Industrial software service combination method based on improved multi-target grey wolf algorithm

Abstract

The invention discloses an industrial software service combination method based on an improved multi-objective grey wolf algorithm, and the method comprises the following steps: building an industrial software resource service combination optimization problem framework, and completing the evaluation of QoS indexes including service time, service cost, service reputation, service delivery quality and the like and availability indexes; an SCOS model with optimal service quality (QoS) and availability as targets is constructed; tent chaotic mapping and a Levy flight improved differential evolution strategy are fused into a multi-target grey wolf algorithm model, the improved multi-target grey wolf algorithm model is used for solving an SCOS model, and the effectiveness and feasibility of the method are verified through algorithm performance verification and example testing. According to the improved multi-target grey wolf algorithm model, the convergence speed can be effectively increased, manual combination is liberated, and the problem of local optimization of industrial software combination is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of service composition optimization technology, and specifically to an industrial software service composition method based on an improved multi-objective gray wolf algorithm. Background Technology

[0002] In the wave of digital transformation, SMEs face constraints in funding, technology, and talent, leading to problems such as information silos, complex system integration, and outdated technology, severely impacting their transformation process. Industrial software, as a key tool driving digital transformation, helps SMEs integrate business processes, optimize production plans, inventory management, and supply chain coordination through data collection, real-time monitoring, resource optimization, and automated control. This improves operational efficiency and reduces management costs, making it a core force for SMEs' digital transformation. Taking printing equipment manufacturers as an example, their demand for industrial software during digital transformation is growing, addressing pain points such as multi-workpiece, small-batch, and customized production. However, given the diversity of software needs and limited resources in printing equipment manufacturing, current production scenarios often rely on manual experience for selection, which is time-consuming and complex. How to quickly select functionally compatible, cost-effective, and collaboratively efficient software combinations from a vast pool of industrial software to support end-to-end optimization in production, management, and service has become a pressing issue. Against this backdrop, industrial software platforms, through standardized data interfaces, have solved the incompatibility problem between different software. The platform can accurately match suitable software tools based on user needs, combine, optimize, and select them, thereby realizing the intelligent configuration of industrial software resources.

[0003] Industrial software portfolio optimization is a multi-objective decision-making process facing complex constraints. In practical applications, simply optimizing QoS indicators is insufficient to guarantee long-term stable system operation; most problems stem from structural flaws in system availability design. Availability, as a key indicator of continuous service provision capability, not only determines the effective service capability of the software portfolio over time but also constructs a defense against internal and external interference through dimensions such as fault recovery efficiency, redundancy design rationality, and fault tolerance mechanism completeness. Therefore, ensuring availability is crucial in industrial software portfolio optimization and is the foundation for ensuring the stable operation of digital businesses. This invention automates the complex process of selecting industrial software based on manual experience using a computer platform, while simultaneously addressing the problems of existing methods failing to consider availability indicators and exhibiting insufficient global optimization capabilities and a tendency to get trapped in local optima when dealing with industrial software service portfolio optimization. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an industrial software service composition method based on an improved multi-objective gray wolf algorithm, comprising:

[0005] Establish a framework for optimizing industrial software resource service portfolios;

[0006] By employing the aforementioned industrial software resource service combination optimization framework, the service quality and availability of multiple industrial software are obtained;

[0007] To optimize the quality and availability of the aforementioned industrial software services, an optimal model for the combination of industrial software services is established.

[0008] The industrial software service composition optimization model is solved using a pre-constructed improved multi-objective gray wolf algorithm model to obtain the industrial software service composition.

[0009] The improved multi-objective gray wolf algorithm model is obtained by improving the multi-objective gray wolf algorithm model using a differential evolution strategy based on Tent chaotic mapping and Lévy flight.

[0010] Preferably, the establishment of the industrial software resource service portfolio optimization framework includes:

[0011] The service quality and functional requirements of industrial software users are searched and matched in the initial search space of the service cloud pool.

[0012] Based on the search and matching results, a software resource set is generated;

[0013] Select the required software from the resource set of each subtask to generate a software service portfolio;

[0014] The aforementioned software service portfolio is used as an optimal framework for industrial software resource service portfolio selection.

