Internet laundry order resource accounting method and system based on LSTM

Abstract

The invention discloses an Internet laundry order resource accounting method and system based on LSTM, and relates to the technical field of intelligent decision making, and the method comprises the steps: constructing an LSTM model, and outputting the real-time fatigue index of each physical resource; receiving and analyzing an internet clothes washing order, and generating a process link; screening out available resources of which the current fatigue index is lower than a safety threshold from all physical resources of the physical resource type to form a candidate resource set; selecting a target physical resource combination for the process link based on the resource combination optimization logic; for a high mechanical load link, selecting a physical resource with the lowest current fatigue index; for a high-heat-load or high-precision requirement link, selecting a physical resource of which the current fatigue index has the largest difference from the average fatigue index of the type of resources; and according to the target physical resource combination and the fatigue index, calculating estimated time consumption and energy consumption of each process link.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, specifically to an LSTM-based method and system for calculating internet laundry order resources. Background Technology

[0002] With the popularization of the O2O model, traditional laundry services are rapidly transforming into a centralized, platform-based internet-based laundry model. In this model, orders are processed centrally at a central laundry plant, placing extremely high demands on the refined management and efficient utilization of internal resources. Current technologies often employ static estimation methods for laundry order resource accounting, such as forecasting based on fixed standard working hours and average energy consumption, or simply linearly adding data based on clothing type and quantity. However, in actual operation, the performance of various physical resources in a laundry plant (such as washing machines, dryers, and ironing machines) is not constant; their processing efficiency and energy consumption dynamically decline with continuous work and accumulated load, exhibiting a fatigue effect. This dynamic change in performance is not effectively quantified or considered in existing technologies, leading to significant discrepancies between resource accounting results and actual consumption, and failing to accurately guide production scheduling, cost control, and energy efficiency management. Summary of the Invention

[0003] The purpose of this invention is to provide an LSTM-based method and system for calculating internet laundry order resources, in order to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an LSTM-based method for calculating internet laundry order resources, the method comprising: S100. Construct an LSTM model. The LSTM model takes the historical workload sequence of each physical resource as input and outputs the real-time fatigue index of each physical resource. The physical resources include washing machines, dryers, ironing machines, and sorting lines. S200: Receive internet laundry orders, parse the clothing categories and process requirements included in the laundry orders, and generate process links; the process links include the types and order of physical resources used to complete the laundry orders; S300. Based on the process chain, for each process step in the order, select available resources from all physical resources of the physical resource type of the laundry order whose current fatigue index is lower than the safety threshold, and form a candidate resource set for each step. S400. Based on resource combination optimization logic, select target physical resource combinations for the process link; the optimization logic includes selecting the physical resource with the lowest current fatigue index for high mechanical load links; and selecting the physical resource with the largest difference between the current fatigue index and the average fatigue index of this type of resource for high thermal load or high precision requirements. S500. Based on the target physical resource combination and fatigue index, use the efficiency mapping table to calculate the estimated time and energy consumption of each process step, and then summarize to obtain the total resource calculation result.

[0005] Based on the above, step S100 includes: S110. The historical workload sequence is a set of parameters arranged in chronological order and representing the historical working status of each physical resource. The parameters include the continuous working duration of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and the equipment start-up and shutdown events. During the parameter acquisition process, the acquisition time and acquisition equipment identifier of the parameters are recorded synchronously. The acquired parameters are deduplicated, outlier removed, and missing value filled to ensure the integrity and consistency of the sequence data. S120, the fatigue index, is a normalized value representing the expected performance degradation and potential failure risk level of the physical resource due to accumulated workload. This index is negatively correlated with the real-time operating efficiency of the equipment and positively correlated with the required maintenance urgency. The smaller the value, the lower the fatigue level and the better the performance of the physical resource; the larger the value, the higher the fatigue level and the more obvious the performance degradation. Real-time operating efficiency decreases linearly with the increase of the fatigue index, while maintenance urgency increases stepwise with the increase of the fatigue index. When the index reaches the preset warning value, a maintenance reminder is triggered. The S130 LSTM model consists of an input layer, an LSTM hidden layer, and an output layer. The input layer is used to receive multidimensional time series data of historical workload sequences. The LSTM hidden layer is used to learn long-term dependencies and dynamic patterns in the workload sequences. The output layer is used to map the state of the LSTM hidden layer to a fatigue index. S140. Using historical operational data collected during historical operations as the training set, an LSTM model is trained in a supervised learning manner. The labeled data of the training set corresponds to the historical workload sequence and is based on the baseline fatigue level constructed from actual equipment performance degradation records or maintenance logs. Historical operational data is collected continuously at fixed time intervals, with the collection period preset based on the operational characteristics of physical resources. The training set is divided into a training sample set and a validation sample set according to a preset ratio for model training and performance verification. The baseline fatigue level is comprehensively calibrated by the ratio of actual performance parameters to rated performance parameters, the frequency of fault occurrence recorded in the maintenance log, and the maintenance interval duration. The calibration process follows a unified quantitative standard.

