Multi-process collaborative scheduling optimization method for pomegranate peeling production line

Abstract

The invention relates to the field of manufacturing industry, in particular to a pomegranate peeling production line-oriented multi-process collaborative scheduling optimization method, which comprises the following steps of: acquiring and processing a first business data stream from an internet of things interface, and inputting historical sensor data into a business prediction model to generate a second business data stream; generating a third service data stream by using the data of the multi-mode sensor array; constructing a multi-target business cost function through a business decision optimization engine; starting an iterative optimization business simulation process to obtain an optimized scheduling scheme; and generating a corresponding digital production work order according to the optimized scheduling scheme. According to the method, equipment state, equipment health degree cost and quality indexes are analyzed through fusion of three business data streams and a multi-objective cost function construction method, comprehensive commercial value evaluation is generated, iterative optimization simulation and closed-loop optimization are started when conditions are met, scheduling optimization of timeliness and value fluctuation in perishable material scheduling is achieved, and the scheduling efficiency of perishable materials is improved. And the commercial value creation and self-adaptive scheduling capability of production is improved.

Description

Technical Field

[0001] The present invention relates to the field of food manufacturing, and in particular to a multi-process collaborative scheduling optimization method for a pomegranate peeling production line. Background Art

[0002] In modern manufacturing, production scheduling is a core process, and its efficiency directly determines a company's production costs and market competitiveness. Currently, companies widely use Manufacturing Execution Systems (MES) to manage production processes. One of the core functions of this system is job shop scheduling, which involves rationally arranging the processing sequence of various production tasks within limited resources to achieve specific optimization goals, such as minimizing total completion time or maximizing equipment utilization. Existing technologies, such as flow shop scheduling methods based on MES systems, are suitable for production lines with fixed process flows, such as food processing and electronic assembly. In these scenarios, scheduling systems typically use heuristic algorithms based on fixed rules to generate production plans. Tasks are generally assigned to single processes or local links based on preset static parameters, achieving a certain degree of automated production management.

[0003] While the aforementioned scheduling methods based on MES and heuristic algorithms have been widely adopted, they also exhibit significant inherent drawbacks when addressing complex and dynamic production environments. Traditional scheduling methods often generate fixed plans based on static models established before production begins. However, dynamic events such as equipment failures, raw material batch quality fluctuations, and urgent order insertions frequently occur in real-time. Static plans are unable to respond to these disruptions in real time, leading to extensive waiting or downtime in subsequent processes and overall production line inefficiency. Heuristic algorithms often focus solely on the local optimum of a single process. For example, the shortest processing time priority rule prioritizes the fastest tasks in the current process, but this can lead to long waits in downstream processes, creating new bottlenecks. Furthermore, scheduling decisions made by each process are independent, lacking a coordinated mechanism to examine the overall situation, making it difficult to achieve an optimal overall completion time for the production line. Existing scheduling models are generally designed for general manufacturing and fail to fully consider the process constraints of specific industries. For example, in food processing, materials are generally perishable. Long wait times not only affect efficiency but also directly lead to reduced product quality and increased waste.

[0004] In summary, existing technologies generally have problems such as low overall production line efficiency, slow response to abnormal events, and difficulty in balancing efficiency and product quality due to the lack of dynamic and collaborative optimization capabilities among multiple processes.

[0005] Therefore, a multi-process collaborative scheduling optimization method for pomegranate peeling production line was proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-process collaborative scheduling optimization method for a pomegranate peeling production line to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-process collaborative scheduling optimization method for a pomegranate peeling production line, comprising: Collect and process the first business data stream reflecting the physical status of equipment and materials in the production line in real time from the Internet of Things interface; Inputting historical sensor data into a business prediction model to generate a second business data stream, wherein the second business data stream includes an expected performance degradation cost of the equipment and an expected commercial value decay trajectory of the material batch; Generating a third business data stream including composite quality indicators of the material batch by performing product value assessment on data from the multimodal sensor array; inputting the first business data stream, the second business data stream, and the third business data stream into a business decision optimization engine to construct a multi-objective business cost function including a perishable business cost item; initiating an iterative optimization business simulation process based on the multi-objective business cost function to obtain an optimized scheduling solution; The optimized scheduling plan is used to generate corresponding digital production work orders, which are sent to processing operation nodes through the interface with the manufacturing execution system to control the production line, conduct closed-loop management of the production line, and track business performance.

[0008] Preferably, the specific implementation process of collecting and processing the first service data flow includes: Through the sensor group associated with each processing unit in the production line, equipment layer data containing equipment operating parameters and working status identification are periodically obtained; at the same time, through the non-contact sensors arranged on the material flow path between each processing unit, inventory data representing the number of work-in-progress between processes is obtained; at the data processing node, the collected equipment layer data and the inventory data are subjected to data cleaning, format standardization and timestamp alignment operations to obtain the first business data stream.

[0009] Preferably, the specific implementation process of generating the second service data flow includes: Extract feature vectors that are strongly correlated with the health status of the equipment from historical sensor data, use the feature vectors to train a business prediction model, establish a mapping relationship between equipment performance degradation and historical operating time, and generate the expected performance degradation cost of the equipment; extract records containing the initial quality of the material, ambient temperature and humidity, and waiting time of each process from the historical data, train a kinetic model to describe the attenuation law of material quality indicators with time and environmental changes, and obtain the expected commercial value attenuation trajectory of the material batch; combine the expected performance degradation cost with the expected commercial value attenuation trajectory to form the second business data stream.

[0010] Preferably, the specific implementation process of generating the third service data flow includes: An optical sensor is used to obtain image data reflecting the appearance characteristics of the material batch; a spectral sensor is used to obtain spectral data reflecting the internal components of the material batch; appearance feature parameters are extracted from the image data, and internal component feature parameters are extracted from the spectral data; the appearance feature parameters are combined with the internal component feature parameters through a preset weighted fusion algorithm to obtain a composite quality indicator that quantitatively characterizes the current comprehensive quality of the material batch; the composite quality indicator is encapsulated to constitute the third business data stream.