[0015] Preferably, the step of establishing an optimal industrial software service portfolio model with the goal of maximizing the quality and availability of the multiple industrial software services includes:

[0016] We construct weighted functions covering four dimensions: time, cost, reputation, and quality. The goal is to minimize the fusion value of the weighted functions of the four dimensions and their corresponding weights, thus forming the service quality indicator objective function.

[0017] Construct a usability objective function based on the usability of individual software programs;

[0018] Based on the service quality index objective function and the availability objective function, an optimal industrial software service portfolio model is formed with the goal of optimizing the quality and availability of the multiple industrial software services.

[0019] Preferably, the pre-construction process of the improved multi-objective gray wolf algorithm model includes:

[0020] Initialize the gray wolf population of the multi-objective gray wolf algorithm model using Tent chaotic mapping;

[0021] The historical industrial software service portfolio optimization model is used as the gray wolf population. The multi-objective gray wolf algorithm model is trained with the goal of outputting the historical industrial software service portfolio optimization model corresponding to the historical industrial software service portfolio optimization model. During the training process, the evolutionary process of the gray wolf population is perturbed and crossovered using a differential evolution strategy based on Levy flight improvement. A mixed perturbation is applied based on a set stagnation threshold to obtain the improved multi-objective gray wolf algorithm model.

[0022] Preferably, during the training process, a differential evolution strategy based on Levy's flight improvement is used to perturb and crossover the evolutionary process of the gray wolf population; and a mixed perturbation is applied based on a set stagnation threshold to obtain an improved multi-objective gray wolf algorithm model, including:

[0023] During the training process, the traditional differential step size of the differential evolution strategy is replaced with the Lévy step size, and a parameter adaptive mechanism is designed. The type of the Lévy step size is controlled by setting the λ parameter. The scaling factor F is adaptively adjusted using the statistical characteristics of the Lévy step size. The crossover probability CR is adaptively adjusted based on the individual sparsity of the gray wolf population. Gaussian Lévy perturbation is applied to individuals that have remained unchanged for 10 generations, and finally, an improved multi-objective gray wolf algorithm model is obtained.

[0024] Preferably, the objective function of the service quality index satisfies the following formula:

[0025]

[0026] In the formula, QoS (CMS) represents the Quality of Service Index function. Represents time weight, Q T Represents time attribute functions, Represents cost weight, Q C Represents cost attribute functions, Represents reputation weight, Q Re Represents reputation attribute functions, Represents quality weight, Q Q Represents a quality attribute function. and .

[0027] Preferably, the construction of the usability objective function based on the usability of a single software application includes:

[0028] Based on the relationships between software, construct multiple software dependency matrices; determine the software dependency strength and overall coupling degree according to the dependency matrices.

[0029] To adjust the availability of each software based on the inter-software dependency strength, an integration uncertainty factor is introduced.

[0030] Based on the correction results, the integrated uncertainty factor was set;

[0031] Based on the availability of individual software, the integration uncertainty factor, the dependency matrix between multiple software, the dependency strength between software, and the overall coupling degree, an availability objective function is constructed.

[0032] Preferably, the availability objective function satisfies the following formula:

[0033]

[0034] In the formula, i represents the software index variable, n represents the total number of i, and A combo MTBF represents the availability metric value after combining multiple software components. i MTTR represents the mean time between failures (MTTR) of the i-th software. i This represents the fault diagnosis and recovery efficiency of the i-th software. Let represent the confidence factor of the i-th software, and C represent the overall coupling degree. This is the coupling penalty factor.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] This invention effectively improves the utilization efficiency of industrial software in small and medium-sized enterprises (SMEs) such as those manufacturing printing equipment, ensuring the efficient use of related industrial software. It provides quantifiable and implementable decision support tools for the fault-tolerant design, stability assessment, and intelligent deployment of industrial software systems. A framework for optimizing industrial software service composition is established, considering both the QoS requirements and availability indicators of the demand side in terms of optimization objectives. This invention develops an improved multi-objective gray wolf optimization algorithm to solve the industrial software service composition optimization problem. The Tent chaotic mapping is used to initialize the population, effectively improving the diversity of the initial population. A Levy flight-improved differential evolution strategy is introduced to enhance the algorithm's global optimization capability, avoiding getting trapped in local optima. It exhibits excellent comprehensive performance in achieving software optimization and improving system stability. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating an industrial software service composition method based on an improved multi-objective gray wolf algorithm according to the present invention.