[0006] Based on the above, step S200 includes: S210. Receive internet laundry orders, parse the clothing category information of the garment attributes in the order data, and the service process requirements information for cleaning and treatment methods; receive order data in real time through the interface of the internet laundry platform. The order data is stored in a structured format, including a unique order identifier, user identifier, garment details list, and service requirements fields; use semantic parsing algorithms to extract material attributes and specifications from the garment category information, and decompose the cleaning methods, treatment standards, and special requirements from the service process requirements information to ensure the comprehensiveness and accuracy of the parsed information; S220. Based on the process knowledge base, the parsed clothing categories and process requirements are mapped to one or more discrete process steps necessary to complete the order. Each process step is associated with the physical resource type required to execute that step. The process knowledge base is generated through domain expert experience configuration, establishing mapping rules between clothing categories, material attributes, stain types, user service selections, and standardized process steps. The process knowledge base establishes a regular update mechanism, dynamically adjusting the mapping rules based on newly added clothing categories, new processing technologies, and actual production verification results. A unique standardized process step is precisely matched according to the combination of clothing category, material attributes, stain type, and user service selection. Each process step is pre-bound with the required physical resource type code to ensure the compatibility between process steps and physical resources. S230. Based on the physical dependencies and business logic between each process step, determine the execution order of the process steps and construct the process link; the process link includes the physical resource types and order for completing the laundry order.

[0007] Based on the above, step S300 includes: S310. Based on each process step in the process chain, obtain the real-time fatigue index of all physical resources corresponding to the physical resource type associated with each step. S320. Safety thresholds are obtained based on the rated performance parameters of the equipment, maintenance cycle standards, and historical fault statistics of physical resources, and are configured separately according to the type of physical resource. The rated performance parameters of the equipment include rated load, rated operating time, and rated efficiency. The maintenance cycle standard is the maintenance interval recommended by the equipment manufacturer. The historical fault statistics are the average fatigue index at the time of the fault in the last three maintenance cycles. Safety thresholds are calibrated periodically based on the cumulative operating time of physical resources, the number of maintenance operations, and the effect of fault repair. The calibration cycle is synchronized with the maintenance cycle. S330. Filter out physical resources whose real-time fatigue index is lower than the corresponding safety threshold and mark them as available resources; compare the real-time fatigue index of the physical resources with the corresponding type of safety threshold, and determine that physical resources whose real-time fatigue index is less than the safety threshold are available resources; assign a unique available identifier to the available resources, which includes the physical resource number, the current fatigue index and the determination result, to ensure the traceability of available resources. S340. Organize the available resources of each process step according to the process steps and construct a candidate resource set. The candidate resource set includes the available resource identifier, current fatigue index and equipment operating status information of each step. The candidate resource set is stored in a linked list structure according to the process steps. Each linked list node contains a process step identifier, a list of available resources and a resource status description field. The equipment operating status information includes three states: running, idle and awaiting maintenance. It is updated in real time through the production management system and synchronized to the candidate resource set to ensure the timeliness of the status information.

[0008] Based on the above, step S400 includes: S410, High Mechanical Load Stages are process stages where the mechanical transmission strength of the equipment reaches a preset mechanical threshold during execution; High Thermal Load Stages are process stages where the thermal output intensity of the equipment reaches a preset thermal threshold during execution; High Precision Requirement Stages are process stages where the precision indicators such as dimensional error and flatness must meet preset precision standards during execution; The preset mechanical thresholds are determined based on the rated mechanical strength of physical resources, material fatigue limits, and long-term operation test data; the preset thermal thresholds are based on the rated heat dissipation capacity of the equipment and the trigger threshold of the thermal protection device; the preset precision standards are formulated with reference to industry-standard garment processing precision specifications and user-required thresholds; the load type is determined by collecting equipment operating parameters during the execution of the process stages and comparing them with the preset thresholds; S420. Calculate the average fatigue index of each physical resource type. The average fatigue index is the arithmetic mean of the current fatigue indices of all physical resources under that type. It is calculated in real time when the resource combination optimization process of each order is started. The calculation scope is all registered physical resources under that physical resource type, including resources in operation and idle state. S430. For high mechanical load processes, select the physical resource with the lowest current fatigue index from the corresponding candidate resource set; sort the physical resources in the candidate resource set in ascending order of current fatigue index, and select the physical resource ranked first as the suitable resource for this process; if there are multiple physical resources with the same current fatigue index and the lowest value, select the physical resource in the idle state according to the priority of equipment operation status. S440. For processes with high heat load or high precision requirements, select the physical resource with the largest difference between the current fatigue index and the average fatigue index of the corresponding type of resource from the corresponding candidate resource set; the difference is the absolute difference between the current fatigue index of the physical resource and the average fatigue index of the corresponding type, and the difference is calculated using absolute value operation to ensure that the result is non-negative; sort the physical resources in the candidate resource set in descending order of difference, and select the physical resource with the first position as the suitable resource for this process step; S450. Integrate the physical resources selected in each process step to form a target physical resource combination that adapts to the process link. After integration, perform conflict verification on the target physical resource combination to check whether each physical resource is occupied by multiple process steps at the same time. If there is a conflict, reselect the adaptable resources. The target physical resource combination is stored in a matrix structure, with the row dimension being the process step order and the column dimension being the physical resource identifier, current fatigue index, and operating status.