[0011] Preferably, the specific implementation process of constructing the multi-objective business cost function includes: The first business data stream, the second business data stream and the third business data stream are input into the business decision optimization engine, and based on the equipment operating parameters and working status identifiers in the first business data stream, a first cost component for measuring the order delivery cycle is constructed; based on the inventory data in the first business data stream, a second cost component for measuring the inventory holding cost and capital occupation is constructed; by taking the real-time composite quality index provided by the third business data stream as the initial value and according to the expected business value decay trajectory provided by the second business data stream, the material quality degradation caused by waiting and processing under different scheduling schemes is calculated, and the quality degradation is converted into direct sales revenue loss, to construct the perishable business cost item as the third cost component; based on the expected performance degradation cost of the equipment in the second business data stream, a fourth cost component for measuring the impact of the scheduling scheme on the long-term asset value of the equipment is constructed; and by weighted combination of the first cost component, the second cost component, the third cost component and the fourth cost component, the multi-objective business cost function is formed.

[0012] Preferably, the specific implementation process of obtaining the optimized scheduling solution includes: An initial decision set including production scheduling solutions is constructed, and the expected comprehensive commercial value of each production scheduling solution in the set is quantitatively evaluated based on the multi-objective business cost function; an iterative optimization business simulation process oriented towards overall commercial profit is initiated; the initial decision set is globally explored, and by adjusting the allocation sequence of production resources and the processing priority of material batches, macro-operation strategies are generated and evaluated to identify high-potential profit intervals; local optimization is performed on the production scheduling solutions within the profit interval, and by fine-tuning the processing time window and reordering adjacent processes, marginal profit growth points are discovered and locked in; the optimized production scheduling solution is fed back to the initial decision set to improve the commercial value benchmark of the overall solution; the iterative process continues and converges when the expected profit gain brought about by the adjustment is lower than the preset commercial sensitivity threshold, and the production scheduling solution with the highest expected comprehensive commercial value is determined as the optimized scheduling solution.

[0013] Preferably, the specific implementation process of generating the digital production work order includes: Analyze the optimized scheduling plan to extract the specific processing sequence, start time and estimated completion time of each material batch at each processing operation node; based on the extracted information, generate a structured digital production work order containing a unique job identifier, material batch information and control instructions for each processing operation node; transmit the digital production work order to the manufacturing execution system in real time through the interface, and the system issues it to the corresponding processing operation node to perform automated control; the structured digital production work order includes data fields for recording and returning the actual completion time, material consumption and quality inspection results of each node for subsequent closed-loop management and business performance tracking.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. At the operational decision-making level, by incorporating operational cost indicators such as expected equipment performance degradation costs and inventory deterioration into decision-making, scheduling shifts from simple cost control to maximizing overall future profits. This dynamic operational decision-making can proactively mitigate potential economic losses, transforming risk management from a post-remediation approach to a preventative one, significantly enhancing the company's operational resilience and profitability.

[0015] 2. At the production operations level, by integrating real-time data and multi-stage optimization simulation, we've achieved a leap from static planning to dynamic, adaptive scheduling. The system intelligently responds to real-time changes in equipment, materials, and orders, automatically generating optimal solutions and minimizing ineffective waiting and downtime, significantly improving production line operational efficiency, flexibility, and market responsiveness.

[0016] 3. At the product value management level, by introducing perishable commercial cost items, combined with real-time value assessment, and using the product's value window as the core scheduling basis, the system prioritizes materials with high value decay risk, ensuring they are processed and sold at peak value. This precise management of the value of perishables throughout their lifecycle reduces supply chain losses at the source. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a multi-process collaborative scheduling optimization method for a pomegranate peeling production line proposed in an embodiment of the present invention; Figure 2 A flowchart for generating a multi-objective business cost function is proposed for an embodiment of the present invention; Figure 3 A flowchart of an optimized scheduling solution is proposed for an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] See also Figures 1 to 3 The present invention relates to a multi-process collaborative scheduling optimization method for a pomegranate peeling production line, and the specific implementation process is as follows: Collect and process the first business data stream reflecting the physical status of equipment and materials in the production line in real time from the Internet of Things interface; Inputting historical sensor data into a business prediction model to generate a second business data stream, wherein the second business data stream includes an expected performance degradation cost of the equipment and an expected commercial value decay trajectory of the material batch; Generating a third business data stream comprising real-time composite quality indicators of the material batch by performing product value assessment on data from the multimodal sensor array; inputting the first business data stream, the second business data stream, and the third business data stream into a business decision optimization engine to construct a multi-objective business cost function including a perishable business cost item; initiating an iterative optimization business simulation process based on the multi-objective business cost function to obtain an optimized scheduling solution; The optimized scheduling plan is used to generate corresponding digital production work orders, which are sent to processing operation nodes through the interface with the manufacturing execution system to control the production line, conduct closed-loop management of the production line, and track business performance.

[0020] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0021] Example 1

[0022] The embodiment of the present application discloses a multi-process collaborative scheduling optimization method for a pomegranate peeling production line. In the production line of a pomegranate peeling factory A, the data stream collected by the sensor is processed. Figure 1 The specific steps of the method proposed in the present invention include: S1, collecting and processing a first business data stream from an Internet of Things interface; S2, inputting historical sensor data into a business prediction model to generate a second business data stream; S3, generating a third business data stream from data of a multimodal sensor array; S4, constructing a multi-objective business cost function through a business decision optimization engine; S5, initiating an iterative optimization business simulation process to obtain an optimized scheduling plan; S6, generating a corresponding digital production work order for the optimized scheduling plan.

[0023] Furthermore, a first business data stream reflecting the physical status of equipment and materials in the production line is collected and processed in real time from the IoT interface; corresponding to the above-mentioned step S1; the specific implementation process includes: Equipment-level and inventory data are periodically acquired through a sensor suite associated with each processing unit in the production line (e.g., the cleaning machine and granulator). This suite includes a multi-axis vibration accelerometer (sampling at 1kHz) and a temperature sensor mounted on the granulator's main motor to monitor equipment health. Furthermore, high-precision weighing belt feeders and non-contact ultrasonic level sensors located on the top of the hoppers on the conveyor belts between processing units capture inventory data representing the number of work-in-progress products between processes.