[0038] Figure 2 This is a schematic diagram of the industrial software service composition optimization problem described in this invention, where C represents candidate software resources in the service cloud pool, and CS represents different resource service combinations;

[0039] Figure 3The algorithm provided in this embodiment of the invention is used to solve the optimal solution distribution of the benchmark function ZDT3;

[0040] Figure 4 The Pareto front is used to solve the service combinatorial optimization problem using the algorithm provided in this embodiment of the invention. Detailed Implementation

[0041] like Figure 1 As shown, the present invention provides an industrial software service composition method based on an improved multi-objective gray wolf algorithm, comprising:

[0042] Step 1: Establish an optimal framework for the combination of industrial software resource services;

[0043] Step 2: Using the aforementioned industrial software resource service combination optimization framework, the service quality and availability of multiple industrial software are obtained;

[0044] Step 3: Establish an optimal industrial software service combination model with the goal of achieving the best quality and availability of the multiple industrial software services;

[0045] Step 4: Solve the industrial software service composition optimization model using a pre-constructed improved multi-objective gray wolf algorithm model to obtain the industrial software service composition; wherein, the improved multi-objective gray wolf algorithm model is obtained by improving the multi-objective gray wolf algorithm model using a differential evolution strategy based on Tent chaotic mapping and Lévy flight.

[0046] Furthermore, such as Figure 2 As shown, the establishment of the industrial software resource service combination optimization framework in step 1 above includes:

[0047] Step 1.1 Search and match resources based on the service quality and functional requirements of industrial software users in the initial search space of the service cloud pool;

[0048] Step 1.2 Based on the search and matching results, generate a software resource set;

[0049] Step 1.3 Select the required software from each subtask resource set to generate a software service portfolio set;

[0050] Step 1.4 Use the software service combination set as the preferred framework for industrial software resource service combination.

[0051] In step 1 above, when establishing the industrial software resource service combination optimization framework, the service quality and functional requirements of the acquired industrial software users are searched and matched with corresponding resources in the initial search space of the set service cloud pool to generate a corresponding software resource set. Then, the software required for the task can be selected from each sub-task resource set of the software resource set to generate the optimal service combination set and fed back to the user.

[0052] Furthermore, in step 3 above, establishing an optimal industrial software service portfolio model with the goal of maximizing the quality and availability of the multiple industrial software services includes:

[0053] Step 3.1: Construct weighted functions covering four dimensions: time, cost, reputation, and quality. The goal is to minimize the fusion value of the weighted functions of the four dimensions and their corresponding weights to form the service quality indicator objective function.

[0054] Step 3.2: Construct a usability objective function based on the usability of individual software components;

[0055] Step 3.3: Based on the service quality index objective function and the availability objective function, and with the goal of optimizing the quality and availability of the multiple industrial software services, form an optimal industrial software service portfolio model.

[0056] In step 3.1 above, service quality indicator objective functions covering four dimensions—time, cost, reputation, and quality—are constructed. The objective function aims to minimize the sum of the products of the normalized indicator values ​​and their corresponding weights across the four dimensions.

[0057] The specific process is as follows: The four dimensions of time, cost, reputation, and quality are divided into positive and negative indicators. Time and cost are negative indicators (lower values ​​are better), while reputation and quality are positive indicators (higher values ​​are better). All indicator values ​​are normalized and mapped to the [0, 1] interval.

[0058] The normalization process described above is formulated as follows:

[0059]

[0060] In the formula, x max and x min These represent the maximum and minimum values ​​of the evaluation indicators, respectively; q k This indicates the importance or weight of the k-th indicator after normalization in the entire evaluation system, and it is located in the interval [0,1].

