 0. If the score difference is less than the threshold, the loss is positive, and the network is prompted to adjust its parameters to increase the score difference through backpropagation.

[0127] Within the same snapshot, the sorting loss is calculated for each order pair that satisfies the sequential relationship, and the average value is taken to obtain the sorting loss value for that snapshot.

[0128] Simultaneously, a delay prediction loss term is constructed for each order as an independent sample. A binary classification cross-entropy loss function is used, mapping the network's output urgency score to a delay probability value via a sigmoid function, and then calculating the cross-entropy loss with the actual delay label. The average delay loss of all orders within the same batch is taken to obtain the delay prediction loss term for that batch.

[0129] Specifically, the calculation method for the delay prediction loss term is as follows:

[0130]

[0131] In the formula, For the delay prediction loss item, M is the total number of orders in this batch. This represents the delay probability value predicted by the model for the k-th order. , The urgency score for this order is given by e, where e is the natural constant. Let be the actual delay label for the k-th order, and ln denote the logarithmic function with the natural constant ee as the base.

[0132] The total loss function is a weighted sum of the ranking loss term and the delay prediction loss term, i.e.:

[0133]

[0134] In the formula, The total loss function is defined by α and β, which are preset weighting coefficients (e.g., α=0.7, β=0.3) used to balance the contribution ratio of the two loss terms during training.

[0135] S4. Based on the total loss function, the parameters of the initial score prediction network are iteratively updated using the backpropagation algorithm until the total loss function converges, thus obtaining the preset score prediction model.

[0136] Specifically, historical samples are randomly sampled in batches, with each batch containing multiple historical snapshots. Each snapshot contains the feature vectors, actual priority labels, and actual delay labels for all orders in production at that moment. After each batch is input into the network and undergoes forward propagation, the loss value is calculated based on the constructed total loss function. Then, the gradients corresponding to the weights and biases of each layer in the network are calculated using the backpropagation algorithm. The Adam optimizer automatically adjusts the parameters based on the gradient values, gradually reducing the total loss value. The learning rate is set to 0.001, and the batch size is 256 pairs of orders (i.e., each batch contains 256 order pairs used to calculate the ranking loss and their corresponding order samples). The total number of training rounds is 100.

[0137] Meanwhile, an early stopping method is employed to prevent overfitting: after each training round, the total loss of the current model is evaluated on the validation set. If the validation set loss does not decrease in 10 consecutive training rounds, training is terminated early, and the model parameters with the minimum validation set loss are saved. After training is completed, the final determined weights and biases of each network layer are fixed and saved as a preset score prediction model for use in step S22 for emergency score inference of real-time workshop orders.

[0138] In this embodiment of the invention, the actual processing order and delay results in historical production data are used as supervision signals to jointly optimize and train the initial score prediction network. The ranking loss term forces the model to learn the relative priority relationship between orders, so that orders with higher urgency at the same time receive higher scores. The delay prediction loss term guides the model to learn the mapping rule between order characteristics and whether they will be delayed, so that the score not only reflects the relative ranking, but also implies quantitative information on delay risk. The two loss terms are weighted and combined under the same optimization objective to jointly guide the update direction of the network parameters, so that the model can simultaneously consider ranking accuracy and delay prediction accuracy during training. The score prediction model obtained after training has the complex correlation rules between the multidimensional features of orders and processing priority and delay risk in historical scheduling experience embedded in its weight parameters. In addition, a model self-update mechanism can be set up, with incremental training to fine-tune the model parameters every week and full retraining triggered every month using accumulated new historical data, to ensure that the model can continuously adapt to changes in production conditions and realize the continuous accumulation and optimization of scheduling experience.

[0139] Step 103: Determine the corresponding warning level based on the ranking position and delay probability value of each workshop's orders.

[0140] Preferably, step 103 may include the following sub-steps:

[0141] S31. Based on the preset ranking threshold range and the preset delay probability threshold range, determine the basic early warning level of each workshop's orders.

[0142] The ranking threshold range refers to a pre-defined range of values ​​based on the position in the global priority sequence, used to map the ranking position of each order to the corresponding urgency segment.

[0143] The delay probability threshold range refers to a numerical range pre-divided according to the percentage of delay probability, used to map the delay risk level of each order to the corresponding risk level segment.

