Quality data asset value quantitative evaluation method

Abstract

The invention provides a quality data asset value quantitative evaluation method, and the method comprises the steps: constructing a quality data asset value evaluation system, carrying out the data preparation and preprocessing, constructing a quality data asset value evaluation model based on an SO-BP neural network, and carrying out the parameter initialization, and optimizing and adjusting the quality data asset value evaluation model, and then predicting and evaluating the quality data asset value evaluation model. Based on the SO algorithm and the BP neural network, the value of the quality data is scientifically evaluated and estimated from multiple dimensions, the analysis efficiency is greatly improved, and computing resources and time cost are saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of quality evaluation, and in particular relates to a method for quantitatively evaluating the value of quality data assets. Background Art

[0002] With the continuous development of next-generation information technologies such as big data, the Internet of Things, cloud computing, and artificial intelligence, intelligent manufacturing is replacing traditional manufacturing and becoming one of the dominant models in the manufacturing industry. It's widely recognized that data is a valuable intangible asset. Enterprises and organizations are increasingly recognizing the importance of data assets and are committed to continuously exploring the potential and actual value they hold. Faced with the rapid growth of data volumes in the context of intelligent manufacturing and the urgent need to leverage and activate data to fully unlock its value, existing models, tools, and methods for quality data collection, management, and analysis are no longer adaptable and have become a bottleneck hindering the exploration and utilization of quality data's potential. For complex equipment development processes, due to the complexity of quality data sources, composition, and form, compared to tangible assets with specific market prices, there are few publicly available trading examples to refer to, a mature indicator system, and a systematic valuation model, making its value assessment extremely difficult.

[0003] Therefore, it is necessary to provide a quantitative assessment method for the value of quality data assets. Summary of the Invention

[0004] In order to solve the technical problem of the lack of a mature indicator system and a systematic evaluation model in the development process of complex equipment, and the great difficulty in quantitatively evaluating the value of quality data assets, the present invention provides a method for quantitatively evaluating the value of quality data assets, which greatly improves the analysis efficiency and saves computing resources and time costs. The technical solution is as follows:

[0005] First, a method for quantitatively evaluating the value of quality data assets is provided, including:

[0006] Construct a quality data asset value assessment system: Quantify the value of quality data assets, analyze and identify key information in quality data asset management, determine technical specification value assessment factors from multiple dimensions, and construct a corresponding quality data asset value assessment system based on the analysis of technical specification assessment factors;

[0007] Perform data preparation and preprocessing: Based on the inspection data of parts collected during the manufacturing process by the inspection information system, a sample set of training and test sets for the quality data asset value assessment model is formed. During the data preparation stage, knowledge graph technology is used to integrate and assetize quality data, and a knowledge graph data layer for the quality field is constructed.

[0008] The snake optimization algorithm is used to build a quality data asset value assessment model based on SO-BP neural network and initialize the parameters;

[0009] Optimize and adjust the quality data asset value assessment model;

[0010] Predict and evaluate the quality data asset value assessment model;

[0011] Multiple dimensions include: complexity, standardization, correctness, relevance, completeness, implementation, integration, maintainability and other dimensions. Other dimensions include development specification constraints and standard support.

[0012] The quality data asset value assessment model uses the following value calculation formula to calculate the quality data asset value:

[0013]

[0014] Where, , QDAV represents quality data asset value, C represents complexity, N represents standardization, V represents correctness, D represents relevance, P represents completeness, R represents implementation, I represents integration; M represents maintenance, and O represents other degrees.

[0015] Among them, the quality data in the production process includes the operating parameters of machine tools and equipment related to quality during the manufacturing process, online detection data, offline detection data after production is completed, and sensor data.

[0016] Among them, the snake optimization algorithm is used to build a quality data asset value assessment model based on BP neural network and perform parameter initialization, including:

[0017] A back-propagation BP neural network is constructed, and the weights, thresholds and activation functions in the BP neural network are initialized; the objective function of the BP neural network is set as the mean square error (MSE) between the expected output and the actual output; the pre-processed data values ​​are numerically input to the BP neural network, and the BP neural network is evolutionarily trained using the snake optimization algorithm to obtain a quality data asset value assessment model, which consists of an input layer, a hidden layer, an output layer and a value calculation module.

[0018] Among them, when optimizing and adjusting the quality data asset value assessment model, the optimal weights and thresholds obtained by evolutionary training of the snake optimization algorithm are used to further optimize the quality data asset value assessment model, including adjusting the network structure, learning rate and hyperparameters, to ensure that further optimization of the quality data asset value assessment model can meet the accuracy requirements while maintaining good generalization capabilities.

[0019] Among them, when predicting and evaluating the quality data asset value assessment model, the quality data asset value assessment model is used to predict and evaluate the quality data in the manufacturing process to obtain the eigenvalue of each dimension, and the value range of each eigenvalue is 0~10.

