MOM software quality optimization method and system based on dynamic fuzzy scenario

Abstract

The application discloses a MOM software quality optimization method and system based on a dynamic fuzzy scene, which comprises the following steps: collecting heterogeneous data in multiple dimensions from operation monitoring, user interaction and business process tracking; extracting a core feature subset through information measurement after cleaning and normalization; defining an operation scene in combination with system load and other features and core feature performance; constructing a double-layer weight model and dynamically adjusting the model; establishing a fuzzy mapping of features and preset evaluation levels for each scene to form a judgment matrix; determining the quality level according to the maximum membership degree after the weight and the matrix are combined; and generating software configuration, resource allocation or workflow adjustment instructions based on the evaluation results and executing the instructions to realize closed-loop optimization. The method realizes precise quantitative evaluation and automatic optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of industrial software and artificial intelligence, and particularly relates to a MOM software quality optimization method and system based on dynamic fuzzy scene. BACKGROUND

[0002] As the core platform of digital transformation in manufacturing industry, the software quality of manufacturing operation management system (MOM) is directly related to the reliability of production planning, the accuracy of resource scheduling and the qualification rate of final products. In complex industrial fields such as aerospace and high-end equipment manufacturing, this requirement is particularly stringent. The production activities in these fields present high dynamicity and scene characteristics: for example, in the same set of MOM software, different scenes such as "super-precision part machining", "multi-variety small-batch agile production" and "equipment fault emergency handling" may be alternately run, and the emphasis on quality attributes such as performance, reliability and functional completeness of the software is different in each scene.

[0003] Traditional software quality evaluation methods mostly rely on static index system and fixed weight model, such as the quality model based on ISO / IEC25010 standard. Such methods cannot adapt to the dynamic changes of runtime context, and their evaluation results are often quickly invalidated when the scene is switched, lacking pertinence. Although fuzzy comprehensive evaluation method is introduced to deal with subjectivity and uncertainty in quality evaluation, the membership degree function is used to convert qualitative description into quantitative calculation, but the weight vector is usually subjectively assigned by experts or obtained by offline historical data training. Once determined, this method is fixed and cannot be self-optimized and adjusted according to real-time running data, user feedback and sudden scene changes, resulting in serious lag in evaluation results.

[0004] In the prior art, although some researches have tried to introduce dynamic factors into the evaluation system, they mostly stay at the theoretical level and lack a complete, engineering-based and deeply integrated technical solution with MOM system. Especially in the context of the rise of low-code development platform, how to provide real-time and scene-based quality assurance for rapidly constructed and iterated MOM applications, and to realize automatic optimization based on evaluation results, has become a key bottleneck restricting its in-depth application. SUMMARY

[0005] The main purpose of the present application is to overcome the above-mentioned defects in the prior art, and to provide a MOM software quality optimization method and system based on dynamic fuzzy scene, which realizes dynamic weight optimization and fuzzy comprehensive evaluation, and generates optimization strategies based on evaluation results, outputs accurate quantitative results and optimization instructions.

[0006] The present application adopts the following technical solutions:

[0007] A MOM software quality optimization method based on dynamic fuzzy scene, comprising:

[0008] Collecting software quality multi-source heterogeneous data from multiple dimensions of MOM system operation monitoring, user interaction records and business process tracking, obtaining a standardized dataset through cleaning and normalization processing, and then extracting a core feature subset strongly related to software quality from the standardized dataset based on an information measurement method;

[0009] Defining and dividing software running scenes according to system load state, user behavior pattern, business workflow characteristics and different performance modes of the core feature subset, constructing a double-layer weight model containing basic weight and dynamic weight for each software running scene, and completing dynamic weight adjustment combined with historical experience and real-time feedback;

[0010] For each software running scene, a fuzzy mapping relationship between quality features in the core feature subset and preset evaluation grades is established to form a fuzzy judgment matrix, a synthetic operation is performed based on the weight vector after dynamic adjustment and the corresponding fuzzy judgment matrix to obtain a scene-based quality comprehensive evaluation vector, and the final quality grade is determined according to the maximum membership degree principle;

[0011] Based on the quality comprehensive evaluation vector, optimization instructions for software configuration, resource allocation or workflow adjustment are generated and issued to the MOM system for execution to realize software quality closed-loop optimization.

[0012] The cleaning and normalization processing obtains a standardized dataset, which specifically includes the following:

[0013] The multi-source heterogeneous data is cleaned, and then the min-max normalization method is used to linearly map the cleaned data to the [0, 1] interval, and the calculation formula is:

[0014] ;

[0015] Wherein, is the normalized data value, is the cleaned data, and are the minimum and maximum values of the data set, respectively.