[0144] The basic warning level refers to the preliminary warning level based solely on two quantitative indicators: ranking position and delay probability value, without being corrected by order attribute data. It is used to reflect the relative urgency and delay risk level of an order in the current global ranking.

[0145] Specifically, there are four preset warning levels, which are red, orange, yellow and blue in descending order of urgency. Red represents the highest level of urgency, corresponding to orders that require immediate intervention and priority processing; orange represents a very high level of urgency, corresponding to orders that need to be processed as soon as possible; yellow represents a moderate level of urgency, corresponding to orders that require close attention but can be processed at the normal pace for the time being; and blue represents a normal level, corresponding to orders that can be processed in the normal order.

[0146] The ranking threshold intervals are divided into four intervals according to their position in the global priority sequence: the first to third place are the first interval, the fourth to eighth place are the second interval, the ninth to fifteenth place are the third interval, and the sixteenth place and above are the fourth interval.

[0147] The delay probability threshold range is divided into three intervals based on percentage: greater than or equal to 85% is the high probability interval, greater than or equal to 60% and less than 85% is the medium-high probability interval, greater than or equal to 30% and less than 60% is the medium-low probability interval, and less than 30% is the low probability interval.

[0148] Understandably, the judgment process first determines the ranking interval based on the order's position, then determines the probability interval based on the delay probability value, and finally determines the basic warning level by combining both. To clearly illustrate the judgment logic, the correspondence between the basic warning levels for each combination of ranking interval and delay probability interval is shown in the table below:

[0149] 1st to 3rd place ≥85% red 1st to 3rd place ≥60% and <85% orange color 1st to 3rd place ≥30% and <60% orange color 1st to 3rd place <30% orange color 4th to 8th place ≥85% orange color 4th to 8th place ≥60% and <85% orange color 4th to 8th place ≥30% and <60% yellow 4th to 8th place <30% yellow 9th to 15th place ≥85% yellow 9th to 15th place ≥60% and <85% yellow 9th to 15th place ≥30% and <60% yellow 9th to 15th place <30% blue 16th place and below ≥85% blue 16th place and below ≥60% and <85% blue 16th place and below ≥30% and <60% blue 16th place and below <30% blue

[0150] It should be noted that the division of the above ranking range and delay probability range can be adjusted according to actual needs.

[0151] S32. Obtain the order attribute data of each workshop's orders, and correct the basic warning level based on the order attribute data to obtain the final warning level corresponding to each workshop's orders.

[0152] Order attribute data refers to supplementary information related to the order's own attributes, in addition to the aforementioned feature values. This includes, but is not limited to, customer level and material shortage flag. The customer level is a preset level label (such as S-level, A-level, B-level, etc.) set by the enterprise based on the importance of the customer, which is used to reflect the service priority of the customer corresponding to the order. The material shortage flag is the calculated material shortage index, which takes a value of 0 or 1, and is used to reflect whether the key materials required for the order are sufficient.

[0153] The revised rules are as follows: If the material shortage index of an order is 1 (i.e. there is a shortage of materials), then regardless of the basic warning level, its final warning level will be directly raised to red; if the customer level of the order is S-level customer, then the warning level will be raised by one level based on the basic warning level, i.e., blue will be raised to yellow, yellow to orange, orange to red, and red will remain unchanged (because red is already the highest level and cannot be raised further).

[0154] When an order meets multiple correction rules at the same time, the rule with the largest increase shall be executed. For example, if the basic warning level of an order is yellow, its customer level is S and its material shortage index is 1, then the two rules should be upgraded to orange and red respectively. The rule with the largest increase shall be used and it shall be directly upgraded to red.

[0155] For example, an order's base warning level is yellow. If its customer level is S, the final warning level is raised to orange. If the order also has a material shortage (material tension index of 1), it is directly raised to red. Another order's base warning level is blue. If its customer level is S, it is raised to yellow; if there is a material shortage, it is directly raised to red. The above correction judgment is performed on all in-process orders to obtain the final warning level for each order.

[0156] It should be noted that the final warning level is used to determine the order push strategy and confirmation time limit, while the ranking position is used to determine the processing order of the order in the employee's personal to-do list. The correction of the final warning level does not affect the ranking position value. The two are logically independent and manage orders from the two dimensions of risk notification and execution order, respectively, so as to decouple the information push intensity and the processing execution order and avoid resource misallocation at the execution level due to the over-strengthening of a single dimension.