[0009] Based on the above, step S500 includes: S510, Efficiency Mapping Table: Establishes a mapping relationship between historical time consumption and energy consumption data of each process step based on different physical resource types and different fatigue index ranges; establishes a mapping relationship between time consumption and energy consumption and physical resource type and fatigue index by statistically analyzing historical operating data of each process step under different physical resource types and different fatigue index ranges, using regression analysis algorithm; the efficiency mapping table is updated at a fixed period, the update period is based on the cumulative amount of historical data and the preset optimization adjustment frequency of the process step, and outdated mapping relationships are replaced synchronously during the update; S520. Based on the type of each physical resource in the target physical resource combination and the real-time fatigue index, query the efficiency mapping table to obtain the unit time consumption parameter and unit energy consumption parameter of the corresponding process step. S530. Based on the processing volume requirements of each process step, calculate the estimated time and energy consumption of each process step; and according to the execution order of the process link, summarize the estimated time and energy consumption of all process steps to obtain the total resource calculation result of the laundry order; the total resource calculation result includes the total time, total energy consumption and the sub-item resource consumption data of each process step.

[0010] An LSTM-based internet laundry order resource accounting system, comprising: a resource monitoring module, a process analysis module, a resource screening module, a resource scheduling module, and a resource accounting module; The resource monitoring module is used to build an LSTM model. It takes the historical workload sequence of each physical resource as input and outputs the real-time fatigue index of each physical resource. The physical resources include washing machines, dryers, ironing machines and sorting lines. The process analysis module is used to receive online laundry orders, analyze the clothing categories and process requirements included in the laundry orders, and generate process links; the process links include the types and order of physical resources used to complete the laundry order; The resource filtering module is used to filter available resources with a current fatigue index below the safety threshold from all physical resources of the physical resource type in the laundry order for each process link in the order, based on the process link, and form a candidate resource set for each link. The resource scheduling module is used to select target physical resource combinations for the process link based on resource combination optimization logic. The optimization logic includes selecting the physical resource with the lowest current fatigue index for high mechanical load links and selecting the physical resource with the largest difference between the current fatigue index and the average fatigue index of that type of resource for high thermal load or high precision requirements. The resource accounting module is used to calculate the estimated time and energy consumption of each process step based on the target physical resource combination and fatigue index, using an efficiency mapping table, and then summarizing the results to obtain the total resource accounting result.

[0011] Based on the above, the resource monitoring module includes a sequence construction unit and an LSTM model unit; The sequence construction unit is used to collect historical operation data and construct a set of parameters arranged in chronological order to represent the historical working status of each physical resource. The parameters include the continuous working time of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and the equipment start-up and shutdown events. The LSTM model unit consists of an input layer, an LSTM hidden layer, and an output layer. The input layer receives multidimensional time series data of historical workload sequences. The LSTM hidden layer learns long-term dependencies and dynamic patterns in the workload sequences. The output layer maps the states of the LSTM hidden layer to a normalized fatigue index.