[0024] All sensor data is aggregated to an edge computing node via industrial Ethernet. At this node, the collected data undergoes real-time cleaning, format standardization, and timestamp alignment. Data cleaning uses the 3σ criterion to remove outliers and fills missing values ​​with neighboring point interpolation. Format standardization converts all data into a JSON format that includes device ID, data type, value, and timestamp. For example, vibration sensor data from a peeler is formatted as: {deviceID: Peeler_01, dataType: Vibration_RMS, value: 0.35, timestamp: 2025-07-10T08:30:01.123Z}; inventory data from a raw material hopper is formatted as: {deviceID: Hopper_01, dataType: Inventory_Ton, value: 2.5, timestamp: 2025-07-10T08:30:05.456Z}. Timestamp alignment is based on a central clock source, ensuring that data from different sources are comparable in time, thereby forming a structured, highly reliable first-line business data stream.

[0025] By clearly defining sensor group deployment strategies and data processing flows for both the equipment layer and material flow paths, the real-time, accurate, and consistent nature of primary business data flows is ensured. This overcomes the data silos and information delays inherent in traditional production environments, often caused by diverse data sources, inconsistent formats, and time synchronization. This provides reliable and accurate input for upper-level optimization decisions, significantly improving the analytical quality of subsequent business decision-making optimization engines and the practical feasibility of scheduling solutions.

[0026] Furthermore, the historical sensor data is input into the business prediction model to generate a second business data stream, which includes the expected performance degradation cost of the equipment and the expected commercial value decay trajectory of the material batch; corresponding to the above step S2; the specific implementation process includes: The first step is to quantify the cost of equipment health, which is used to represent the monetized impact of different scheduling schemes on the physical lifespan and long-term health of the equipment. Feature vectors strongly correlated with the equipment's health status are extracted from historical sensor data, including vibration, temperature, and power. This is achieved by first extracting a preliminary feature set from each sensor's raw time series data. This set includes the vibration signal's time domain characteristics (such as root mean square (RMS), peak-to-peak value, margin factor, kurtosis, and skewness), frequency domain characteristics (such as the energy fraction and spectral entropy around the equipment's natural frequency and its multiples obtained through fast Fourier transform), and real-time readings and statistical characteristics of the temperature and power signals (such as mean and standard deviation). Subsequently, to reduce data redundancy and improve model performance, a two-stage feature screening strategy was adopted. In the first stage, weakly correlated features with correlations below a preset threshold (e.g., 0.4) with historical equipment failure records or remaining useful life (RUL) labels were removed through Pearson correlation coefficient analysis. In the second stage, recursive feature elimination (RFE) was used on these features, combined with a benchmark model (e.g., random forest) to select an optimal feature subset (e.g., 10 features) that maximized their contribution to the prediction target. This ultimately formed an N-dimensional feature vector that was used as input to subsequent models. This feature vector was then used to train a commercial prediction model for equipment health index, a long short-term memory (LSTM) network. The network topology consisted of one input layer, two hidden LSTM layers, and a fully connected output layer. Each LSTM layer contained 128 neurons and used the hyperbolic tangent function as the activation function. The hyperbolic tangent function was chosen because its inherent S-shaped curve characteristics effectively fit the observed nonlinear decay of equipment health status over its lifecycle. By analyzing a large amount of historical equipment failure data, we found that equipment health decay follows a three-stage pattern: slow, accelerated, and then steep. The S-shaped curve of the tanh function approximates this physical process: the flat region of the function corresponds to the stable operation period of the equipment, the inflection point region corresponds to the period of accelerated performance decay, and the steep region corresponds to the period approaching failure. This ensures that the neural network learning process closely aligns with the actual physical decay patterns of the equipment, improving the physical interpretability of RUL prediction. This not only improves prediction accuracy but, more importantly, provides clear physical meaning to the prediction results, making them easier for maintenance personnel to understand and make decisions. Compared to activation functions such as ReLU, this improves the accuracy of equipment life prediction. To prevent overfitting, a dropout layer with a dropout rate of 0.2 is placed between the two LSTM layers. The model is trained using historical data covering the entire life cycle of the equipment from operation to failure. Key hyperparameters are set as follows: learning rate of 0.001, batch size of 64, and number of training epochs of 100.The RUL prediction value output by the model is converted into the expected performance degradation cost quantified in monetary units, which is used to quantify the potential impact of different scheduling schemes on the long-term asset value of the equipment in business decisions. At the same time, the confidence interval of its prediction (for example, ±5%) is recorded. This interval information will be used for subsequent more complex risk sensitivity analysis or robustness optimization. The estimation method is: multiply the equipment replacement cost by the health function based on the exponential function, and then multiply it by the risk adjustment coefficient, where the health function is equal to one minus the natural constant e raised to the power of negative λt, λ is the health decay rate parameter, and its value is equal to the negative natural logarithm of the residual value rate divided by the rated service life. The residual value rate is generally 5% to 10%, and t is the current operating time; the risk adjustment coefficient is equal to one plus the product of the sensitivity parameter α and the ratio of the actual failure rate to the expected failure rate minus one. This coefficient can be dynamically determined through historical data analysis based on factors such as the equipment criticality level, historical maintenance costs, and spare parts availability. In this embodiment, α is 0.2 to 0.5.

[0027] Subsequently, an expected commercial value decay trajectory for the material batch is generated. Commercial value decay quantifies the decline in physical quality. Historical data, including records of initial material mass, ambient temperature and humidity, and waiting times for each process step, is extracted and various kinetic or statistical models are employed. When quality decay is primarily driven by temperature-sensitive chemical reactions, the Arrhenius equation is preferred to describe the decay rate of key chemical components driven by temperature. In other cases, such as when quality decay manifests as a probabilistic failure process (e.g., packaging damage, microbial contamination reaching a threshold), or when more complex decay patterns require fitting, statistical models such as the Weibull distribution can be employed. The Arrhenius equation is used to model the decay rate of pomegranate quality (e.g., total soluble solids content) and describes the relationship between the quality decay rate constant k and the ambient absolute temperature T. Specifically, the rate constant k is equal to the pre-exponential factor A multiplied by the natural constant e, which is the negative activation energy Ea divided by the product of the ideal gas constant R and the absolute temperature T. The pre-exponential factor A and the activation energy Ea parameters in the model are obtained by fitting historical experimental data using nonlinear least squares methods. Based on this model, we can predict the quality degradation curve of any batch of materials under different scheduling scenarios, corresponding to different waiting times and ambient temperatures, namely the expected commercial value decay trajectory. Ultimately, the expected performance degradation cost and the expected commercial value decay trajectory are combined to form a second business data stream containing clear prediction basis and quantitative results.