[0020] The beneficial effects of this application are at least:

[0021] This paper combines the SO algorithm and BP neural network to propose a quality data asset valuation model based on the SO-BP neural network. This model can scientifically evaluate and estimate the value of an enterprise's quality data from multiple dimensions. By leveraging the excellent fault tolerance and nonlinear mapping capabilities of the BP neural network and fully integrating the snake optimization algorithm's ability to search a wide range of spaces and seek the global optimal solution, the two complement each other and form an organic integration. Compared with traditional artificial intelligence algorithms, this model significantly improves analysis efficiency and saves computing resources and time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Schematic diagram of the optimization process for the quality data asset value assessment model;

[0023] Figure 2 This is a schematic diagram of the technical specification evaluation indicator system;

[0024] Figure 3 This is a schematic diagram of the quality data asset value assessment model;

[0025] Figure 4 is an example of original quality data;

[0026] Figure 5 This is a schematic diagram of the quality data resource cleaning process;

[0027] Figure 6 A partial display diagram of the manufacturing process quality data knowledge graph;

[0028] Figure 7 It is a schematic diagram of BP neural network;

[0029] Figure 8 Schematic diagram of the distribution of feature importance of neurons in the input layer. DETAILED DESCRIPTION

[0030] The present application is described in further detail below with reference to the accompanying drawings of the embodiments.

[0031] This embodiment of the present invention provides a method for quantitatively assessing the value of quality data assets based on an SO-BP neural network. This method leverages the fault tolerance and nonlinear mapping capabilities of BP neural networks and fully integrates the snake optimization algorithm's ability to search across a wide range of spaces and seek global optimal solutions, achieving complementary and organic integration of their strengths. This method enables scientific evaluation and estimation of the value of an enterprise's quality data from multiple dimensions, helping enterprises efficiently utilize their quality data assets.

[0032] The method described in the embodiment of the present invention is intended to quantify and value quality data assets. First, data from the inspection process is collected, with process inspection data used as the training set and general inspection data used as the test set. Then, a quality data asset value assessment model based on a BP neural network is constructed, and relevant weights, thresholds, and functions are initialized. The BP neural network objective function is the mean square error (MSE) of the expected value. The quality data asset value assessment model is trained through evolution of the snake optimization algorithm (SO) to obtain optimal weights and thresholds. The optimal weights and thresholds obtained through evolutionary training are used to further optimize the quality data asset value assessment model to obtain a quality data asset value assessment model that meets accuracy requirements. The trained quality data asset value assessment model is used to predict and estimate the test samples, ultimately obtaining a prediction output.

[0033] An embodiment of the present invention provides a method for quantitatively evaluating the value of quality data assets, comprising the following steps:

[0034] Step 1: Build a quality data asset value assessment system

[0035] First, starting from the micro level of data quality, the value of quality data assets is quantitatively evaluated, and key information in quality data asset management is analyzed and identified. The technical specification value assessment factors are sorted out from the dimensions of complexity, standardization, correctness, and relevance. Based on the analysis of the technical specification assessment factors, a corresponding quality data asset value assessment system is constructed.

[0036] Step 2: Data preparation and preprocessing

[0037] The training and test sample sets for the quality data asset valuation model are based on part inspection data collected by the inspection information system during the manufacturing process. Quality data from the production process includes quality-related machine tool operating parameters, online inspection data, offline inspection data after production, and sensor data. During the data preparation phase, necessary data preprocessing is performed, and knowledge graph technology is used to integrate and capitalize quality data. A quality domain knowledge graph data layer is constructed by combining data collection, data cleaning, knowledge extraction, knowledge fusion, and storage technologies to improve the efficiency and accuracy of model training.

[0038] Step 3: Construct a quality data asset value assessment model based on BP neural network and initialize parameters

[0039] A back-propagation (BP) neural network was constructed, and the relevant weights, thresholds, and activation functions within the network were initialized. The objective function of the BP neural network was set to the mean squared error (MSE) between the expected and actual outputs, a key metric for measuring the model's predictive accuracy. To optimize the network parameters, the snake optimization algorithm (SOA) was used to evolve and train the neural network, resulting in a quality data asset valuation model. SOA, an optimization algorithm designed to simulate the predatory behavior of snakes, effectively searches for and finds the global optimal solution within the parameter space.

[0040] Step 4: Optimize and adjust the quality data asset value assessment model

[0041] The quality data asset valuation model is further optimized using the optimal weights and thresholds obtained through evolutionary training using the snake optimization algorithm. This step involves adjusting the network structure, learning rate, and other hyperparameters to ensure that the model meets accuracy requirements while maintaining good generalization capabilities. Through this process, we aim to obtain an accurate and reliable quality data asset valuation model.

[0042] Step 5: Predict and evaluate the quality data asset value assessment model

[0043] Finally, the trained quality data asset valuation model is applied to the test sample set to generate a prediction estimate. By comparing the model's predicted output with the actual values, the model's predictive performance can be evaluated. This step is crucial for verifying the model's ability to perform well on unknown data.