[0061] The process of constructing the weighted function across the four dimensions is as follows: The overall attribute index value of the service portfolio is obtained using the aggregation formula under the sequential structure. The aggregation formulas for each index under the sequential structure are as follows:

[0062] For time attributes including service response time, task completion time, and resource scheduling time, the function expression is:

[0063]

[0064] In the formula: Q TRepresents the time attribute value, m represents the total number of j, j is the vector dimension, and q T (ST j ) represents the subtask ST j The time performance of the selected candidate services;

[0065] For cost attributes including service rental costs, computing resource costs, and data storage costs, the function expression is:

[0066]

[0067] In the formula: Q C Represents the cost attribute value, q C (ST j ) represents the subtask ST j Cost performance of the selected candidate services;

[0068] For reputation attributes including user rating, historical call count, service success rate, ecosystem compatibility score, and security certification compliance score, the function expression is:

[0069]

[0070] In the formula: Q Re Represents the reputation attribute value, q Re (ST j ) represents the subtask ST j The credibility of the selected candidate services;

[0071] For quality attributes including interoperability, execution accuracy, and maintainability, the functional expression is:

[0072]

[0073] In the formula: Q Q Represents the quality attribute value, q Q (ST j ) represents the subtask ST j The credibility of the selected candidate service.

[0074] In step 3.2 above, constructing the usability objective function based on the usability of a single software application includes:

[0075] Step 3.2.1: Based on the relationships between software, construct multiple software dependency matrices; determine the software dependency strength and overall coupling degree according to the dependency matrices;

[0076] Step 3.2.2: For the inter-software dependency strength, introduce an integration uncertainty factor to correct the availability of each software;

[0077] Step 3.2.3: Based on the correction results, set the integrated uncertainty factor;

[0078] Step 3.2.4: Construct an availability objective function based on the availability of a single software, the integration uncertainty factor, the dependency matrix between the multiple software, the dependency strength between the software, and the overall coupling degree.

[0079] In step 3.2.1 above, the specific process of constructing multiple software dependency matrices based on the relationships between software, and determining the software dependency strength and overall coupling degree based on the dependency matrices, is as follows:

[0080] The relationships between dependent software are categorized into three types: indirect dependency, control flow dependency (in series), and joint control flow dependency with mandatory control. The dependency strength is 0 for indirect dependency, 0.5 for control flow dependency (in series), and 1.0 for joint control flow dependency with mandatory control. A dependency matrix D=[d] is then constructed. wv ], reaction software S w For S v The strength of dependence;

[0081] Usability modeling requires a quantitative assessment of the coupling degree between software components to serve as a penalty factor for potential risks. A dependency matrix is ​​used to express the strength of dependencies between software components. Based on this dependency matrix, the formula for the overall coupling degree C between software components is as follows:

[0082]

[0083] In the formula, e represents the number of components, d wv Let w represent the elements in the dependency matrix, where w = 1.2...w and v = 1.2...v.

[0084] Furthermore, in step 3.2.2, the process of introducing an integrated uncertainty factor to correct the availability of each software is as follows:

[0085] Determine the confidence factor based on the strength of software dependencies. Let represent the reliability level of the i-th software in collaboration with other software. When a software has a history of successful integration, high technological maturity, or good interface standardization, let . =1 indicates high integration reliability; however, for newly introduced services, or in situations with complex interfaces and significant vendor heterogeneity, ... =0.5, representing its potential uncertainty risk. In the calculation, the availability of each software component is... Multiply by the corresponding This results in the improved software usability. ;

[0086] Furthermore, in step 3.2.4, when constructing the availability objective function, a service software set S is first defined. This set includes multiple industrial software programs from different vendors and using different technology stacks. The software set contains n individual software programs s, as shown in the following formula:

[0087]

[0088] Calculate the availability of individual software programs, and for each software program s, its availability A. i The formula is as follows:

[0089]

[0090] Where: MTBF i MTTR represents the mean time between failures (MTTR) of the i-th software. i Let represent the fault diagnosis and recovery efficiency of the i-th software component. To accurately reflect the operational reliability of the entire software suite, the availability objective function is constructed by comprehensively considering the above factors and satisfying the following formula:

[0091]

[0092] In the formula, A combo MTBF represents the availability metric value after combining multiple software components. i MTTR represents the mean time between failures (MTTR) of the i-th software. i This represents the fault diagnosis and recovery efficiency of the i-th software. Let represent the confidence factor of the i-th software. This is a coupling penalty factor between software components.