[0028] By training a business forecasting model using historical data, this approach quantifies future equipment and business risks into specific, comparable cost terms. Specifically, by employing established dynamic models such as the Arrhenius equation to describe the decay of material value over time and environmental changes, the system can foresee the direct impact of different scheduling decisions on the commercial value of the final product. This ability to translate future uncertainties into current decision variables enables the system to make more profound economic trade-offs across time, thereby achieving preventative, value-preserving, and intelligent scheduling.

[0029] Furthermore, the data from the multimodal sensor array is used to generate a third business data stream containing real-time composite quality indicators of the material batch through product value assessment; corresponding to the above-mentioned step S3; the specific implementation process includes: First, a first-tier rapid online assessment system was established. Relatively low-cost, fast-response sensors, such as industrial cameras and load cells, were deployed at key production line nodes (e.g., after cleaning and before granulation). This system continuously and rapidly inspected each batch of material passing through, capturing easily analyzable appearance parameters (such as color uniformity, average size, and estimated bruise area ratio) and weight data, generating preliminary, rapid quality assessment results. These results were primarily used to monitor production stability online and quickly screen out batches with obvious appearance defects.

[0030] Next, establish a second-tier precision sampling inspection system. Dedicated bypasses or inspection stations are set up on the production line, deploying the aforementioned high-precision multimodal sensor array (such as hyperspectral imagers and visible / near-infrared spectrometers). This system does not inspect all materials, but rather performs precision sampling inspections based on pre-set strategies. Sampling strategies may include: 1) periodic sampling, such as randomly selecting one batch every hour for in-depth analysis; and 2) triggered sampling. When the first-tier rapid assessment system detects that a batch's quality parameters exceed a preset threshold, the batch is automatically imported into the second-tier system for precision diagnosis.

[0031] Finally, data fusion and model self-calibration are performed. Detailed appearance features (such as precise bruise area) acquired by the second-tier precision detection system are combined with internal compositional features (such as soluble solids content and total acidity) to calculate the precise composite quality index (CQI) for the sampled batch using a weighted fusion algorithm. This weighted fusion algorithm is preferably a linear weighted summation model, multiplying each normalized quality characteristic parameter by its corresponding weight. These products are then summed to produce a comprehensive quality index score. Weights are objectively determined through regression analysis of historical sales data. The specific steps include: 1) collecting quality characteristic parameters and their corresponding market unit prices for historical batches of material; 2) developing a multivariate linear regression model that expresses the market unit price as the sum of the base price and multiple quality characteristics multiplied by their respective regression coefficients; 3) fitting the model to the collected historical data to obtain standardized regression coefficients for each quality characteristic parameter; and 4) normalizing the absolute values ​​of each coefficient to obtain the weights used in the final CQI calculation. More importantly, this precise CQI value and its corresponding rapid assessment parameter serve as new, high-quality training samples, periodically validating and calibrating the predictive model of the first-tier rapid assessment system. This enables the system to continuously improve the accuracy of large-scale, low-cost testing using a small amount of high-precision data, forming an intelligent, adaptive closed-loop quality control system. Ultimately, the composite quality indicator output by this layered system, combining rapid assessment and precise sampling results, constitutes the third service data stream. For example, a complete third-service data stream JSON object example is as follows: {batchID: PG20250710-A01, timestamp: 2025-07-10T08:45:10.500Z, rawFeatures: {bruiseAreaRatio: 0.02, colorUniformity: 0.85, TSS: 14.5, totalAcidity: 1.9}, normalizedFeatures: {bruiseAreaRatio: 0.13, colorUniformity: 0.81, TSS: 0.65, totalAcidity: 0.47}, activeWeights: {bruiseAreaRatio: 0.30, colorUniformity: 0.20, TSS: 0.35, totalAcidity: 0.15}, CQI_score: 0.785, qualityGrade: A}.

[0032] By deploying a multimodal sensor array and combining it with a weighted fusion algorithm, real-time, objective, and multi-dimensional quantification of the current comprehensive quality of material batches is achieved. Using optical sensors to capture appearance characteristics and spectral sensors to capture internal composition characteristics overcomes the subjectivity, partiality, and latency inherent in traditional manual sampling or single-sensor testing. By generating real-time composite quality indicators, this method enables the scheduling system to differentiate processing based on the immediate, real-world value of each batch of materials, such as prioritizing high-value batches, thereby maximizing the utilization of limited production resources and improving the value conversion rate from raw materials to finished products.

[0033] Furthermore, the first business data stream, the second business data stream and the third business data stream are input into a business decision optimization engine to construct a multi-objective business cost function including a perishable business cost item; corresponding to the above step S4; refer to Figure 2 The specific implementation process includes: First, based on the equipment operating parameters and working status identifiers in the first business data stream, a first cost component is constructed to measure the order delivery cycle. This component quantifies the total cost required to complete all orders under a specific scheduling plan. It is calculated by summing the total energy consumption cost and total labor cost of all equipment involved in the plan. Second, based on the inventory data in the first business data stream, a second cost component is constructed to measure inventory holding costs and capital tie-up. This is calculated by integrating the product of the value of the work-in-progress inventory at each buffer node and the holding time, and then multiplying it by the capital holding rate per unit time.