[0012] Based on the above, the process analysis module includes an information analysis unit and a process link unit; The information parsing unit is used to receive online laundry orders, parse the clothing category information of the clothing attributes in the order data, and the service process requirements information of cleaning and treatment methods; The process link unit is used to map the parsed clothing categories and process requirements to one or more discrete process steps necessary to complete the order, based on a pre-built process knowledge base. Each process step is associated with the physical resource type necessary to execute that step. The execution order of the process steps is determined according to the physical dependencies between them and the business logic, thus constructing the process link.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention introduces an LSTM model to predict the fatigue of physical resources in real time, enabling resource accounting to fully consider the dynamic degradation of equipment performance and improving the accuracy of time consumption and energy consumption prediction. 2. This invention uses a fatigue-based resource combination optimization logic to route high-load orders to the healthiest equipment, promoting a balanced distribution of fatigue across equipment, avoiding excessive wear and tear on a single device, improving resource utilization efficiency, and extending equipment lifespan. 3. This invention reduces equipment idle and inefficient operation time by precisely matching process links and resources, while the resource accounting based on fatigue optimization can accurately control energy consumption and time consumption. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the structure of an Internet laundry order resource accounting system based on LSTM according to the present invention; Figure 2 This is a flowchart illustrating the steps of an LSTM-based internet laundry order resource accounting method according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example: Figures 1-2 As shown, this invention provides a technical solution: an LSTM-based method for calculating internet laundry order resources. The method includes the following steps: S100. Construct an LSTM model. The LSTM model takes the historical workload sequence of each physical resource as input and outputs the real-time fatigue index of each physical resource. The physical resources include washing machines, dryers, ironing machines, and sorting lines. Specifically, step S100 includes: S110. The historical workload sequence is a set of parameters arranged in chronological order and representing the historical working status of each physical resource. The parameters include the continuous working duration of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and the equipment start-up and shutdown events. During the parameter acquisition process, the acquisition time and acquisition equipment identifier of the parameters are recorded synchronously. The acquired parameters are deduplicated, outlier removed, and missing value filled to ensure the integrity and consistency of the sequence data. For example, taking a washing machine as an example, its historical workload sequence contains a set of parameters arranged in hourly order. One sequence segment is: [Continuous working time: 4.5 hours, Order category: down jackets, Quantity: 3 pieces, Processing intensity level: Level 2, Equipment start / stop event: Start]. When collecting these parameters, the collection time and equipment identifier are recorded simultaneously. The collected raw data is processed to remove obviously abnormal values ​​such as -1 or 999 in the continuous working time, and missing values ​​caused by temporary sensor failures are filled in by linear interpolation of data from previous and subsequent time points. This is only an example and is not a limitation. S120, the fatigue index, is a normalized value representing the expected performance degradation and potential failure risk level of the physical resource due to accumulated workload. This index is negatively correlated with the real-time operating efficiency of the equipment and positively correlated with the required maintenance urgency. The smaller the value, the lower the fatigue level and the better the performance of the physical resource; the larger the value, the higher the fatigue level and the more obvious the performance degradation. Real-time operating efficiency decreases linearly with the increase of the fatigue index, while maintenance urgency increases stepwise with the increase of the fatigue index. When the index reaches the preset warning value, a maintenance reminder is triggered. For example: the fatigue index is normalized to the range of 0-100; a newly put into use dryer has an initial fatigue index of 5; after continuous high-intensity work, the index rises to 75; a maintenance reminder is triggered when the index reaches 80; the relationship between real-time operating efficiency and fatigue index is: for every 1 increase in fatigue, efficiency decreases linearly by 0.6%; the maintenance urgency is set as follows: an index below 60 is normal, 60-80 is of concern, and above 80 is a tiered level requiring urgent maintenance; this is only an example and is not a limitation. The S130 LSTM model consists of an input layer, an LSTM hidden layer, and an output layer. The input layer is used to receive multidimensional time series data of historical workload sequences. The LSTM hidden layer is used to learn long-term dependencies and dynamic patterns in the workload sequences. The output layer is used to map the state of the LSTM hidden layer to a fatigue index. For example: the model input layer receives a time series with a dimension of 5, i.e., 5 parameters, and the sequence length is set to 24, i.e., it examines data from the past 24 hours; the LSTM hidden layer has 50 neurons and introduces the Dropout mechanism to prevent overfitting; the output layer uses the Sigmoid activation function to scale the final output value to the range of 0-100 as a fatigue index; this is just an example and is not a limitation. S140. Using historical operational data collected during historical operations as the training set, an LSTM model is trained in a supervised learning manner. The labeled data of the training set corresponds to the historical workload sequence and is based on the baseline fatigue level constructed from actual equipment performance degradation records or maintenance logs. Historical operational data is collected continuously at fixed time intervals, with the collection period preset based on the operational characteristics of physical resources. The training set is divided into a training sample set and a validation sample set according to a preset ratio for model training and performance verification. The baseline fatigue level is comprehensively calibrated by the ratio of actual performance parameters to rated performance parameters, the frequency of fault occurrence recorded in the maintenance log, and the maintenance interval duration. The calibration process follows a unified quantitative standard. For example: Collect historical operational data of all physical resources collected hourly over the past year, totaling 1 million records; randomly divide these records into training and validation sets in an 8:2 ratio; when the equipment performance tester measures that the actual rotation speed of a washing machine is lower than 95% of the rated speed, set the label corresponding to that time point and the workload sequence of the previous 24 hours to 85; if there are no performance degradation records and the time since the last maintenance is within the standard cycle, set the label to 10; using the training set, train the LSTM model using the Adam optimizer with mean squared error as the loss function until its loss on the validation set no longer decreases significantly; this is just an example and is not a limitation. S200: Receive internet laundry orders, parse the clothing categories and process requirements included in the laundry orders, and generate process links; the process links include the types and order of physical resources used to complete the laundry orders; Specifically, step S200 includes: S210. Receive internet laundry orders, parse the clothing category information of the garment attributes in the order data, and the service process requirements information for cleaning and treatment methods; receive order data in real time through the interface of the internet laundry platform. The order data is stored in a structured format, including a unique order identifier, user identifier, garment details list, and service requirements fields; use semantic parsing algorithms to extract material attributes and specifications from the garment category information, and decompose the cleaning methods, treatment standards, and special requirements from the service process requirements information to ensure the comprehensiveness and accuracy of the parsed information; For example: Order ID: ORD123456, User ID: USER789, Clothing details: [{Category: Shirt, Material: Cotton, Quantity: 2}, {Category: Trousers, Material: Wool, Quantity: 1}], Service requirements: Fine washing + ironing; Using a rule-based and keyword-based semantic parsing algorithm, the clothing category information is extracted as [Shirt, Trousers], the material as [Cotton, Wool], and the service process requirements as [Fine washing, ironing]; This is just an example and is not a limitation. S220. Based on the process knowledge base, the parsed clothing categories and process requirements are mapped to one or more discrete process steps necessary to complete the order. Each process step is associated with the physical resource type required to execute that step. The process knowledge base is generated through domain expert experience configuration, establishing mapping rules between clothing categories, material attributes, stain types, user service selections, and standardized process steps. The process knowledge base establishes a regular update mechanism, dynamically adjusting the mapping rules based on newly added clothing categories, new processing technologies, and actual production verification results. A unique standardized process step is precisely matched according to the combination of clothing category, material attributes, stain type, and user service selection. Each process step is pre-bound with the required physical resource type code to ensure the compatibility between process steps and physical resources. For example, a rule in the knowledge base states: if the product category is shirts and the service is fine cleaning, then the process steps are: [pre-stain removal, washing, drying, ironing]. This maps the order to four discrete process steps: pre-stain removal, washing, drying, and ironing. The physical resource types associated with these steps are: sorting line, washing machine, dryer, and ironing machine. This is just an example and is not a limitation. S230. Based on the physical dependencies and business logic between each process step, determine the execution order of the process steps and construct the process link; the process link includes the physical resource types and order for completing the laundry order; For example, based on physical dependencies, such as the requirement to wash before drying, and business logic, such as the requirement for pre-stain removal to be done first and ironing to be done last, the process flow is determined as: pre-stain removal - washing - drying - ironing; the physical resource types and order corresponding to this flow are: sorting line - washing machine - dryer - ironing machine; this is just an example and is not a limitation. S300. Based on the process chain, for each process step in the order, select available resources from all physical resources of the physical resource type of the laundry order whose current fatigue index is lower than the safety threshold, and form a candidate resource set for each step. Specifically, step S300 includes: S310. Based on each process step in the process chain, obtain the real-time fatigue index of all physical resources corresponding to the physical resource type associated with each step. For example: Based on the process flow, obtain the fatigue index of all relevant resources: Sorting line: [Sorting line A: 15, Sorting line B: 40]; Washing machine: [Washing machine 001: 65, Washing machine 002: 28, Washing machine 003: 80]; Dryer: [Dryer 01: 50, Dryer 02: 72]; Ironing machine: [Ironing machine 1: 10, Ironing machine 2: 55]; This is just an example and is not a limitation. S320. Safety thresholds are obtained based on the equipment's rated performance parameters, maintenance cycle standards, and historical fault statistics, and are configured separately according to the physical resource type. For example, the safety threshold uses a weighted average method, with the formula: A = α × E + β × W + γ × L; where A represents the safety threshold; E represents the converted value of the rated performance parameters; W represents the converted value of the maintenance cycle; L represents the average fatigue value of historical faults; α, β, and γ are weighting coefficients, and α + β + γ = 1. The weighting coefficients are set by domain experts based on the different physical resource types; for washing machines, α = 0.5, β = 0.3, and γ = 0.2 are set; the converted value of the rated performance parameters is a normalized number of rated load and rated operating time. The values ​​are linearly combined; the maintenance cycle is calculated based on the ratio of the current time to the last maintenance duration to the standard maintenance cycle; the historical failure fatigue average is the arithmetic mean of the fatigue index recorded when all failure events occurred within the last three maintenance cycles; this is only an example and is not a limitation; the equipment's rated performance parameters include rated load, rated operating time, and rated efficiency; the maintenance cycle standard is the maintenance interval recommended by the equipment manufacturer; the historical failure statistics are the average fatigue index at the time of failure within the last three maintenance cycles; the safety threshold is calibrated periodically based on the cumulative operating time of physical resources, the number of maintenance operations, and the effectiveness of failure repair, with the calibration cycle synchronized with the maintenance cycle; for example, a comprehensive calibration is performed every three months; S330. Filter out physical resources whose real-time fatigue index is lower than the corresponding safety threshold and mark them as available resources; compare the real-time fatigue index of the physical resources with the corresponding type of safety threshold, and determine that physical resources whose real-time fatigue index is less than the safety threshold are available resources; assign a unique available identifier to the available resources, which includes the physical resource number, the current fatigue index and the determination result, to ensure the traceability of available resources. For example: compare the real-time fatigue index with the safety threshold; for washing machines, washing machine 001: 65 < 70 (available), washing machine 002: 28 < 70 (available), washing machine 003: 80 > 70 (unavailable); assign an identifier to the available resource washing machine 001: {Resource ID: washing machine 001, fatigue level: 65, judgment: available}; this is only an example and is not a limitation. S340. Organize the available resources for each process step in sequence and construct a candidate resource set. The candidate resource set includes the available resource identifier, current fatigue index, and equipment operating status information for each step. The candidate resource set is stored in a linked list structure in the order of process steps. Each linked list node contains a process step identifier, a list of available resources, and a resource status description field. The equipment operating status information includes three states: running, idle, and pending maintenance. It is updated in real time through the production management system and synchronized to the candidate resource set to ensure the timeliness of the status information. For example: Organize available resources according to the process steps to form a linked list of candidate resources: Node 1 (pre-stain removal): [Sorting line A, sorting line B]; Node 2 (washing): [washing machine 001, washing machine 002]; Node 3 (drying): [drying machine 01]; Node 4 (ironing): [ironing machine 1, ironing machine 2]; This is just an example and is not a limitation. S400. Based on resource combination optimization logic, select target physical resource combinations for the process link; the optimization logic includes selecting the physical resource with the lowest current fatigue index for high mechanical load links; and selecting the physical resource with the largest difference between the current fatigue index and the average fatigue index of this type of resource for high thermal load or high precision requirements. Specifically, step S400 includes: S410, High Mechanical Load Stages are process stages where the mechanical transmission strength of the equipment reaches a preset mechanical threshold during execution; High Thermal Load Stages are process stages where the thermal output intensity of the equipment reaches a preset thermal threshold during execution; High Precision Requirement Stages are process stages where the precision indicators such as dimensional error and flatness must meet preset precision standards during execution; The preset mechanical thresholds are determined based on the rated mechanical strength of physical resources, material fatigue limits, and long-term operation test data; the preset thermal thresholds are based on the rated heat dissipation capacity of the equipment and the trigger threshold of the thermal protection device; the preset precision standards are formulated with reference to industry-standard garment processing precision specifications and user-required thresholds; the load type is determined by collecting equipment operating parameters during the execution of the process stages and comparing them with the preset thresholds; For example, based on preset thresholds, washing and drying are identified as high mechanical load processes, and ironing is identified as a high precision requirement process; this is only an example and is not a limitation. S420. Calculate the average fatigue index of each physical resource type. The average fatigue index is the arithmetic mean of the current fatigue indices of all physical resources under that type. It is calculated in real time when the resource combination optimization process of each order is started. The calculation scope is all registered physical resources under that physical resource type, including resources in operation and idle state. S430. For high mechanical load processes, select the physical resource with the lowest current fatigue index from the corresponding candidate resource set; sort the physical resources in the candidate resource set in ascending order of current fatigue index, and select the physical resource ranked first as the suitable resource for this process; if there are multiple physical resources with the same current fatigue index and the lowest value, select the physical resource in the idle state according to the priority of equipment operation status. For example, in the washing process, the washing machine with the lowest fatigue level is selected from the candidate resources [washing machine 001, washing machine 002]; this is just an example and is not a limitation. S440. For processes with high heat load or high precision requirements, select the physical resource with the largest difference between the current fatigue index and the average fatigue index of the corresponding type of resource from the corresponding candidate resource set. For processes with high heat load or high precision, select equipment whose fatigue index deviates significantly from the average level to ensure that these processes with high stability requirements are performed by equipment with the best performance, thereby ensuring process quality and reducing the risk of failure. At the same time, avoid a few devices being in an overprotected state for a long time, and promote a balanced distribution of fatigue in the entire resource pool. The difference is the absolute difference between the current fatigue index of the physical resource and the average fatigue index of the corresponding type. The difference is calculated using absolute value operation to ensure that the result is non-negative. Sort the physical resources in the candidate resource set in descending order of difference, and select the physical resource with the highest difference as the suitable resource for this process. For example: In the ironing process, the average fatigue level of the ironing machine is calculated to be 32.5; the absolute difference between each resource and the average value is calculated to be 22.5, and the value of ironing machine 02 is 22.5; if the differences are the same, then according to the additional rules, the idle resource is selected first, such as ironing machine 01 being idle, so ironing machine 01 is selected; this is only an example and is not a restriction. S450. Integrate the physical resources selected in each process step to form a target physical resource combination that adapts to the process link. After integration, perform conflict verification on the target physical resource combination to check whether each physical resource is occupied by multiple process steps at the same time. If there is a conflict, reselect the adaptable resources. The target physical resource combination is stored in a matrix structure, with the row dimension being the process step order and the column dimension being the physical resource identifier, current fatigue index, and operating status. For example: forming a target combination: [pre-stain removal: sorting line A, washing: washing machine 002, drying: dryer 01, ironing: ironing machine 1]; performing conflict checks to ensure that these resources are not occupied by other orders at the current moment, and finally storing the combination in a matrix; this is just an example and is not a limitation.