[0157] In this embodiment of the invention, a two-dimensional joint judgment is first made based on the ranking position and the delay probability value. The ranking reflects the relative priority order between orders, and the delay probability reflects the absolute risk level of the order. The two are mutually verified and complementary, ensuring that the basic warning level takes into account both the objectivity of the global ranking and the absolute level of delay risk. Subsequently, order attribute data is introduced for differentiated correction, so that special factors such as customer level and material shortage can be reasonably reflected in the warning, avoiding misjudgment or omission of special orders by general rules. The final warning level is output with a visual color label (red, orange, yellow, blue), and each order carries a unique final warning level label.

[0158] Step 104: Based on the preset workshop resource configuration information table, assign each workshop order to the corresponding employee and generate an exclusive to-do list for each employee.

[0159] Preferably, step 104 may include the following sub-steps:

[0160] S41. Obtain the workshop resource configuration information table, which records the relationship between the current process of each workshop order and the corresponding operator.

[0161] The workshop resource allocation information table refers to a data table that is pre-configured and stored in the system. It is used to record the relationship between the current process of each workshop order and the operator responsible for that process. The information table includes fields such as order number, current process number, equipment number, operator name, and operator ID. It is pre-entered by workshop managers according to the actual production resource allocation and is regularly maintained and updated. By reading this table, the specific operator information that each work-in-process order should belong to can be obtained.

[0162] S42. Based on the workshop resource allocation information table, find the employee corresponding to the current process in each workshop order, and assign each workshop order to the found employee.

[0163] Specifically, based on the obtained workshop resource allocation information table, for each work-in-process order whose final warning level has been determined, its current process information is extracted, the operator's name and employee number associated with the current process are found in the workshop resource allocation information table, and the order is assigned to the found operator's name, thus establishing the affiliation relationship between the order and the operator.

[0164] For example, if the current process of an order is CNC milling precision machining, and the operator corresponding to this process is found to be Zhang San in the workshop resource allocation information table, then the order will be assigned to Zhang San.

[0165] Perform the above search and allocation operations on all in-process orders one by one. If a certain order cannot be found in the workshop resource configuration information table (e.g., because the process has not yet been assigned a responsible person), then the order is assigned to the team leader of the work group where the process is located, and a configuration missing log is generated for management personnel to check and supplement, so as to ensure that each order has a clear responsibility.

[0166] S43. For each employee, sort all workshop orders assigned to that employee according to their ranking position and generate an employee's personal to-do list.

[0167] Specifically, for each operator, all workshop orders assigned to them are filtered out and sorted in ascending order according to their global ranking position (i.e., the numerical value of the ranking position, with smaller values ​​indicating higher priority). Orders with higher ranking positions are placed at the top of the to-do list, and orders with lower ranking positions are placed at the bottom, thus generating a personalized to-do list for that operator. Each operator's personalized to-do list includes, but is not limited to, fields such as order number, ranking position, delay probability value, final warning level, and suggested actions.

[0168] For example, if an operator has three orders assigned to them, ranking 2nd, 5th, and 8th globally, then after sorting them in ascending order of ranking, the order in the operator's personal to-do list will be the 2nd order, the 5th order, and the 8th order, respectively. This aggregation and sorting operation is performed on all operators one by one to generate a personal to-do list for each operator.

[0169] In this embodiment of the invention, a resource configuration information table recording the relationship between orders and operators is first obtained. Then, each order is accurately matched to the specific operator corresponding to the current process, ensuring that orders are accurately dispatched according to responsibility. Finally, the orders are aggregated and sorted by operator, and the global ranking sequence is decomposed into an ordered to-do list from the perspective of each operator. After the above processing, each order in production has a clear specific responsibility, and each operator can clearly know the order and urgency of all their current pending orders without having to manually filter and identify them from factory-wide mass notifications. This effectively avoids information overload and ambiguity of responsibilities. It is worth mentioning that the generated to-do list for each employee is stored in a Redis cache, preferably refreshed every 5 minutes to keep it synchronized with the actual production status.

[0170] Step 105: Based on the warning level of orders in each workshop, determine the push channel and push frequency corresponding to the warning level, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

[0171] Specifically, for orders under red alert, push notifications can be sent simultaneously through four channels: SMS, voice call, DingTalk message, and full-screen pop-up window on the workbench. The recipients of the notifications are the employees who own the order and their workshop directors. The notifications are sent every 5 minutes until a confirmation reply is received.