[0034] The key to this invention lies in constructing a perishable commercial cost item as the third cost component. Using the real-time composite quality index (CQI) provided by the third service data stream as the initial value, and based on the expected commercial value decay trajectory provided by the second service data stream, the system calculates the material quality degradation caused by waiting and processing under different scheduling scenarios and converts this quality degradation into direct sales revenue loss. This conversion is achieved using a preset quality-price function, which describes the commercial value of the material as an S-shaped curve. Specifically, the commercial value of the material is equal to the lowest market price plus the added value that increases with quality improvement. This added value is calculated by multiplying the difference between the highest and lowest market prices by an adjustment coefficient. This adjustment coefficient is calculated as one plus the inverse of the natural constant e raised to a specific power. This power is the product of a negative steepness coefficient k and the difference between the current composite quality index (CQI) and the inflection point (CQI_0) where the quality-price relationship is most sensitive. All parameters in the function, such as the highest price, lowest price, inflection point, and steepness coefficient, are objectively determined through nonlinear fitting of historical sales data. The expected equipment performance degradation cost is also constructed as a fourth cost component. This cost item directly uses the expected equipment performance degradation cost generated in the second service data stream to quantify the impact of different scheduling options on the long-term asset value of the equipment. Subsequently, a weighted combination of the first, second, third, and fourth cost components is formed to form the multi-objective service cost function. The weights of each cost component are scientifically determined using the analytic hierarchy process. This process includes: first, establishing a hierarchical model with total cost minimization as the objective layer and the four cost components as the criterion layer; then, constructing a pairwise comparison judgment matrix, in which decision makers perform pairwise comparisons of the importance of each criterion using a 1-9 scale; then, calculating the weight vector for each criterion; and finally, performing a consistency check, ensuring that the consistency ratio is less than 0.1 to ensure the logical consistency of the decision judgment. The resulting weight combination is used to construct the final multi-objective service cost function.

[0035] A comprehensive cost function was constructed that uniformly measures and balances multiple core business objectives. By weighting order delivery cycles, inventory holding costs, and the perishable business costs of innovation, a unified optimization objective was created. This eliminates the need for the optimization engine to pursue optimal performance in isolation for a single technical metric. Instead, it intelligently seeks the balance point within the overall business framework that minimizes overall business costs. This unified quantification of multiple objectives is a key technical prerequisite for truly optimizing business decisions.

[0036] Furthermore, based on the multi-objective business cost function, an iterative optimization business simulation process is started to obtain an optimized scheduling solution; corresponding to the above step S5; refer to Figure 3 The specific implementation process includes: To solve this scheduling problem, this method initiates an iterative optimization business simulation process guided by overall commercial profit. This process utilizes a hybrid metaheuristic algorithm combining a genetic algorithm (GA) and simulated annealing (SA). The system first constructs an initial decision set containing multiple production scheduling solutions, known as the GA population. Each scheduling solution is represented as a global sequence of all pending operations using an operation-based encoding. Subsequently, the system uses the GA to conduct a global exploration, generating offspring solutions in each generation through operations such as crossover and mutation that preserve priority order. After each GA iteration, a simulated annealing local search is initiated for the best individuals (elites) in the population. To prevent the algorithm from prematurely converging on local optima, the selection strategy for elite individuals, in addition to primarily relying on the comprehensive cost function, also incorporates auxiliary evaluation mechanisms designed to promote solution diversity. For example, the Hamming distance between individual solutions in the elite population can be periodically calculated, or a penalty term can be introduced during the selection process to moderately reduce the probability of selecting solutions that are too similar to existing elite individuals. This mechanism helps the algorithm strike a better balance between exploration and exploitation, increasing the likelihood of finding high-quality solutions while maintaining solution diversity to address diverse production scenarios. This improves the algorithm's robustness and practicality. When faced with sudden production changes, the elite pool always contains an adaptable solution, shortening decision response time and enhancing solution quality stability. SA generates neighborhood solutions by reordering two adjacent processes on key equipment, identifying and targeting areas with increasing marginal profit margins. If the SA search result outperforms the original elite individual, it is replaced, and the optimized elite individual then participates in the next generation of crossover and mutation. This iterative process continues, converging when the expected profit gain from the adjustment falls below a preset business sensitivity threshold. This threshold is not a fixed value but is dynamically calculated based on multiple factors, including current production costs, market volatility, and order urgency. A diminishing marginal returns model is established based on historical optimization data. If the profit gain for three consecutive iterations is less than 80% of the current threshold, the sensitivity threshold is automatically lowered to continue the optimization search. When market demand surges, the threshold is raised to enable faster response. This prevents the algorithm from over-iterating in low-value areas while ensuring that optimization opportunities are not missed when high-value opportunities arise. This method achieves intelligent allocation of computing resources. Compared with the fixed threshold method, it reduces some ineffective computing time and improves the profit optimization effect of key orders, truly realizing the intelligent convergence of algorithm resources to focus on commercial value. The production scheduling plan with the highest expected comprehensive commercial value is determined as the optimized scheduling plan. In addition, to ensure the applicability of the algorithm in different production scenarios, this method does not use fixed algorithm parameters. Instead, it introduces experimental design techniques such as the Taguchi method. Through a series of simulation experiments, it systematically searches for the optimal algorithm parameter combination under different production constraints, such as crossover probability and cooling rate, thereby transforming parameter selection from a one-time decision to a repeatable optimization method.

[0037] By distinguishing between global exploration and local optimization, we can effectively avoid falling into the trap of local optimal solutions. The global exploration phase adjusts macro-operation strategies to quickly identify high-potential profit zones, while the local optimization phase conducts fine-tuning within these high-potential zones to discover and target growth points for marginal profits. This combination of coarse and fine tuning ensures the algorithm is both broad and deep, allowing it to steadily converge to a scheduling solution with significantly better commercial value at a reasonable computational cost.