[0016] Interaction information is used as a feature correlation measurement index, and based on the standardized dataset, the most relevant features to software quality are selected from all features according to a predetermined selection strategy to form a core feature subset, and the specific expression is as follows:

[0017] ;

[0018] Wherein, is a single feature variable to be screened in the standardized dataset, It refers to the specific values ​​that the characteristic variable X takes. It is a quality level variable. Representing characteristic variables With quality target variables The amount of information exchanged between them Representing the eigenvector Values And the quality grade variable takes the value of The joint probability distribution, and They represent the characteristic variables respectively. Pick Marginal probability distribution and quality grade variables Pick The marginal probability distribution.

[0019] Based on a core feature subset and the quality levels of expert annotations, a structured training sample library is constructed. A random forest regression algorithm is used, and a feature importance scoring mechanism is employed to calculate the first [value] in the core feature subset. The quality characteristic in the first Initial base weights for each scenario ;

[0020] For each quality feature in the core feature subset, based on the current scenario Based on the real-time evaluation error and scene change signals, the basic weights are adjusted through a weight adaptive update mechanism, the update formula of which is as follows:

[0021] ;

[0022] in, Indicates the adjusted number A quality characteristic in the scene Dynamic weights under; For adaptive learning rate parameters, This represents the evaluation error in the current scenario. This refers to the weight adjustment amount calculated based on real-time scene switching signals and a business priority rule base.

[0023] The basic weights and dynamic weights are combined to obtain the scene. The combined weight of each quality feature in the core feature subset :

[0024] ;

[0025] in, Indicates the total number of quality characteristics, ensuring .

[0026] Adaptive learning rate parameter An annealing strategy is adopted, and the value gradually decreases as the number of evaluation iterations increases. The update rule is as follows:

[0027] ;

[0028] in, Indicates the first Learning rate during iteration The initial learning rate, The attenuation coefficient is... This represents the current iteration number. .

[0029] For the scenario , will be comprehensive weight With fuzzy relation matrix Perform a synthesis operation to obtain the comprehensive evaluation vector for this scenario. ;

[0030] ;

[0031] in, This represents the weighted average composition operator;

[0032] Based on the comprehensive evaluation vector The final quality level is determined using the principle of maximum membership, i.e., a comprehensive evaluation vector is selected. The evaluation level corresponding to the maximum value is used as the software quality assessment result in this scenario.

[0033] Based on quality-driven dynamic optimization decision-making, at least one of the following optimization instructions is generated and executed according to the comprehensive evaluation vector:

[0034] Resource scheduling command: Dynamically adjust the CPU quota or memory allocation of the container where the MOM software is located;

[0035] Business process refactoring instruction: Rearrange the execution sequence of pending work orders according to task priority;

[0036] Parameter configuration commands: Adjust the number of thread pools or database connection parameters related to software performance.

[0037] A MOM software quality optimization system based on dynamic fuzzy scene mapping, including

[0038] The data acquisition, preprocessing, and feature extraction module is used to collect multi-source heterogeneous data on software quality from multiple dimensions, including MOM system operation monitoring, user interaction records, and business process tracking. The module cleans and normalizes the collected heterogeneous data to obtain a standardized dataset, and extracts a subset of core features that are strongly correlated with software quality from the standardized dataset based on information measurement methods.

[0039] The scene definition and weight adjustment module is configured to define and divide software running scenes according to system load states, user behavior patterns, business workflow features and different performance modes of the core feature subset, and to construct a double-layer weight model including basic weights and dynamic weights for each software running scene, and to complete dynamic adjustment of the weights in combination with historical experience and real-time feedback.

[0040] The fuzzy comprehensive evaluation module is configured to establish a fuzzy mapping relationship between quality features in the core feature subset and preset evaluation grades for each software running scene to form a fuzzy judgment matrix, and to perform synthetic operation based on the dynamically adjusted weight vector and the corresponding fuzzy judgment matrix to obtain a scene-based quality comprehensive evaluation vector, and to determine a final quality grade according to the maximum membership degree principle.

[0041] The closed-loop optimization execution module is configured to generate optimization instructions for software configuration, resource allocation or workflow adjustment based on the quality comprehensive evaluation vector, and to issue the optimization instructions to the MOM system for execution to realize closed-loop optimization of software quality.

[0042] The micro-service architecture is adopted and deployed in a Kubernetes cluster through containerization technology, including:

[0043] The data collection pod is configured to implement the multi-source heterogeneous data collection function in the data collection preprocessing and feature extraction module.

[0044] The feature processing pod is configured to implement the data cleaning, normalization processing and core feature subset extraction function in the data collection preprocessing and feature extraction module.

[0045] The scene recognition pod is configured to implement the software running scene definition and classification function in the scene definition and weight adjustment module.

[0046] The weight optimization pod is configured to implement the double-layer weight model construction and dynamic adjustment function in the scene definition and weight adjustment module.