[0093] In step 3.3 above, the objective function of the service quality indicator satisfies the following formula:

[0094]

[0095] In the formula, QoS (CMS) represents the Quality of Service Index function. Represents time weight, Q T Represents time attribute functions, Represents cost weight, Q C Represents cost attribute functions, Represents reputation weight, Q Re Represents reputation attribute functions, Represents quality weight, Q Q Represents a quality attribute function. and .

[0096] In step 3.3, the resulting industrial software service portfolio optimization model satisfies the following formula:

[0097]

[0098] In the formula, QoS (CMS) is the utility value of software service quality, A combo (CMS) is the energy consumption value of software service composition, F η Let f1 represent the objective function vector of the multi-objective optimization problem, where f1 represents the service quality optimization objective and f2 represents the system availability optimization objective.

[0099] Furthermore, in step 4 above, the pre-construction process of the improved multi-objective gray wolf algorithm model includes:

[0100] Step a, initialize the gray wolf population of the multi-objective gray wolf algorithm model using Tent chaotic mapping;

[0101] Step b involves using the obtained historical industrial software service combination optimization model as the gray wolf population, and training the multi-objective gray wolf algorithm model with the output of the historical industrial software service combination corresponding to the obtained historical industrial software service combination optimization model as the objective. During the training process, a differential evolution strategy based on Levy flight improvement is used to perturb and crossover the evolutionary process of the gray wolf population. A mixed perturbation is applied based on a set stagnation threshold to obtain the improved multi-objective gray wolf algorithm model.

[0102] In step a above, the gray wolf population is initialized using the Tent chaotic mapping formula:

[0103]

[0104] In the formula: x n The information is for the individual wolves in the current wolf pack. μ is a control parameter. When μ=2, the multi-objective gray wolf algorithm model enters a completely chaotic state and can generate non-periodic orbits with uniform distribution characteristics.

[0105] In step b above, the specific process is as follows:

[0106] Step B1: Use the historical industrial software service portfolio optimization model as the gray wolf population, set the variable dimension, population size, maximum number of iterations, archive size, grid parameters, leader selection pressure parameters, archive deletion pressure parameters, and set the stagnation threshold.

[0107] Step B2: Calculate the initial non-dominated solution and initialize the archive, construct the hypercube network and calculate the network index of the archived solution;

[0108] Step B3: Update the convergence factor a, adaptively adjust the scaling factor F and crossover probability CR, and set the current iteration number t=1;

[0109] Step B4: Select α, β, and δ leaders from the archive based on grid fitness and selection pressure, and update the gray wolf positions;

[0110] Step B5: Perturbation is performed using the Levy-flight improved differential evolution strategy;

[0111] Step B6: Perform a binomial crossover operation, dynamically adjust the CR based on the distribution of the Pareto front region, and greedily select and update the individual positions;

[0112] Step B7: If the gray wolf's stagnation value is greater than or equal to the stagnation threshold, apply a Levi-Gaussian mixed perturbation to reset the stagnation value; otherwise, do not change it. If the gray wolf's stagnation value is less than the stagnation threshold, return to step B4.

[0113] Step B8: Traverse the population size, merge the new dominant solutions into the archive, rebuild or update the hypercube, update the grid index of the archive members, let t=t+1, and return to step B3;

[0114] Step B9: If the current file is larger than the external file's capacity limit, then delete redundant solutions based on grid density;

[0115] Step B10: When t reaches the maximum number of iterations, the final improved multi-objective gray wolf algorithm model is obtained.

[0116] In step B4 above, selecting α, β, and δ leaders from the archive and updating the gray wolf positions based on grid fitness and selection pressure includes:

[0117] The gray wolf locates potential prey through a global search, a process that simulates the algorithm's mechanism of exploring unknown regions in the solution space. The pursuit behavior is as follows:

[0118]

[0119] in, , , These represent the distances between the current gray wolf individual and the leader. , , These represent the positions of the current best leader; This indicates the current position of the individual gray wolf in the solution space. , , These represent different random perturbation vectors.