[0017] S500. Based on the target physical resource combination and fatigue index, use the efficiency mapping table to calculate the estimated time and energy consumption of each process step, and then summarize to obtain the total resource calculation result. Specifically, step S500 includes: S510, the efficiency mapping table, establishes a mapping relationship between historical time and energy consumption data of each process step based on different physical resource types and different fatigue index ranges. Through statistical analysis of historical operating data of each process step under different physical resource types and different fatigue index ranges, a regression analysis algorithm is used to establish the mapping relationship between time and energy consumption and physical resource type and fatigue index. For example, the efficiency mapping table is set to be updated monthly, or triggered when the accumulated new historical data exceeds 10% of the original data volume, to ensure that the mapping relationship can reflect the latest changes in equipment performance in a timely manner. During updates, a sliding window model is used to gradually eliminate the oldest historical data and synchronously replace outdated mapping relationships. This is only an example and is not a limitation. For example: for a washing machine in the fatigue range [20, 40), the unit time consumption parameter is 30 minutes / piece, and the unit energy consumption parameter is 0.8kWh / piece; S520. Based on the type of each physical resource in the target physical resource combination and the real-time fatigue index, query the efficiency mapping table to obtain the unit time consumption parameter and unit energy consumption parameter of the corresponding process step. S530. Based on the processing volume requirements of each process step, calculate the estimated time and energy consumption of each process step; and according to the execution order of the process link, summarize the estimated time and energy consumption of all process steps to obtain the total resource calculation result of the laundry order; the total resource calculation result includes the total time, total energy consumption and the sub-item resource consumption data of each process step.