[0047] The comprehensive evaluation pod is configured to implement the fuzzy judgment matrix construction, quality comprehensive evaluation vector calculation and final quality grade determination function in the fuzzy comprehensive evaluation module.

[0048] The strategy execution pod is configured to implement the optimization instruction generation and execution function of the closed-loop optimization execution module.

[0049] The pods communicate with each other through a service mesh to achieve distributed and elastic expansion.

[0050] The closed-loop optimization execution module includes a business optimization strategy library configured to map the final quality grade and the quality comprehensive evaluation vector into three types of executable instructions.

[0051] Resource scheduling instructions for dynamically adjusting the CPU and memory resource limits of the MOM software Pod through the Kubernetes API;

[0052] Process reconstruction instructions for rearranging the execution priority and path of the production ticket through the workflow engine API of the MOM system;

[0053] Parameter configuration instructions for dynamically adjusting key operating parameters such as database connection pool and thread pool through the configuration management center API.

[0054] From the above description of the present application, compared with the prior art, the present application has the following beneficial effects:

[0055] 1. The present application can identify the quality bottleneck in different software running scenarios through the dynamic weight adjustment mechanism based on scene definition and the fuzzy comprehensive evaluation method, and provide reliable basis for subsequent decisions such as resource optimization and process reconstruction, which improves the accuracy of quality problem positioning.

[0056] 2. The present application relies on real-time data acquisition preprocessing, dynamic weight self-adaptive updating and real-time evaluation error feedback mechanism, so that the quality evaluation model can be dynamically adjusted following the system running state, effectively avoiding the deviation of the evaluation result caused by data lag.

[0057] 3. The present application automatically generates executable optimization instructions such as resource scheduling, process reconstruction and parameter configuration based on the evaluation result and issues them to the MOM system for execution, forming a closed-loop mechanism of "data acquisition-evaluation-optimization-execution", and realizing the automatic and continuous optimization of software quality.

[0058] 4. The present application generates a quantitative quality comprehensive evaluation vector and a clear quality level through fuzzy comprehensive evaluation, which provides a quantifiable reference for system optimization decision-making and improves the objectivity and operability of decision-making.

[0059] 5. The present application adopts micro-service architecture combined with containerization technology and is deployed in Kubernetes cluster, and each functional module is distributedly deployed and communicates through service mesh through Pod, which supports dynamic elastic scaling according to system load and adapts to the computing power demand in different business scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 It is the overall flowchart of the method of the present application.

[0061] Figure 2 It is the schematic diagram of the dynamic weight adjustment mechanism in the embodiment of the present application.

[0062] Figure 3 It is the schematic diagram of the micro-service architecture of the system of the present application.

[0063] The application will be further described in the following with reference to the accompanying drawings and specific embodiments. DETAILED DESCRIPTION

[0064] The application will be further described in the following with reference to the accompanying drawings and specific embodiments.

[0065] Reference Figure 1 A MOM software quality optimization method based on dynamic fuzzy scene, comprising:

[0066] S1 multi-source data acquisition preprocessing and key feature extraction, multi-dimensional software quality multi-source heterogeneous data is collected from MOM system operation monitoring, user interaction record and business process tracking, and a standardized data set is obtained after cleaning and normalization processing, and a core feature subset strongly related to software quality is extracted from the standardized data set based on information measurement method;

[0067] In this step, CPU load, memory occupancy and network I / O indicators are collected through system performance monitor, user interaction log analyzer and business workflow tracker. The multi-source heterogeneous data is cleaned and normalized to obtain a standardized data set, which specifically includes the following:

[0068] The multi-source heterogeneous data is cleaned, and then the cleaned data is linearly mapped to the [0, 1] interval by using the min-max normalization method, and the calculation formula is:

[0069] ;

[0070] Wherein, is the normalized data value, is the cleaned data, and are the minimum and maximum values of the data set respectively.

[0071] In this step, the interaction information is used as the feature correlation measurement index, and based on the standardized data set, the most relevant features to software quality are selected from all features according to the predetermined selection strategy to form a core feature subset, and the specific expression is as follows:

[0072] ;

[0073] Wherein, is a single feature variable to be screened in the standardized data set, is the specific value of the feature variable X, is the quality level variable, represents the interaction information between the feature variable and the quality target variable , represents the value of the feature vector X and the quality grade variable takes value , respectively, denote the marginal probability distribution of the feature variable X taking x and the marginal probability distribution of the quality grade variable Y taking y, respectively. and respectively denote the marginal probability distribution of the feature variable X taking x and the marginal probability distribution of the quality grade variable Y taking y, respectively.