[0172] For orders under orange alert, notifications can be sent via DingTalk messages and highlighted items in the workbench list. The notifications should be sent to the employees who own the order, and the notifications should be sent every 15 minutes until a confirmation reply is received.

[0173] For orders with a yellow warning, the yellow mark can be displayed only in the workbench list, without proactive push notifications or requirements for repeated pushes.

[0174] For orders under blue alert, they can simply be displayed normally in the workbench list without being actively pushed to the system, and there is no requirement for repeated pushes.

[0175] It should be noted that SMS and voice calls are sent through the enterprise SMS gateway and voice gateway, DingTalk messages are pushed through the DingTalk enterprise application interface, and the workbench display is rendered through the system's web page. The status of all orders in processing is scanned every 5 minutes. For orders in red and orange alerts that have not yet been confirmed, push notifications are executed according to the above strategy. After a push is completed, the push record (including push time, push channel, push recipient, and push content summary) is written to the database for subsequent traceability and querying.

[0176] In this embodiment of the invention, through the above-mentioned differentiated push strategy, orders with different urgency levels are notified in different intensities. Red alerts ensure that the information is delivered with the highest frequency and the most channels, orange alerts ensure timely response with a moderate frequency and channels, and yellow and blue alerts are displayed silently on the interface to avoid excessive interference to employees, thus achieving a precise match between notification intensity and urgency.

[0177] Preferably, step 105 may include the following sub-steps:

[0178] S51. Determine the corresponding confirmation time limit based on the warning level.

[0179] Confirmation time limit refers to the maximum allowed time interval from the first push of the exclusive to-do list to the corresponding employee until the employee must complete the confirmation operation. It is used to constrain the response time of orders with different levels of urgency.

[0180] Specifically, the confirmation time limit for red alert orders is 5 minutes, the confirmation time limit for orange alert orders is 15 minutes, and there is no mandatory confirmation time limit for yellow and blue alert orders (i.e., no employee needs to perform a confirmation operation). The start time for the confirmation time limit is the moment when the first push is executed, and the end time is the moment when the employee's confirmation instruction is received. If the time limit exceeds the confirmation time limit of the corresponding alert level, the timeout processing logic will be triggered.

[0181] S52. If a confirmation instruction for the exclusive to-do list is received from an employee within the confirmation period, the exclusive to-do list will not be repeatedly pushed to the employee.

[0182] After pushing the personalized to-do list, continuously monitor the employee's confirmation actions. When the employee completes the confirmation by clicking the confirmation button on the workbench interface or by clicking the confirmation button on the DingTalk message card, receive the confirmation instruction and record the confirmer's name, confirmation time and confirmation method.

[0183] Determine whether the receipt time of the confirmation instruction is within the confirmation time limit of the corresponding warning level. If so, immediately stop repeatedly pushing the exclusive to-do list information of the order to the employee (i.e., stop subsequent repeated pushes via SMS, phone, DingTalk, etc.), and downgrade the warning level of the order by one level, i.e., from red to orange, from orange to yellow, from yellow to blue, while blue remains unchanged. The downgraded warning status is stored in the system cache and used to update the workbench display label and adjust the subsequent push strategy.

[0184] S53. If no confirmation instruction is received from the employee within the confirmation period, the early warning escalation mechanism will be triggered, and the exclusive to-do list will be pushed to the employee's corresponding supervisor.

[0185] Understandably, the confirmation time limit is used as the monitoring window to continuously monitor the confirmation status of each order. If a red alert order does not receive a confirmation instruction from the employee within 5 minutes after the first push, or an orange alert order does not receive a confirmation instruction from the employee within 15 minutes after the first push, it is determined to be unconfirmed within the time limit, triggering the alert escalation mechanism.

[0186] Specifically, if a red alert remains unconfirmed after a timeout, an automatic SMS and voice call notification is sent to the workshop director of the employee, and the order information is pushed to the workshop director's workstation. Once the workshop director confirms receipt, a record is created indicating that the process has been initiated. If an orange alert remains unconfirmed after a timeout, a DingTalk message notification is automatically sent to the team leader of the employee, and the order information is pushed to the team leader's watchlist. Yellow and blue alerts do not have a mandatory confirmation timeout and therefore do not trigger alert escalation. After an alert escalates, the confirmation status of the recipient continues to be monitored. If no confirmation is received within the preset time window after the escalation, the alert is pushed to the next higher level of management, forming a hierarchical confirmation process.