[0018] This invention provides another technical solution: an Internet laundry order resource accounting system based on LSTM, which includes: a resource monitoring module, a process analysis module, a resource screening module, a resource scheduling module, and a resource accounting module. The resource monitoring module is used to build an LSTM model. It takes the historical workload sequence of each physical resource as input and outputs the real-time fatigue index of each physical resource. The physical resources include washing machines, dryers, ironing machines and sorting lines. The process analysis module is used to receive online laundry orders, analyze the clothing categories and process requirements included in the laundry orders, and generate process links; the process links include the types and order of physical resources used to complete the laundry order; The resource filtering module is used to filter available resources with a current fatigue index below the safety threshold from all physical resources of the physical resource type in the laundry order for each process link in the order, based on the process link, and form a candidate resource set for each link. The resource scheduling module is used to select target physical resource combinations for the process link based on resource combination optimization logic. The optimization logic includes selecting the physical resource with the lowest current fatigue index for high mechanical load links and selecting the physical resource with the largest difference between the current fatigue index and the average fatigue index of that type of resource for high thermal load or high precision requirements. The resource accounting module is used to calculate the estimated time and energy consumption of each process step based on the target physical resource combination and fatigue index, using an efficiency mapping table, and then summarizing the results to obtain the total resource accounting result.