[0074] S2 runs the scene definition classification and adaptive weight determination, defines and divides the software running scenes according to the system load state, user behavior mode, business workflow characteristics and different performance modes of the core feature subset, constructs a double-layer weight model containing basic weights and dynamic weights for each software running scene, and completes dynamic weight adjustment in combination with historical experience and real-time feedback;

[0075] S21 constructs a structured training sample library based on the core feature subset and in combination with the quality grades marked by experts; adopts a random forest regression algorithm to calculate the initial basic weight of the i-th quality feature in the core feature subset under the j-th scene through a feature importance scoring mechanism; ;

[0076] S22 adjusts the basic weight of each quality feature in the core feature subset through a weight adaptive updating mechanism based on the real-time evaluation error and scene change signal under the current scene , and the updating formula is as follows:

[0077] ;

[0078] wherein, denotes the dynamic weight of the i-th quality feature under the j-th scene after adjustment; denotes the initial weight or the weight value of the i-th quality feature under the j-th scene before adjustment, which is the basis for weight updating; is an adaptive learning rate parameter, is the evaluation error under the current scene, is a weight adjustment amount calculated based on real-time scene switching signals and a business priority rule library; S23 synthesizes the basic weight and the dynamic weight to obtain the comprehensive weight of each quality feature in the core feature subset under the j-th scene :

[0079] wherein, denotes the total number of quality features, and ensures ;

[0080] ;

[0081] wherein, denotes the total number of quality features, and ensures​​​​ .

[0082] In this embodiment, the adaptive learning rate parameter An annealing strategy is adopted, which gradually decreases with the increase of the number of evaluation iterations, and the update rule is:

[0083] ;

[0084] wherein, denotes the learning rate at the i-th iteration, is the initial learning rate, is the decay coefficient, is the current iteration number, .

[0085] S3 multi-scene fuzzy comprehensive evaluation, for each software running scene, a fuzzy mapping relationship between the quality characteristics in the core feature subset and the preset evaluation level is established to form a fuzzy judgment matrix, based on the dynamically adjusted weight vector and the corresponding fuzzy judgment matrix, a scene quality comprehensive evaluation vector is obtained and the final quality level is determined according to the maximum membership degree principle.

[0086] wherein, the preset evaluation level set V={excellent, good, medium, poor}, is a qualitative conclusion set jointly set by domain experts according to business fault tolerance standards and performance expectations before system deployment. When constructing the fuzzy judgment matrix, the quality characteristics followed are the core feature subset determined in step S2.

[0087] For scene , the weighted average composition operator is used to synthesize the comprehensive weight and the fuzzy relationship matrix to obtain the comprehensive evaluation vector under the scene;

[0088] ;

[0089] wherein, denotes the weighted average composition operator;

[0090] According to the comprehensive evaluation vector , the maximum membership degree principle is used to determine the final quality level, that is, the evaluation level corresponding to the maximum value of the comprehensive evaluation vector is selected as the software quality evaluation result under the scene.

[0091] S4 closed-loop optimization execution, based on the final quality level and the quality comprehensive evaluation vector, optimization instructions for software configuration, resource allocation or workflow adjustment are generated and issued to the MOM system for execution to realize software quality closed-loop optimization. ​

[0092] Based on the quality-based dynamic optimization decision, at least one of the following optimization instructions is generated and executed according to the final quality level and the comprehensive evaluation vector:

[0093] Resource scheduling instruction: dynamically adjusting the CPU quota or memory allocation of the container where the MOM software is located;

[0094] Business process reconstruction instruction: rearranging the execution sequence of the to-be-processed work order according to the task priority;

[0095] Parameter configuration instruction: adjusting the thread pool quantity or database connection parameter related to the performance inside the software.

[0096] The application is based on a dynamic fuzzy scenario-based MOM software quality optimization method, which builds a data foundation through multi-source data acquisition preprocessing and core feature extraction, and combines software running scenario definition to implement dynamic weight adjustment and fuzzy comprehensive evaluation, which can accurately identify quality bottlenecks in different scenarios; relying on real-time data support to realize dynamic update of evaluation to avoid lagging problem, and can automatically generate optimization instructions based on the evaluation results to form a closed-loop optimization; at the same time, output the quantitative evaluation results to support decision-making.