[0038] Furthermore, the optimized scheduling plan is used to generate a corresponding digital production work order, which is then sent to the processing operation node through the interface with the manufacturing execution system to control the production line, perform closed-loop management of the production line, and track business performance. This corresponds to the above-mentioned step S6. The specific implementation process includes: By analyzing the optimized scheduling plan, the specific processing sequence, start time, and estimated completion time for each material batch at each processing operation node are extracted. A structured digital production work order is generated for each processing operation node. This work order is transmitted to the manufacturing execution system in real time via the standardized OPC UA interface and distributed to the corresponding processing node. A JSON example of a digital production work order is as follows: {workOrderID: WO-2025-A-001, batchID: PG20250710-A01, nodeID: Peeler_01, startTime: 2025-07-10T09:00:00Z, estimatedEndTime: 2025-07-10T09:25:00Z, controlCommands: {speed: 1200, temperature: 25}, feedbackFields: {actualEndTime: null, materialConsumed_kg: null, qualityCheckResult: null}}. The feedbackFields field in the work order is used to record and transmit the actual completion time, material consumption, and quality inspection results of each node. These completed work orders with actual execution results form a closed-loop data flow, which is used to track business performance and serve as new training samples for input into business forecasting and dynamic models. Through this periodic model retraining mechanism, the system can adaptively correct and optimize its forecast accuracy. For example, it can adjust equipment reliability parameters based on actual equipment failure frequencies and update value decay coefficients based on actual material losses. This ensures that the decision quality of the entire system continuously improves over time, forming a healthy and adaptive intelligent manufacturing ecosystem.

[0039] By parsing optimized scheduling plans into structured digital production work orders and distributing them through an interface with the manufacturing execution system, the system ensures that optimization decisions are executed accurately and automatically. More importantly, by including data fields in the work orders for transmitting actual execution results, a mechanism for continuous learning and self-improvement is established. This transmitted data is not only used for performance tracking but also as new training samples to iteratively update and optimize the prediction model. This allows the entire system's prediction accuracy and decision quality to continuously improve over time, forming a healthy and adaptive intelligent manufacturing ecosystem.

[0040] This method integrates real-time physical data streams, predictive business data streams, and real-time product quality data streams, and feeds these into a multi-objective optimization engine that incorporates perishable business costs. This creates a decision-making framework capable of dynamically balancing multiple, sometimes conflicting, business objectives, including production efficiency, equipment depreciation, inventory costs, and product commercial value decay. By converting optimization plans into digital work orders and implementing closed-loop management, this approach not only enhances the intelligence and adaptability of production scheduling but, more importantly, directly links production operations to the company's core financial objectives, thereby maximizing overall commercial profits in the production environment of perishable products.

[0041] Example 2

[0042] In the production line of pomegranate peeling factory B, the specific implementation method for obtaining three business data flows is as follows: First, the system collects and processes the first business data stream in real time from the IoT interface. The system periodically acquires device-level data and inventory data through sensor groups associated with each processing unit in the production line. For example, the sensor group in the cleaning processing unit collects operating parameters and working status indicators such as motor power of 52kW and speed of 1200rpm; the sensor group in the peeling processing unit collects data such as vibration acceleration. Simultaneously, the ultrasonic liquid level sensor in the raw material buffer detects 2.8 tons of pomegranate raw material inventory, and the photoelectric sensor in the post-cleaning buffer detects 2.1 tons of clean fruit inventory. At the data processing node, the system cleans, standardizes the format, and aligns the timestamps of these data to form the first business data stream.

[0043] The system then inputs historical sensor data into the business prediction model to generate a second business data stream. To generate the expected performance degradation cost of the equipment, the system extracts feature vectors from the historical data and trains a long short-term memory (LSTM) model, predicting the remaining useful life of the threshing machine to be 2847.5 hours. Based on the equipment health cost quantification method, for a threshing machine with a replacement cost of 20,000 yuan, a rated total service life of 10,000 hours, and a current operating time of 7152.5 hours, the following calculation can be performed: First, the residual value is set to 8% and the health decay rate parameter λ is calculated to be approximately 0.00025257. Based on this rate and operating time, the health function is calculated to be 0.8358. Furthermore, if the current actual failure rate is roughly consistent with the expected failure rate, the risk adjustment factor is 1. Finally, by multiplying the equipment replacement cost, the health function, and the risk adjustment factor, the expected performance degradation cost of the equipment is estimated to be approximately 16,716 yuan. To generate the expected commercial value decay trajectory of a material batch, the system extracts historical data to train a kinetic model to describe the quality decay law. The model uses the Arrhenius equation and uses nonlinear least squares fitting of historical experimental data to determine the key parameters of the model: activation energy (Ea) of 75 kJ / mol and pre-exponential factor (A) of 1.87 x 10 11 h -1 This calculation shows that, at the current temperature of 24.5 degrees Celsius, the quality degradation rate of material batch PG20250709_B_002847 is 0.0142 per hour. This degradation trajectory will be used in subsequent cost calculations. Ultimately, the calculated expected performance degradation cost is combined with this expected commercial value degradation trajectory to form the second business data stream.

[0044] First, material batch PG20250709_B_002847 passed through the first-level rapid online assessment system. An industrial camera initially assessed its color uniformity as 0.75, and estimated the bruise area ratio to be 0.04, slightly above the average for a normal batch. Next, because the estimated bruise area ratio triggered a preset threshold, the system initiated the second-level precision sampling inspection. The batch was directed to the inspection station for in-depth analysis using a hyperspectral imager and visible / near-infrared spectrometer. The results accurately identified the bruise area ratio as 0.038, calculated the color uniformity index as 0.762, and non-destructively predicted the soluble solids content to be 13.8% and the total acidity to be 2.14g / 100g. The system then normalized these precise characteristic parameters. For example, the bruise area ratio is mapped to the interval [0, 0.15], color uniformity is mapped to the interval [0.3, 1.0], soluble solids is mapped to the interval [8.0, 18.0], and total acidity is mapped to the interval [0.5, 3.5], resulting in normalized values ​​of 0.253, 0.660, 0.580, and 0.547, respectively. Based on the weighting configuration determined by regressing historical sales prices (with a weight of 0.25 for bruise area, 0.20 for color uniformity, 0.15 for soluble solids, 0.10 for acidity, and a total weight of 0.30 for other features), a weighted fusion algorithm is used to calculate the current precise composite quality index (CQI) for this material batch: 0.695. This precise CQI value of 0.695 is then encapsulated into the third service data stream. At the same time, this set of data (rapid evaluation parameters and precise CQI values) is stored in the training database for subsequent iterative optimization of the first-level evaluation model to improve its estimation accuracy.