[0019] Based on the above, the resource monitoring module includes a sequence construction unit and an LSTM model unit; The sequence construction unit is used to collect historical operation data and construct a set of parameters arranged in chronological order to represent the historical working status of each physical resource. The parameters include the continuous working time of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and the equipment start-up and shutdown events. The LSTM model unit consists of an input layer, an LSTM hidden layer, and an output layer. The input layer receives multidimensional time series data of historical workload sequences. The LSTM hidden layer learns long-term dependencies and dynamic patterns in the workload sequences. The output layer maps the states of the LSTM hidden layer to a normalized fatigue index.

[0020] Based on the above, the process analysis module includes an information analysis unit and a process link unit; The information parsing unit is used to receive online laundry orders, parse the clothing category information of the clothing attributes in the order data, and the service process requirements information of cleaning and treatment methods; The process link unit is used to map the parsed clothing categories and process requirements to one or more discrete process steps necessary to complete the order, based on a pre-built process knowledge base. Each process step is associated with the physical resource type necessary to execute that step. The execution order of the process steps is determined according to the physical dependencies between them and the business logic, thus constructing the process link.

[0021] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for calculating internet laundry order resources based on LSTM, characterized in that: The method includes: S100. Construct an LSTM model. The LSTM model takes into account the historical workload sequence of each physical resource and outputs the real-time fatigue index of each physical resource. The physical resources include washing machines, dryers, ironing machines, and sorting lines. S200: Receive an internet laundry order, parse the clothing categories and process requirements included in the laundry order, and generate a process link; the process link includes the physical resource types and order for completing the laundry order; S300. According to the process chain, for each process step in the order, select available resources whose current fatigue index is lower than the safety threshold from all physical resources of the physical resource type of the laundry order, and form a candidate resource set for each step. S400. Based on resource combination optimization logic, select target physical resource combinations for the process link; the optimization logic includes: for high mechanical load links, select the physical resource with the lowest current fatigue index; for high thermal load or high precision requirement links, select the physical resource with the largest difference between the current fatigue index and the average fatigue index of that type of resource. S500. Based on the target physical resource combination and fatigue index, use the efficiency mapping table to calculate the estimated time and energy consumption of each process step, and then summarize to obtain the total resource calculation result.

2. The method for calculating internet laundry order resources based on LSTM according to claim 1, characterized in that: Step S100 includes: S110. The historical workload sequence is a set of parameters arranged in chronological order and representing the historical working status of each physical resource; the parameters include the continuous working time of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and equipment start-up and shutdown events. S120. The fatigue index is a normalized value, representing the expected performance degradation and potential failure risk level of the physical resource due to accumulated workload; the index is negatively correlated with the real-time operating efficiency of the equipment and positively correlated with the urgency of required maintenance. S130. The LSTM model includes an input layer, an LSTM hidden layer, and an output layer; the input layer is used to receive multidimensional time series data of the historical workload sequence; the LSTM hidden layer is used to learn long-term dependencies and dynamic patterns in the workload sequence; the output layer is used to map the state of the LSTM hidden layer to the fatigue index. S140. Using historical operation data collected during historical operations as a training set, the LSTM model is trained in a supervised learning manner; the label data of the training set is a baseline fatigue level corresponding to the historical workload sequence and constructed based on the actual performance degradation records or maintenance logs of the equipment.

3. The method for calculating internet laundry order resources based on LSTM according to claim 1, characterized in that: Step S200 includes: S210. Receive internet laundry orders, parse the clothing category information of the clothing attributes in the order data, and the service process requirements information of cleaning and treatment methods; S220. Based on the process knowledge base, the parsed clothing categories and process requirements are mapped to one or more discrete process steps necessary to complete the order; each process step is associated with the physical resource type necessary to execute the step; the process knowledge base is generated through domain expert experience configuration to establish mapping rules between clothing categories, material attributes, stain types, user service selections and standardized process steps; S230. Based on the physical dependencies and business logic between each process step, determine the execution order of the process steps and construct the process link; the process link includes the physical resource types and order for completing the laundry order.