[0097] Based on this, the embodiment of the application also proposes a dynamic fuzzy scenario-based MOM software quality optimization system, which comprises

[0098] The data acquisition preprocessing and feature extraction module is used for collecting software quality multi-source heterogeneous data from MOM system running monitoring, user interaction record and business process tracking in multiple dimensions, performing cleaning and normalization processing on the collected heterogeneous data to obtain a standardized data set, and extracting a core feature subset strongly related to software quality from the standardized data set based on an information measurement method;

[0099] The scene definition and weight adjustment module is used for defining and dividing software running scenarios according to system load state, user behavior mode, business workflow characteristics and different performance modes of the core feature subset, and constructing a double-layer weight model containing basic weight and dynamic weight for each software running scenario, and completing dynamic adjustment of the weight in combination with historical experience and real-time feedback;

[0100] The fuzzy comprehensive evaluation module is used for establishing a fuzzy mapping relationship between quality features in the core feature subset and preset evaluation levels to form a fuzzy judgment matrix for each software running scenario, and performing synthetic operation based on the dynamically adjusted weight vector and the corresponding fuzzy judgment matrix to obtain a scenario-based quality comprehensive evaluation vector, and determining a final quality level according to the maximum membership degree principle;

[0101] A closed-loop optimization execution module is configured to generate optimization instructions for software configuration, resource allocation or workflow adjustment based on the final quality level and the quality comprehensive evaluation vector, and to issue the optimization instructions to the MOM system for execution to achieve closed-loop optimization of software quality.

[0102] Referring to Figure 3 In this embodiment, the system adopts a micro-service architecture and is deployed in a Kubernetes cluster through containerization technology, including:

[0103] A data collection Pod is configured to implement the multi-source heterogeneous data collection function in the data collection preprocessing and feature extraction module.

[0104] A feature processing Pod is configured to implement the data cleaning, normalization processing and core feature subset extraction function in the data collection preprocessing and feature extraction module.

[0105] A scene recognition Pod is configured to implement the software running scene definition and classification function in the scene definition and weight adjustment module.

[0106] A weight optimization Pod is configured to implement the double-layer weight model construction and dynamic adjustment function in the scene definition and weight adjustment module.

[0107] A comprehensive evaluation Pod is configured to implement the fuzzy judgment matrix construction, quality comprehensive evaluation vector calculation and final quality level determination function in the fuzzy comprehensive evaluation module.

[0108] A strategy execution Pod is configured to implement the optimization instruction generation and execution function of the closed-loop optimization execution module.

[0109] The Pods communicate with each other through a service mesh to achieve distributed elastic expansion.

[0110] The closed-loop optimization execution module includes a business optimization strategy library configured to map the final quality level and the quality comprehensive evaluation vector into three types of executable instructions:

[0111] A resource scheduling instruction is configured to dynamically adjust the CPU and memory resource limits of the MOM software Pod through the Kubernetes API.

[0112] A process reconstruction instruction is configured to rearrange the execution priority and path of the production order through the workflow engine API of the MOM system.

[0113] A parameter configuration instruction is configured to dynamically adjust the key running parameters such as database connection pool and thread pool through the configuration management center API.

[0114] Embodiment: MOM software quality optimization in the context of aero-engine blade machining

[0115] The embodiment is applied to MOM software quality dynamic evaluation and optimization in the scene of aero-engine blade ultra-precision pneumatic forming machining and flexible multi-configuration agile machining, and the overall process comprises the following steps:

[0116] S1 multi-source data acquisition preprocessing and key feature extraction

[0117] From the MOM system of an aerospace manufacturing enterprise, the running data of the software module for engine blade machining in the next 7 working days is collected. The data sources include:

[0118] System performance monitor: collect CPU average load rate, memory occupancy rate, network I / O throughput and other original indexes.

[0119] User interaction log analyzer: records user operation response time, operation failure rate, function point call frequency and other original logs.

[0120] Business workflow tracker: collects business process original data such as work order processing time, task queue depth, device state switching delay, etc.

[0121] Firstly, the above multi-source heterogeneous raw data is cleaned, and the abnormal values caused by network instantaneous interruption and sensor failure are eliminated. Then, the min-max standardization method is used to map the data to the [0, 1] interval. For example, the CPU load rate original value is 0.72, and the historical minimum and maximum values are 0.10 and 0.95 respectively, and after standardization:

[0122]

[0123] The core features most related to the quality of the aero-engine blade machining software are extracted from the massive data.

[0124] Firstly, according to the existing software quality model and the experience of field experts, a basic quality feature set (20 features in total) covering 6 dimensions (function integrity, performance efficiency, compatibility, usability, reliability, portability) is established, as shown in Table 1:

[0125] Table 1 Software quality feature table

[0126]

[0127] Subsequently, the feature selection training process is started. This process relies on a historical data sample library, which is constructed in the following way: the data collected in the past 3 months and after standardization preprocessing is organized according to the basic quality feature set, and each data sample is labeled with a quality level (excellent, good, qualified, unqualified).

[0128] Based on this "feature-label" sample library, mutual information is used as a measure index to select the 8 most relevant features from the 20 basic features to form the core feature subset. The 8 core features with the highest mutual information value are shown in Table 2, and the 8-dimensional feature vector F = {f1, f2,..., f8} is the input for subsequent scenario-based evaluation.