[0120] Subsequently, the gray wolf gradually narrowed the encirclement, simulating a fine-grained search within a local area. Based on the positions of the α, β, and δ wolves, the gray wolf adjusted its own direction:

[0121]

[0122] In the formula, , , Indicates the leader's current location. , , The three wolves' attraction coefficients to the leader are used to ultimately update the position to the average of the three. :

[0123]

[0124] In step b above, during the training process, a differential evolution strategy based on Levy flight improvement is used to perturb and crossover the evolutionary process of the gray wolf population; and a mixed perturbation is applied based on a set stagnation threshold to obtain an improved multi-objective gray wolf algorithm model, including:

[0125] Step b1: Replace the traditional differential step size of the differential evolution strategy with the Lévy step size, and design a parameter adaptation mechanism, as follows:

[0126]

[0127] In the formula: v i (g) represents the velocity vector of the i-th individual in the g-th generation, x r1 x r2 x r3 For three randomly selected distinct individuals, L(λ) replaces the original scaling factor F, and λ is used as the global adjustment parameter;

[0128] Step b2: Control the type of the Levi step size by setting the λ parameter. The step size ratio is controlled by the parameter λ∈(0,1): λ<0.5 is a short step size, which enhances local exploration; λ>0.5 is a long step size, which strengthens global exploration and completes differential mutation.

[0129] Step b3: Adaptively adjust the scaling factor F using the statistical characteristics of the Lévy step size, as follows:

[0130]

[0131] In the formula: F represents the new value of the scaling factor used in the next generation, F min F represents the minimum value of the scaling factor. max µ represents the maximum value of the scaling factor. L Let σ be the mean of the current Levy step size. L Let σ be the standard deviation of the current Levy step size. max This represents the maximum value of the Lévy step size standard deviation, µ. maxThis represents the maximum value of the Lévy step size mean. In the high-variance phase, it represents a global search; decreasing F suppresses the perturbation amplitude. In the low-variance phase, increasing F enhances local convergence. µ 阈值 =0.5µ max The threshold for the mean of the Levy step size;

[0132] Step b4: Adaptively adjust the crossover probability CR based on the individual sparsity of the gray wolf population, as follows:

[0133]

[0134] In the formula: CR represents the crossover probability of the next generation, CR max CR is the maximum crossover probability. min ρ is the minimum crossover probability, λ is the global adjustment factor used to control the adjustment magnitude; 阈值 =0.5 is the sparsity threshold, ρ∈[0,1] is the local normalized sparsity of an individual, and ρ is obtained by crowding distance normalization under non-dominated sorting.

[0135] Step b5: Apply the Gauss-Levie perturbation to individuals that have remained unchanged for 10 generations, ultimately obtaining the improved multi-objective gray wolf algorithm model, as follows:

[0136]

[0137] In the formula, X new For the updated individual position, X old Let z be the current individual position, z be the perturbation intensity coefficient, and L(β) be the Lévy step size. Let σ represent a normally distributed random variable. f It represents the standard deviation of population fitness.

[0138] To further illustrate and verify the effectiveness of the method of the present invention, two sets of experiments were designed in this specification.

[0139] Case 1: The effectiveness of the improved multi-objective gray wolf algorithm against the benchmark function

[0140] The experimental examples are designed as follows: Figure 3 As shown, ten standard benchmark functions widely used in multi-objective optimization research were selected to evaluate the performance of the proposed improved multi-objective Grey Wolf algorithm. The benchmark functions used include unconstrained test function 1...unconstrained test function 3 (ZDT1-ZDT3) and constrained function 1...constrained function 7 (CF1-CF3). All parameter values ​​were taken from the original literature of each algorithm or their common configurations in the field of multi-objective optimization. Unless otherwise specified, no special tuning was performed. The algorithm parameter settings are shown in Table 1.