[0187] It should be noted that the hierarchical reporting can be promoted step by step in the order of employee - team leader - workshop director - production allocation supervisor, with each level having a corresponding response confirmation deadline. If the recipient at the current level does not perform the confirmation operation within the corresponding response confirmation deadline, the exclusive to-do list will be automatically pushed to the next higher level of management, and the timestamp and recipient of each promotion push will be recorded.

[0188] Once the dedicated to-do list is pushed to the highest-level manager who reports through the chain of command, if no confirmation instruction is received within the corresponding response confirmation period, an emergency allocation work order will be automatically generated. The highest-level manager and the production director will be notified simultaneously via SMS and telephone voice, and the order will be marked as a global intervention status.

[0189] In this embodiment of the invention, different confirmation time limits are set for different warning levels, linking timeliness control with risk level. Once confirmed, repeated pushes are immediately stopped to avoid excessive interference with operators. If confirmation is not made within the time limit, a hierarchical reporting mechanism is automatically triggered. This effectively solves the defects of traditional verbal or mass notification methods, such as easy omission of information and difficulty in tracing responsibility, ensuring that urgent orders are responded to and processed in a timely manner, and further improving the closed-loop management level of production allocation.

[0190] Example 2:

[0191] Please see Figure 2 The present invention provides a workshop order allocation system based on multi-feature evaluation, comprising:

[0192] The data acquisition module 201 is used to collect the raw data of all workshop orders and calculate the corresponding feature values ​​based on the raw data.

[0193] The quantization processing module 202 is used to quantify the feature values ​​of orders from each workshop to obtain the ranking position and delay probability value of each workshop order.

[0194] The grade determination module 203 is used to determine the corresponding warning grade based on the ranking position and delay probability value of each workshop's orders.

[0195] The task allocation module 204 is used to allocate orders from each workshop to the corresponding employees based on a preset workshop resource configuration information table, and generate an exclusive to-do list for each employee.

[0196] The list push module 205 is used to determine the push channel and push frequency corresponding to the warning level of each workshop's orders, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

[0197] Since the above is a system corresponding to a workshop order allocation method based on multi-feature evaluation, its implementation principle is consistent with a workshop order allocation method based on multi-feature evaluation. For the sake of convenience and brevity, those skilled in the art can clearly understand that the specific working process of the system and modules described above can be referred to the corresponding process in the aforementioned method embodiments, and will not be repeated here.

[0198] Example 3:

[0199] Please see Figure 3 An electronic device according to an embodiment of the present invention includes: a memory 301 and a processor 302. The memory 301 stores a computer program. When the computer program is executed by the processor, the processor 302 executes the workshop order allocation method based on multi-feature evaluation as described in any of the above embodiments.

[0200] Memory 301 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Memory 301 has storage space 303 for program code 304 for performing any of the method steps described above. For example, storage space 303 for program code may include various program codes 304 for implementing the various steps in the methods described above. This program code may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the methods described above.

[0201] Example 4:

[0202] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed, implements the workshop order allocation method based on multi-feature evaluation of any of the above embodiments.

[0203] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0204] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0205] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0206] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0207] If the integrated unit is implemented as 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 invention, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0208] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A workshop order allocation method based on multi-feature evaluation, characterized in that, include: Collect raw data of all workshop orders and calculate the corresponding feature values ​​based on the raw data; The feature values ​​of each workshop order are quantified to obtain the ranking position and delay probability value of each workshop order; Based on the ranking position and delay probability value of each workshop order, the corresponding warning level is determined; Based on the preset workshop resource configuration information table, each workshop order is assigned to the corresponding employee, and a unique to-do list is generated for each employee. Based on the warning level of each workshop order, determine the push channel and push frequency corresponding to the warning level, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

2. The workshop order allocation method based on multi-feature evaluation according to claim 1, characterized in that, The specific method for calculating the corresponding feature value based on the original data is as follows: The raw data is parsed to obtain the promised delivery date, standard working hours, reported working hours, current inventory quantity, and required quantity for each workshop order; Obtain the current time, calculate the time urgency based on the promised delivery date, the current time, and the standard working hours, and obtain the time pressure index; The processing progress is calculated based on the reported working hours and the standard working hours to obtain the progress completion rate. By comparing the current inventory quantity with the quantity required by the order, the material shortage index is obtained; The time pressure index, the progress completion rate, and the material shortage index are used as characteristic values ​​corresponding to each workshop order.