[0129] Table 2 Core features of software quality in the scenario of precision machining of aero-engine blades

[0130]

[0131] S2 Definition of running scenario classification and adaptive weight determination

[0132] According to the machining requirements of aero-engine blades and the running characteristics of MOM software, the following two typical scenarios are defined:

[0133] Ultra-precision pneumatic forming machining scenario This scenario is related to the core requirement of optimizing the aerodynamic performance of the blade through precise surface design, and the control accuracy and long-term running stability of the software are extremely high. Any slight jitter or calculation error may cause product scrap.

[0134] Flexible multi-configuration agile machining scenario This scenario deals with various blade configurations due to differences in engine models and performance requirements, requiring the software to quickly switch between machining tasks, efficiently calculate new tool paths, and adapt to different materials. The response speed and task adaptability of the software are required to be higher.

[0135] S21 Based on the core feature subset, combined with the labeled quality level, a structured training sample library is constructed. Using the random forest regression algorithm, 80% of the historical samples are used as the training set to learn the initial basic weight vector in the two scenarios.

[0136] For the ultra-precision pneumatic forming machining scenario, the initial basic weight obtained by training is:

[0137] ;

[0138] This weight indicates that in this scenario, the machining operation integrity (f1) and long-term running stability (f6) are the most important.

[0139] When the system switches from scenario to scenario . According to the business requirements of this new scenario (focusing on efficiency and agility), the weights of features f2 (tool path calculation speed), f6 (long-term running stability), and f7 (fault warning accuracy) need to be improved.

[0140] Take feature f2 as an example, when the period-based weight , set the learning rate , the prediction error , the weight adjustment amount .

[0141] Calculate the new weight, that is, the dynamic weight:

[0142] ;

[0143] The final comprehensive weight vector under the scene is obtained by normalizing the adjusted weights of all features:

[0144] =[0.24,0.17,0.13,0.13,0.13,0.24,0.06,0.06];

[0145] This is the final comprehensive weight vector under the flexible multi-configuration agile manufacturing scene.

[0146] S3 multi-scene fuzzy comprehensive evaluation

[0147] Set the preset evaluation level set V={v1, v2, v3, v4}={excellent, good, medium, poor}. Define the membership function for each quality feature and each evaluation level .

[0148] Repeat this process for all 8 features, and combine the 8 membership vectors obtained, that is, form an 8-row × 4-column fuzzy judgment matrix :

[0149] =[

[0150] [0.60,0.30,0.10,0.00],#f1 processing operation integrity

[0151] [0.60,0.80,0.20,0.00],#f2 tool path calculation speed

[0152] [0.40,0.45,0.10,0.05],#f3 simulation processing real-time performance

[0153] [0.20,0.60,0.15,0.05],#f4 equipment control compatibility

[0154] [0.60,0.20,0.20,0.00],#f5 user interface friendliness

[0155] [0.80,0.10,0.10,0.00],#f6 Long-term operational stability

[0156] [0.70,0.25,0.05,0.00],#f7 Fault warning accuracy

[0157] [0.30,0.50,0.10,0.10]#f8 Cross-platform deployment convenience ]

[0159] For flexible, multi-configuration, and agile manufacturing scenarios, the overall weight is:

[0160] =[0.24,0.17,0.13,0.13,0.13,0.24,0.06,0.06];

[0161] Using the weighted average synthesis operator, the overall weights are... With the above fuzzy judgment matrix Perform a synthesis operation to obtain the comprehensive evaluation result vector for this scenario. .

[0162] Taking ultra-precision pneumatic forming processing as an example:

[0163] ;

[0164] The result vector indicates that, in the current scenario, the overall membership of the MOM software quality is 63% for "excellent", 33% for "good", 13% for "medium", and 3% for "poor".

[0165] Final quality level determination: Based on the principle of maximum membership, select... The highest value (0.63) corresponds to the "Excellent" rating, which is the final conclusion of this evaluation. Furthermore, by comparing it with the "Ultra-precision pneumatic forming process scenario" (whose evaluation result is [0.55, 0.28, 0.12, 0.03], also "Excellent"), it can be quantitatively analyzed that the software is more suitable for the "flexible multi-configuration agile processing scenario," providing precise data support for subsequent resource allocation and software optimization.

[0166] S4 Closed-Loop Optimization Execution

[0167] Based on the comprehensive quality evaluation vector, a corresponding optimization strategy is generated:

[0168] Resource scheduling optimization: When the membership degree of the performance efficiency dimension is lower than When this happens, the resource elastic expansion strategy will be automatically triggered, and the new CPU quota calculation formula is as follows:

[0169] ;

[0170] wherein, , is the resource elasticity coefficient, represents the membership value of performance efficiency dimension in the comprehensive evaluation vector.