[0141] Table 1 Algorithm Parameter Settings

[0142]

[0143] like Figure 4 As shown, generation distance and inverse generation distance are used to evaluate the convergence and distribution of each algorithm. To reduce the impact of random fluctuations, each algorithm is run independently 30 times. The mean and standard deviation of the generation distance and inverse generation distance are listed in Tables 2 and 3, respectively. Mean represents the average value of the results; Std represents the standard deviation of the results, which is used to measure the stability and convergence performance of the algorithm.

[0144] Table 2. Generational Distance Data for Benchmark Function Test

[0145]

[0146] Table 3. Benchmark Function Test Reverse Generation Distance Data

[0147]

[0148] As shown in Table 2, the improved multi-objective gray wolf algorithm (HEMOGWO) of this invention achieved optimal results on all benchmark functions except CF1, CF5, and CF6. The multi-objective gray wolf algorithm (MOGWO) performed best on CF6. Although the decomposition-based multi-objective evolutionary algorithm – Differential Evolution (MOEAD_DE) only achieved optimal results on CF1, its performance on this function was extremely excellent. However, the multi-objective particle swarm optimization algorithm (MOPSO) and the non-dominated sorting genetic algorithm III (NSGA-III) failed to achieve the optimal generation distance index in all test functions. As shown in Table 3, the improved multi-objective gray wolf algorithm achieved the optimal inverse generation distance value in all seven benchmark functions except CF2 and CF5. The multi-objective gray wolf algorithm showed an advantage on CF2 and CF5. The decomposition-based multi-objective evolutionary algorithm – Differential Evolution, the multi-objective particle swarm optimization algorithm, and the non-dominated sorting genetic algorithm III performed poorly overall in terms of the inverse generation distance index, failing to achieve optimal results in any test function, and exhibiting deficiencies in convergence and solution set distribution.

[0149] To comprehensively evaluate the overall performance of each algorithm and determine its ranking, Friedman nonparametric tests were performed on the generation distance and inverse generation distance results in Tables 2 and 3. The test results for generation distance and inverse generation distance are shown in Table 4.

[0150] Table 4. Friedman test results for the benchmark function test

[0151] (Significance level: 0.05)

[0152]

[0153] As shown in Table 4, the p-values ​​for both generation distance and inverse generation distance are significantly lower than the set significance level of 0.05, indicating statistically significant differences in the performance of the five algorithms in these two metrics. Furthermore, all algorithms rank consistently in both generation distance and inverse generation distance, with the improved multi-objective gray wolf algorithm ranking first in both metrics, demonstrating superior convergence accuracy and distributional stability compared to other algorithms. This result is highly consistent with the numerical performance in Tables 2 and 3.

[0154] Case 2: Improving the performance of the multi-objective Grey Wolf algorithm on problems of different scales

[0155] The experimental design involves decomposing the software requirements into five sub-tasks and performing service matching combinations. The experimental scale is divided into two groups: 5-20 and 5-100, meaning each sub-task has 20 and 100 candidate services respectively. Before calculating the overall QoS index, the weights of the four attributes—time, cost, reputation, and quality—are preset to 0.25. This section still uses generation distance and inverse generation distance indices to evaluate the convergence and distribution of each algorithm. To reduce the impact of random fluctuations, each algorithm is run independently 30 times. The mean and standard deviation of the obtained generation distance and inverse generation distance indices are shown in Table 5.

[0156] Table 5. Generational distance and inverse generational distance data for the service composition optimization problem.

[0157]

[0158] Analyzing Table 5, the improved multi-objective Grey Wolf algorithm outperforms other algorithms in both generational and inverse generational distance values ​​for service composition problems of different scales. Overall, the solutions generated by the improved multi-objective Grey Wolf algorithm mostly exhibit broad convergence and diversity, demonstrating good algorithm performance.

[0159] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in the details for the sake of brevity.

[0160] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for composing industrial software services based on an improved multi-objective gray wolf algorithm, characterized in that, include: Establish a framework for optimizing industrial software resource service portfolios; By employing the aforementioned industrial software resource service combination optimization framework, the service quality and availability of multiple industrial software are obtained; To optimize the quality and availability of the aforementioned industrial software services, an optimal model for the combination of industrial software services is established. The industrial software service composition optimization model is solved using a pre-constructed improved multi-objective gray wolf algorithm model to obtain the industrial software service composition. The improved multi-objective gray wolf algorithm model is obtained by improving the multi-objective gray wolf algorithm model using a differential evolution strategy based on Tent chaotic mapping and Lévy flight.