4. The method for calculating internet laundry order resources based on LSTM according to claim 1, characterized in that: Step S300 includes: S310. Based on each process step in the process chain, obtain the real-time fatigue index of all physical resources corresponding to the physical resource type associated with each step. S320. The safety threshold is obtained based on the rated performance parameters of the equipment, maintenance cycle standards, and historical fault statistics of physical resources, and is configured separately according to the type of physical resource. S330. Filter out physical resources whose real-time fatigue index is lower than the corresponding safety threshold, and mark the physical resources as available resources; S340. According to the process steps, organize the available resources of each step and construct a candidate resource set; the candidate resource set includes the available resource identifier, current fatigue index and equipment operating status information of each step.

5. The method for calculating internet laundry order resources based on LSTM according to claim 1, characterized in that: Step S400 includes: S410, the high mechanical load stage is a process stage in which the mechanical transmission strength of the equipment reaches a preset mechanical threshold during execution; the high thermal load stage is a process stage in which the thermal output intensity of the equipment reaches a preset thermal threshold during execution; the high precision requirement stage is a process stage in which the precision indicators such as dimensional error and flatness need to meet preset precision standards during execution. S420. Calculate the average fatigue index of each physical resource type, wherein the average fatigue index is the arithmetic mean of the current fatigue indices of all physical resources under that type. S430. For high mechanical load links, select the physical resource with the lowest current fatigue index from the corresponding candidate resource set; S440. For processes with high heat load or high precision requirements, select the physical resource with the largest difference between the current fatigue index and the average fatigue index of the corresponding type of resource from the corresponding candidate resource set. S450. Integrate the physical resources selected in each process step to form a target physical resource combination that adapts to the process link.

6. The method for calculating internet laundry order resources based on LSTM according to claim 1, characterized in that: Step S500 includes: S510. The efficiency mapping table establishes a mapping relationship between different physical resource types and different fatigue index ranges and the historical time and energy consumption data of each process step. S520. Based on the type of each physical resource in the target physical resource combination and the real-time fatigue index, query the efficiency mapping table to obtain the unit time consumption parameter and unit energy consumption parameter of the corresponding process step. S530. Based on the processing volume requirements of each process step, calculate the estimated time and energy consumption of each process step; and according to the execution order of the process link, summarize the estimated time and energy consumption of all process steps to obtain the total resource calculation result of the laundry order; the total resource calculation result includes the total time, total energy consumption and the sub-item resource consumption data of each process step.

7. An internet-based laundry order resource accounting system based on LSTM, characterized in that: The system includes: a resource monitoring module, a process analysis module, a resource screening module, a resource scheduling module, and a resource accounting module; The resource monitoring module is used to construct an LSTM model, input the historical workload sequence of each physical resource, and output the real-time fatigue index of each physical resource; the physical resources include washing machines, dryers, ironing machines, and sorting lines; The process analysis module is used to receive internet laundry orders, analyze the clothing categories and process requirements included in the laundry orders, and generate a process link; the process link includes the physical resource types and order for completing the laundry order; The resource filtering module is used to filter available resources with a current fatigue index lower than the safety threshold from all physical resources of the physical resource type of the laundry order for each process link in the order, based on the process link, to form a candidate resource set for each link. The resource scheduling module is used to select target physical resource combinations for the process link based on resource combination optimization logic. The optimization logic includes selecting the physical resource with the lowest current fatigue index for high mechanical load links and selecting the physical resource with the largest difference between the current fatigue index and the average fatigue index of that type of resource for high thermal load or high precision requirements. The resource accounting module is used to calculate the estimated time and energy consumption of each process step based on the target physical resource combination and fatigue index, using an efficiency mapping table, and then summarizing the results to obtain the total resource accounting result.

8. The Internet laundry order resource accounting system based on LSTM according to claim 7, characterized in that: The resource monitoring module includes a sequence construction unit and an LSTM model unit; The sequence construction unit is used to collect historical operation data and construct a set of parameters arranged in chronological order to represent the historical working status of each physical resource; the parameters include the continuous working time of the equipment, the type and quantity of orders processed, the intensity level of the process executed, and the equipment start-up and shutdown events; An LSTM model unit includes an input layer, an LSTM hidden layer, and an output layer. The input layer is used to receive multidimensional time series data of the historical workload sequence. The LSTM hidden layer is used to learn long-term dependencies and dynamic patterns in the workload sequence. The output layer is used to map the state of the LSTM hidden layer to a normalized fatigue index.

9. The Internet laundry order resource accounting system based on LSTM according to claim 7, characterized in that: The process analysis module includes an information analysis unit and a process link unit; The information parsing unit is used to receive online laundry orders, parse the clothing category information of the clothing attributes in the order data, and the service process requirements information of cleaning and treatment methods; The process link unit is used to map the parsed clothing categories and process requirements to one or more discrete process steps necessary to complete the order, based on a pre-built process knowledge base. Each process step is associated with the physical resource type necessary to execute that step. Based on the physical dependencies and business logic between each process step, the execution order of the process steps is determined, and the process link is constructed.

Images