[0045] Example 3

[0046] In the production line of pomegranate peeling factory B, see Figure 2 The specific implementation of inputting the first business data stream, the second business data stream, and the third business data stream into a business decision optimization engine to construct a multi-objective business cost function including a perishable business cost item is as follows: First, the system constructs the first cost component based on the data in the first business data stream. Based on the equipment operating parameters, the cleaning equipment runs at 52 kilowatts for 6 hours, the threshing equipment runs at 43 kilowatts for 5.765 hours, and the packaging equipment runs at 26 kilowatts for 4.5 hours. Based on the industrial electricity price of 0.65 yuan per kilowatt-hour, the total energy cost is 439.98 yuan. Adding the known labor cost of 2,456 yuan, the first cost component (order delivery cycle cost) is 2,895.98 yuan. Next, the system constructs the second cost component based on the real-time inventory data in the first business data stream. Raw material inventory is 2.8 tons at a unit price of 9.2 yuan per kilogram; post-cleaning inventory is 2.1 tons at a unit price of 10.7 yuan per kilogram; and post-threshing inventory is 0.7 tons at a unit price of 48.3 yuan per kilogram. Using an annualized capital holding rate of 8.5% to convert this into hourly holding costs, and integrating the value of work-in-process inventory over the entire scheduling cycle, the second cost component (inventory holding costs) is calculated to be 182.5 yuan.

[0047] The system then constructs the third cost component, the core of the present invention—the perishable commercial cost. Using an initial composite quality index (CQI) of 0.695 and a quality decay rate of 0.0142 per hour, a quality-price sigmoid function is applied. The function's parameters, determined through historical data fitting, are: a high market price of 65 yuan per kilogram, a low market price of 10 yuan per kilogram, an inflection point (CQI_0) of 0.75 (the most sensitive inflection point in the quality-price relationship), and a steepness coefficient (k) of 10. Based on this function, the initial commercial value is first calculated: equal to the low market price of 10 yuan, plus an additional value. This value is the difference between the high and low prices (55 yuan) multiplied by an adjustment coefficient. This adjustment coefficient is calculated by adding the inverse of the natural constant e raised to a specific power (the negative steepness coefficient of 10) and the difference between the initial CQI value of 0.695 and the inflection point of 0.75. The resulting initial commercial value is 30.12 yuan per kilogram. After a total processing time of 5.25 hours, the system, using a comprehensive attenuation model based on the Arrhenius equation, predicted the attenuation of each key internal component affecting the CQI. Subsequently, based on these predicted values, a pre-set weighted fusion algorithm was used to recalculate the final CQI prediction, which dropped to 0.646. The final commercial value was calculated again using the quality-price function, resulting in a value of 14.36 yuan per kilogram. Therefore, the expected direct sales revenue loss due to the quality degradation was 15.76 yuan per kilogram. Based on the batch weight of 680 kilograms, the expected value of the third cost component (perishable commercial cost) was 10,716.8 yuan.

[0048] Next, the system constructs a fourth cost component: the expected performance degradation cost of the equipment. This cost component quantifies the impact of different scheduling options on the long-term asset value of the equipment. Its value is directly derived from the expected performance degradation cost in the second service data stream and is determined to be 16,716 yuan.

[0049] Finally, the system forms a multi-objective business cost function by weighting the four cost components. The weights are determined using the Analytic Hierarchy Process (AHP). In this example, to balance short-term production benefits with long-term asset maintenance, the weights are set as follows: order delivery weighted at 0.3, inventory holding weighted at 0.2, perishability cost weighted at 0.3, and expected equipment performance degradation cost weighted at 0.2. Each cost component is multiplied by its corresponding weight and summed: (2,895.98 yuan multiplied by 0.3) plus (182.5 yuan multiplied by 0.2) plus (10,716.8 yuan multiplied by 0.3) plus (16,716 yuan multiplied by 0.2). The final calculated multi-objective business cost function value for this scheduling scheme is 7,463.534 yuan.

[0050] Example 4

[0051] In the production line of pomegranate peeling factory B, see Figure 3 , based on the multi-objective business cost function, the iterative optimization business simulation process is started to obtain the specific implementation method of the optimized scheduling solution as follows: The system initiates an iterative optimization business simulation process focused on overall business profit. This process utilizes a hybrid metaheuristic algorithm that combines a genetic algorithm (GA) for global exploration with simulated annealing (SA) for local optimization. First, an initial decision set containing production scheduling solutions, known as the GA's initial population, is constructed. The population size is set to 50 individuals. Each scheduling solution uses an operation-based encoding to represent the global sequence of all pending operations. In each GA iteration, the system evaluates the fitness of each scheduling solution based on a multi-objective business cost function, with lower costs associated with higher fitness. A priority-preserving crossover operator (with a crossover probability of 0.8) and an exchange mutation operator (with a mutation probability of 0.05) are used to generate offspring solutions, exploring new, high-potential macro-operational strategies. After each GA iteration, the system initiates a simulated annealing local search for the best individuals (elites) in the population. The SA algorithm, with an initial temperature of 1000 and a cooling rate of 0.95, generates neighborhood solutions by reordering two adjacent processes on key equipment, identifying and targeting growth points for marginal profits. The optimal solution obtained through local optimization is fed back into the genetic algorithm population to replace the original elite individuals. This iterative process continues, and the algorithm converges when the improvement in the optimal solution over multiple generations falls below a preset business sensitivity threshold of 0.1%, or when the preset maximum number of iterations, 500, is reached. Ultimately, the system selects the production schedule with the lowest total business cost discovered during the entire optimization process as the optimal schedule.

[0052] Example 5

[0053] In the production line of pomegranate peeling factory B, the specific implementation method of generating the corresponding digital production work order from the optimized scheduling plan is as follows: The system first analyzes the optimized scheduling plan, extracting the specific processing sequence, start time, and estimated completion time for each material batch at each processing operation node. For example, for batch PG20250709_B_002847, the system analyzes that it is ranked third in the cleaning process, with a start time of 09:30 and an estimated completion time of 09:58. Based on this information, the system generates a structured digital production work order for the corresponding processing operation node, containing a unique job identifier, material batch information, and control instructions. This work order is transmitted to the manufacturing execution system in real time via the OPC UA standardized interface, and then issued by the system to the corresponding processing operation node for automated control. A JSON example of a digital production work order is as follows: {workOrderID: WO-2025-B-1847, batchID: PG20250709_B_002847, nodeID: Cleaner_01, startTime: 2025-07-10T09:30:00Z, estimatedEndTime: 2025-07-10T09:58:00Z, controlCommands: {speed: 520, temperature: 38}, feedbackFields: {actualEndTime: null, materialConsumed_kg: null, qualityCheckResult: null}}.