3. The workshop order allocation method based on multi-feature evaluation according to claim 1, characterized in that, The quantification of the feature values ​​of each workshop order to obtain the ranking position and delay probability value of each workshop order includes: Based on the feature values ​​corresponding to each workshop order, an order feature set is constructed; The order feature set is input into a preset score prediction model to obtain the emergency score corresponding to each workshop order; Based on the urgency score of each workshop order, all workshop orders are sorted in descending order to construct a global priority sequence, and the ranking position of each workshop order in the global priority sequence is determined. The urgency score of each workshop order is converted based on a preset mapping function to obtain the delay probability value.

4. The workshop order allocation method based on multi-feature evaluation according to claim 3, characterized in that, The training method for the preset score prediction model is as follows: Obtain the historical feature set of historical workshop orders, along with the corresponding actual delay tags and actual priority tags; Construct an initial score prediction network, input the historical feature set into the initial score prediction network, and obtain the emergency score corresponding to each historical workshop order; Based on the emergency score, the actual priority label, and the actual delay label, a total loss function is constructed that includes a ranking loss term and a delay prediction loss term, wherein the ranking loss term is used to constrain the ranking of the predicted emergency score to be consistent with the ranking of the actual priority label; Based on the total loss function, the parameters of the initial score prediction network are iteratively updated using the backpropagation algorithm until the total loss function converges, thus obtaining the preset score prediction model.

5. The workshop order allocation method based on multi-feature evaluation according to claim 1, characterized in that, The determination of the corresponding early warning level based on the ranking position and delay probability value of each workshop order includes: Based on the preset ranking threshold range and the preset delay probability threshold range, the basic early warning level of each workshop order is determined; Obtain the order attribute data of each workshop order, and correct the basic warning level based on the order attribute data to obtain the final warning level corresponding to each workshop order.

6. The workshop order allocation method based on multi-feature evaluation according to claim 1, characterized in that, Based on a preset workshop resource allocation information table, each workshop order is assigned to a corresponding employee, generating a personalized to-do list for each employee, including: Obtain the workshop resource configuration information table, which records the association between the current process and the corresponding operator for each workshop order; Based on the workshop resource allocation information table, find the employee corresponding to the current process in each workshop order, and assign each workshop order to the found employee. For each employee, all workshop orders assigned to that employee are sorted according to their ranking position, generating a personalized to-do list for that employee.

7. The workshop order allocation method based on multi-feature evaluation according to claim 1, characterized in that, After determining the push channel and push frequency corresponding to the warning level based on the warning level of each workshop order, and pushing the exclusive to-do list to the corresponding employee based on the push channel and push frequency, the process includes: The corresponding confirmation time limit is determined based on the aforementioned warning level; If a confirmation instruction for the personalized to-do list is received from the employee within the confirmation time limit, the repeated pushing of the personalized to-do list to the employee will be stopped. If no confirmation instruction is received from the employee within the specified confirmation period, an escalation warning mechanism will be triggered, and the exclusive to-do list will be pushed to the employee's corresponding supervisor.

8. A workshop order allocation system based on multi-feature evaluation, characterized in that, include: The data acquisition module is used to collect the raw data of all workshop orders and calculate the corresponding feature values ​​based on the raw data; The quantization processing module is used to quantify the feature values ​​of each workshop order to obtain the ranking position and delay probability value of each workshop order. The level determination module is used to determine the corresponding warning level based on the ranking position and delay probability value of each workshop order; The task allocation module is used to allocate each workshop order to the corresponding employee based on a preset workshop resource configuration information table, and generate an exclusive to-do list for each employee. The list push module is used to determine the push channel and push frequency corresponding to the warning level of each workshop order, and push the exclusive to-do list to the corresponding employee based on the push channel and push frequency.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any one of claims 1 to 7, which is a workshop order allocation method based on multi-feature evaluation.

10. A computer-readable storage medium, characterized in that, The computer program is stored and can be loaded by a processor and executed as described in any one of claims 1 to 7, which is a workshop order allocation method based on multi-feature evaluation.

Images