[0171] Business process reconfiguration: dynamically adjust the order execution sequence according to task priority and device status, and the optimization objective function is:

[0172] ;

[0173] Constraint conditions:

[0174] ;

[0175] wherein, is the total number of orders in the order queue to be processed, is the order priority, is the decision variable (1 represents execution, and 0 represents non-execution), is the expected completion rate, is the resource consumption, is the maximum resource constraint.

[0176] Parameter configuration adaptation: based on the quality evaluation results, adjust the MOM software running parameters, including:

[0177] Database connection pool size: , wherein is the adjustment coefficient, is the basic connection number, is the membership value of reliability in the comprehensive evaluation vector under specific scenarios ;

[0178] Task processing thread number: dynamically adjust according to system load, when , the number of threads increases by 20%;

[0179] Cache strategy: adjust cache size and expiration time based on , and is the membership value of response speed in the comprehensive evaluation vector under specific scenarios .

[0180] After all instructions are executed, its effect is monitored in real time by the data acquisition preprocessing and feature extraction module, and feedback to the fuzzy comprehensive evaluation module, so as to form a complete, self-optimizing closed loop.

[0181] The application further provides a computer readable medium, which can be included in the electronic device described in the above embodiments, or can exist independently without being assembled into the electronic device.

[0182] The computer readable medium described above carries one or more programs, which, when executed by the electronic device, enable the electronic device to implement the method in the above embodiments.

[0183] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, the division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into several modules or units.

[0184] Through the above description of the embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.) or a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0185] Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the specification and practice of the present disclosure. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure following the general principles thereof and including the general and specific knowledge in the art not disclosed in the present disclosure.

[0186] The above is merely specific embodiments of the present application, but the design concept of the present application is not limited thereto, and any non-essential modification of the present application using the concept shall be deemed as an infringement of the protection scope of the present application.

Claims

1. A method for optimizing the quality of MOM software based on dynamic fuzzy scene mapping, characterized in that, include: Multi-source heterogeneous data on software quality was collected from multiple dimensions, including MOM system operation monitoring, user interaction records, and business process tracking. After cleaning and normalization, a standardized dataset was obtained. Based on the amount of interaction information as a feature relevance metric, a subset of core features strongly correlated with software quality was extracted from the standardized dataset. The specific expression for the amount of interaction information is as follows: ； in, It is a single feature variable to be selected from a standardized dataset. It refers to the specific values ​​that the characteristic variable X takes. It is a quality level variable. Representing characteristic variables With quality target variables The amount of information exchanged between them This indicates that the eigenvector X takes the value of And the quality grade variable takes the value of The joint probability distribution, and These represent the marginal probability distributions of feature variable X taking the value x and quality grade variable Y taking the value y, respectively. Based on system load status, user behavior patterns, business workflow characteristics, and different performance patterns of the core feature subset, software operation scenarios are defined and segmented. For each software operation scenario, a two-layer weight model containing basic weights and dynamic weights is constructed, and the weights are dynamically adjusted by combining historical experience and real-time feedback. Based on the core feature subset and the quality level of expert annotations, a structured training sample library is constructed. A random forest regression algorithm is used, and through a feature importance scoring mechanism, the first [feature] in the core feature subset is calculated. The quality characteristic in the first Initial base weights for each scenario ; For each quality feature in the core feature subset, based on the current scenario Based on the real-time evaluation error and scene change signals, the basic weights are adjusted through a weight adaptive update mechanism, the update formula of which is as follows: ； in, Indicates the adjusted number A quality characteristic in the scene Dynamic weights under; For adaptive learning rate parameters, This represents the evaluation error in the current scenario. This refers to the weight adjustment amount calculated based on real-time scene switching signals and a business priority rule base. The basic weights and dynamic weights are combined to obtain the scene. The combined weight of each quality feature in the core feature subset : ； in, This represents the basic weight of the j-th quality feature in scenario s. This represents the dynamic weight of the j-th quality feature in scenario s. Indicates the total number of quality characteristics; The adaptive learning rate parameter An annealing strategy is adopted, and the value gradually decreases as the number of evaluation iterations increases. The update rule is as follows: ； in, Indicates the first Learning rate during iteration The initial learning rate, The attenuation coefficient is... This represents the current iteration number. ; For each software operation scenario, a fuzzy mapping relationship between quality features in the core feature subset and preset evaluation levels is established to form a fuzzy judgment matrix. Based on the dynamically adjusted weight vector and the corresponding fuzzy judgment matrix, a synthesis operation is performed to obtain a scenario-based comprehensive quality evaluation vector, and the final quality level is determined according to the maximum membership principle. Based on the comprehensive quality evaluation vector, optimization instructions for software configuration, resource allocation, or workflow adjustment are generated and sent to the MOM system for execution to achieve closed-loop optimization of software quality.