[0054] The structured digital production work order contains data fields, known as feedbackFields, for recording and transmitting the actual execution results of each node. After the work order is completed, these fields are populated with actual data, such as actual completion time, actual material consumption, and final quality inspection results. These completed work orders with actual execution results form a closed-loop feedback data stream. This data stream is fed back to the business decision optimization engine for performance tracking and accurate cost accounting analysis. More importantly, it serves as new, high-quality training samples for input into the business forecasting model. The system utilizes this real-time feedback data to establish a periodic model retraining mechanism. For example, it updates the mapping of equipment performance degradation based on actual equipment failure frequency and maintenance records, or adjusts the parameters of the value decay trajectory based on actual material loss. Through this mechanism, the system's forecasting accuracy is continuously optimized.

[0055] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line, characterized in that: include: Collect and process the first business data stream reflecting the physical status of equipment and materials in the production line in real time from the Internet of Things interface; Inputting historical sensor data into a business prediction model to generate a second business data stream, wherein the second business data stream includes an expected performance degradation cost of the equipment and an expected commercial value decay trajectory of the material batch; Generating a third business data stream including composite quality indicators of the material batch by performing product value assessment on data from the multimodal sensor array; Inputting the first business data stream, the second business data stream, and the third business data stream into a business decision optimization engine to construct a multi-objective business cost function including a perishable business cost item; Initiating an iterative optimization business simulation process based on the multi-objective business cost function to obtain an optimized scheduling solution; The optimized scheduling plan is used to generate corresponding digital production work orders, which are sent to processing operation nodes through the interface with the manufacturing execution system to control the production line, conduct closed-loop management of the production line, and track business performance.

2. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of collecting and processing the first business data stream includes: periodically obtaining equipment layer data containing equipment operating parameters and working status identification through a sensor group associated with each processing unit in the production line; at the same time, obtaining inventory data representing the number of work-in-progress between processes through non-contact sensors arranged on the material flow path between each processing unit; at the data processing node, performing data cleaning, format standardization and timestamp alignment operations on the collected equipment layer data and the inventory data to obtain the first business data stream.

3. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of generating the second business data stream includes: extracting feature vectors related to the health status of the equipment from historical sensor data, using the feature vectors to train a business prediction model, establishing a mapping relationship between equipment performance degradation and historical operating time, and generating the expected performance degradation cost of the equipment due to operation; extracting records of the initial quality of the material, ambient temperature and humidity, and waiting time of each process from the historical data, training a dynamic model for describing the attenuation law of material quality indicators with time and environmental changes, and obtaining the expected commercial value attenuation trajectory of the material batch; combining the expected performance degradation cost with the expected commercial value attenuation trajectory to form the second business data stream.

4. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of generating the third business data stream includes: using an optical sensor to obtain image data reflecting the appearance characteristics of the material batch; using a spectral sensor to obtain spectral data reflecting the internal components of the material batch; extracting appearance feature parameters from the image data, and extracting internal component feature parameters from the spectral data; combining the appearance feature parameters with the internal component feature parameters through a preset weighted fusion algorithm to obtain a composite quality indicator that quantitatively characterizes the current comprehensive quality of the material batch; encapsulating the composite quality indicator to constitute the third business data stream.

5. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of constructing the multi-objective business cost function includes: inputting the first business data stream, the second business data stream and the third business data stream into the business decision optimization engine, and constructing a first cost component for measuring the order delivery cycle based on the equipment operating parameters and working status identifiers in the first business data stream; constructing a second cost component for measuring inventory holding costs and capital occupation based on the inventory data in the first business data stream; by using the real-time composite quality index provided by the third business data stream as the initial value and according to the expected business value decay trajectory provided by the second business data stream, calculating the material quality degradation caused by waiting and processing under different scheduling schemes, and converting the quality degradation into direct sales revenue loss, constructing the perishable business cost item as the third cost component; constructing a fourth cost component for measuring the impact of the scheduling scheme on the long-term asset value of the equipment based on the expected performance degradation cost of the equipment in the second business data stream; and forming the multi-objective business cost function by weighted combination of the first cost component, the second cost component, the third cost component and the fourth cost component.

6. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of obtaining the optimized scheduling scheme includes: constructing an initial decision set including production scheduling schemes, and quantitatively evaluating the expected comprehensive commercial value of each production scheduling scheme in the set based on the multi-objective business cost function; initiating an iterative optimization business simulation process oriented towards overall commercial profit; conducting a global exploration of the initial decision set, generating and evaluating macro-operation strategies by adjusting the allocation sequence of production resources and the processing priority of material batches, and identifying high-potential profit intervals; conducting local optimization of the production scheduling schemes within the profit interval, and discovering and locking in marginal profit growth points by fine-tuning the processing time window and reordering adjacent processes; feeding back the optimized production scheduling scheme to the initial decision set to improve the commercial value benchmark of the overall scheme; the iterative process continues, and converges when the expected profit gain brought by the adjustment is lower than the preset commercial sensitivity threshold, and the production scheduling scheme with the highest expected comprehensive commercial value is determined as the optimized scheduling scheme.

7. A multi-process collaborative scheduling optimization method for a pomegranate peeling production line according to claim 1, characterized in that: The specific implementation process of generating the digital production work order includes: parsing the optimized scheduling plan to extract the specific processing sequence, start time and estimated completion time of each material batch at each processing operation node; based on the extracted information, generating a structured digital production work order containing a unique job identifier, material batch information and control instructions for each processing operation node; transmitting the digital production work order to the manufacturing execution system in real time through the interface, and the system sends it to the corresponding processing operation node to perform automated control; the structured digital production work order includes data fields for recording and returning the actual completion time, material consumption and quality inspection results of each node, which are used for subsequent closed-loop management and business performance tracking.

Images