2. The method according to claim 1, characterized in that, The establishment of the industrial software resource service combination optimization framework includes: The service quality and functional requirements of industrial software users are searched and matched in the initial search space of the service cloud pool. Based on the search and matching results, a software resource set is generated; Select the required software from the resource set of each subtask to generate a software service portfolio; The aforementioned software service portfolio is used as an optimal framework for industrial software resource service portfolio selection.

3. The method according to claim 1, characterized in that, The goal is to optimize the quality and availability of the multiple industrial software services, and to establish an optimal industrial software service combination model, including: We construct weighted functions covering four dimensions: time, cost, reputation, and quality. The goal is to minimize the fusion value of the weighted functions of the four dimensions and their corresponding weights, thus forming the service quality indicator objective function. Construct a usability objective function based on the usability of individual software programs; Based on the service quality index objective function and the availability objective function, an optimal industrial software service portfolio model is formed with the goal of optimizing the quality and availability of the multiple industrial software services.

4. The method according to claim 1, characterized in that, The pre-construction process of the improved multi-objective gray wolf algorithm model includes: Initialize the gray wolf population of the multi-objective gray wolf algorithm model using Tent chaotic mapping; The historical industrial software service portfolio optimization model is used as the gray wolf population. The multi-objective gray wolf algorithm model is trained with the goal of outputting the historical industrial software service portfolio optimization model corresponding to the historical industrial software service portfolio optimization model. During the training process, the evolutionary process of the gray wolf population is perturbed and crossovered using a differential evolution strategy based on Levy flight improvement. A mixed perturbation is applied based on a set stagnation threshold to obtain the improved multi-objective gray wolf algorithm model.

5. The method according to claim 4, characterized in that, During the training process, a differential evolution strategy based on Levy's flight improvement is used to perturb and crossover the evolutionary process of the gray wolf population; and a mixed perturbation is applied based on a set stagnation threshold to obtain an improved multi-objective gray wolf algorithm model, including: During the training process, the traditional differential step size of the differential evolution strategy is replaced with the Lévy step size, and a parameter adaptive mechanism is designed; the type of the Lévy step size is controlled by setting the λ parameter; the scaling factor F is adaptively adjusted using the statistical characteristics of the Lévy step size; and the crossover probability CR is adaptively adjusted based on the individual sparsity of the gray wolf population. By applying the Gauss-Levi perturbation to individuals that have remained unchanged for 10 generations, an improved multi-objective gray wolf algorithm model is obtained.

6. The method according to claim 3, characterized in that, The objective function of the service quality index satisfies the following formula: In the formula, QoS (CMS) represents the Quality of Service Index function. Represents time weight, Q T Represents time attribute functions, Represents cost weight, Q C Represents cost attribute functions, Represents reputation weight, Q Re Represents reputation attribute functions, Represents quality weight, Q Q Represents a quality attribute function. and .

7. The method according to claim 3, characterized in that, The usability objective function, based on the usability of a single software application, includes: Based on the relationships between software, construct multiple software dependency matrices; determine the software dependency strength and overall coupling degree according to the dependency matrices. To adjust the availability of each software based on the inter-software dependency strength, an integration uncertainty factor is introduced. Based on the correction results, the integrated uncertainty factor was set; Based on the availability of individual software, the integration uncertainty factor, the dependency matrix between multiple software, the dependency strength between software, and the overall coupling degree, an availability objective function is constructed.

8. The method according to claim 7, characterized in that, The availability objective function satisfies the following formula: In the formula, i represents the software index variable, n represents the total number of i, and A combo MTBF represents the availability metric value after combining multiple software components. i MTTR represents the mean time between failures (MTTR) of the i-th software. i This represents the fault diagnosis and recovery efficiency of the i-th software. Let represent the confidence factor of the i-th software, and C represent the overall coupling degree between the software. This is the coupling penalty factor.

Images