2. The MOM software quality optimization method based on dynamic fuzzy scene mapping as described in claim 1, characterized in that, The cleaning and normalization processes yield a standardized dataset, specifically including the following: The multi-source heterogeneous data is cleaned, and then the min-max normalization method is used to linearly map the cleaned data to the [0,1] interval. The calculation formula is as follows: ； in, These are the normalized data values. For the cleaned data, and These are the minimum and maximum values ​​of the data set, respectively.

3. The MOM software quality optimization method based on dynamic fuzzy scene mapping as described in claim 1, characterized in that, For the scenario , will be comprehensive weight With fuzzy relation matrix Perform a synthesis operation to obtain the comprehensive evaluation vector for this scenario. ; ； in, This represents the weighted average composition operator; Based on the comprehensive evaluation vector The final quality level is determined using the principle of maximum membership, i.e., a comprehensive evaluation vector is selected. The evaluation level corresponding to the maximum value is used as the software quality assessment result in this scenario.

4. The MOM software quality optimization method based on dynamic fuzzy scene mapping as described in claim 1, characterized in that, Based on quality-driven dynamic optimization decision-making, at least one of the following optimization instructions is generated and executed according to the comprehensive evaluation vector: Resource scheduling command: Dynamically adjust the CPU quota or memory allocation of the container where the MOM software is located; Business process refactoring instruction: Rearrange the execution sequence of pending work orders according to task priority; Parameter configuration commands: Adjust the number of thread pools or database connection parameters related to software performance.

5. A MOM software quality optimization system based on dynamic fuzzy scene mapping, characterized in that, A method for optimizing MOM software quality based on dynamic fuzzy scene mapping, as described in any one of claims 1 to 4, includes: The data acquisition, preprocessing, and feature extraction module is used to collect multi-source heterogeneous data on software quality from multiple dimensions, including MOM system operation monitoring, user interaction records, and business process tracking. The module cleans and normalizes the collected heterogeneous data to obtain a standardized dataset, and extracts a subset of core features that are strongly correlated with software quality from the standardized dataset based on information measurement methods. The scenario definition and weight adjustment module is used to define and divide software operation scenarios based on system load status, user behavior patterns, business workflow characteristics, and different performance patterns of the core feature subset. For each software operation scenario, a two-layer weight model containing basic weights and dynamic weights is constructed, and the weights are dynamically adjusted by combining historical experience and real-time feedback. The fuzzy comprehensive evaluation module is used to establish a fuzzy mapping relationship between quality features in the core feature subset and preset evaluation levels for each software operation scenario to form a fuzzy judgment matrix. Based on the dynamically adjusted weight vector and the corresponding fuzzy judgment matrix, a synthesis operation is performed to obtain a scenario-based comprehensive quality evaluation vector. Finally, the final quality level is determined according to the maximum membership principle. The closed-loop optimization execution module is used to generate optimization instructions for software configuration, resource allocation or workflow adjustment based on the comprehensive quality evaluation vector, and send the optimization instructions to the MOM system for execution to achieve closed-loop optimization of software quality.

6. The MOM software quality optimization system based on dynamic fuzzy scene mapping as described in claim 5, characterized in that, Employing a microservices architecture and deploying it on a Kubernetes cluster using containerization technology, including: The data acquisition Pod is used to implement the multi-source heterogeneous data acquisition function in the data acquisition preprocessing and feature extraction module. The Feature Processing Pod is used to implement data cleaning, normalization, and core feature subset extraction functions in the data acquisition, preprocessing, and feature extraction modules. The Scene Recognition Pod is used to implement the software operation scene definition and classification functions in the Scene Definition and Weight Adjustment module; The Weight Optimization Pod is used to implement the two-layer weight model construction and dynamic adjustment function in the scenario definition and weight adjustment module. The Comprehensive Evaluation Pod is used to implement the functions of fuzzy judgment matrix construction, quality comprehensive evaluation vector calculation, and final quality level determination in the fuzzy comprehensive evaluation module. The strategy execution Pod is used to generate optimization instructions in the closed-loop optimization execution module and send them to the MOM system for execution. Each Pod communicates with the others through a service mesh to achieve distributed elastic scaling.

7. The MOM software quality optimization system based on dynamic fuzzy scene mapping as described in claim 5, characterized in that, The closed-loop optimization execution module includes a business optimization strategy library, which maps the final quality level and comprehensive quality evaluation vector into three types of executable instructions: Resource scheduling commands are used to dynamically adjust the CPU and memory resource limits of MOM software Pods via the Kubernetes API; Process refactoring instructions are used to rearrange the execution priority and path of production work orders through the workflow engine API of the MOM system; Parameter configuration commands are used to dynamically adjust key operating parameters such as database connection pools and thread pools through the configuration management center API.

Images