[0044] Another embodiment of the present invention provides a method for quantitatively evaluating the value of quality data assets, which may specifically include the following steps:

[0045] Step 1: Build a quality data asset value assessment system

[0046] The technical specification evaluation factors are used to quantitatively evaluate the value of quality data assets from the micro level of data quality, which is conducive to the analysis and identification of key information in quality data asset management. The present invention sorts out the technical specification value evaluation factors from the dimensions of complexity, standardization, correctness, relevance, etc., and constructs a corresponding technical specification evaluation index system based on the analysis of the technical specification evaluation factors, namely the quality data asset value evaluation system, such as Figure 2 shown.

[0047] Complexity reflects the volume of quality data, broken down into the number of entities, attributes, and relationships. Complexity varies with different application scenarios, enterprise expansion, or the development of new models. Standardization interprets quality data from a database storage perspective, categorized as 1NF, 2NF, and 3NF. With the continuous advancement of informatization, structured quality data is increasing, and most of it is stored in relational databases. This necessitates the introduction of standardization to evaluate quality data. Correctness analyzes quality data from a business perspective for ambiguous expressions and data completeness. For example, for a supplier of the same enterprise, the supplier name recorded in the system may have a full name or a code to represent the supplier; relevance is to analyze the strength of the relationship between quality data from the perspective of association, which can be divided into the relevance of entities, attributes, relationships and constraints. For the analysis scenario of quality problems in quality management activities, it is hoped to establish the association relationship of quality problem concepts from multiple dimensions as much as possible, so as to facilitate the mining of quality problems; completeness mainly evaluates the value of quality data from the coverage of quality data, and judges whether there is a missing quality data object, whether the entity extraction is complete, and whether the objects are strongly constrained or weakly correlated. These are all aspects that need to be carefully considered; feasibility is used to evaluate The management of quality data, whether it can be collected efficiently, whether quality data needs to be classified and processed according to different granularities, and whether heterogeneous quality data can achieve compatibility; integration is to consider the integration of quality data from the perspective of multi-source integration; maintainability is to evaluate quality data from the perspective of system management, whether it is convenient and efficient to govern quality data to meet subsequent quality management, and use these data for mining and analysis to guide enterprise operations management; with the continuous introduction of smart devices into the product manufacturing process, different types of quality data will be generated. Based on this consideration, the technical specification evaluation index system can be subsequently expanded to introduce new data standards and data models to evaluate quality data.

[0048] According to the technical specification evaluation index system, the value of quality data assets can be measured using nine dimensions: complexity, standardization, correctness, relevance, completeness, feasibility, integration, maintainability, and other dimensions. This embodiment of the present invention defines these nine dimensions as having equal weights. Figure 3 Provides a pre-built quality data asset value assessment model.

[0049] Step 2: Data preparation and preprocessing

[0050] The data from the inspection process is selected as the data set to be input into the model. Specifically, the process inspection data is first screened out as the training set of the model to facilitate the model to learn the key features of the manufacturing process. At the same time, the total inspection data is selected as the test set to evaluate the prediction performance of the trained model in actual applications. The original quality data sample is as follows Figure 4 As shown, it is necessary to preprocess these data.

[0051] Most of the collected quality data is stored in relational databases. The data types are mainly divided into integer, floating point and other digital data, time and date data, fixed-length and variable-length text data. A considerable part of the quality data still has redundancy, conflict, incompleteness and other problems, which is not conducive to subsequent entity extraction and relationship fusion. In view of these characteristics of quality data in the process of complex equipment development, the present invention designs a quality data resource cleaning method, such as Figure 5 shown.

[0052] After the quality data resources are cleaned, all the entities and relationships in the quality field are entered into the knowledge graph database to form a quality field knowledge graph based on the manufacturing process quality data and a knowledge graph of the inspection data, which is used to analyze the relationship between the part and people, machines, materials, methods, environment, and measurement, such as Figure 6 shown.

[0053] The processed data set is input into the quality data asset value assessment model for training. The training of the quality data asset value assessment model based on the SO-BP neural network consists of multiple batches, each batch has 15 iterations, and each iteration requires 64 sets of data. The data of each batch will be randomly sorted and then input into the model for training.

[0054] Step 3: Build a quality data asset value assessment model and initialize parameters

[0055] BP neural network is a multi-layer feedforward network trained by the error back propagation algorithm. It is a widely used neural network algorithm. The network mainly optimizes the threshold and weight of the network based on back propagation to obtain the minimum sum of squared network errors. This network mainly covers three levels: input layer, hidden layer and output layer. The layers are connected by neurons, but there is no link between the same layer. Its structure is as follows: Figure 7 shown.

[0056] assumed is the transfer function connecting the hidden layer and the output layer, is the transfer function of the output layer, and the input layer is expressed as follows:

[0057]

[0058] The output of the hidden layer is expressed as follows:

[0059]

[0060] The output layer is represented as follows:

[0061]

[0062] Then, the error between the output value and the expected value in the neural network is expressed as follows:

[0063]

[0064] Where n is the number of neurons in the input layer, r is the number of neurons in the hidden layer, and m is the number of neurons in the output layer. is the weight of input layer node i and hidden layer node j, is the weight of hidden layer node j and output layer node k.

[0065] See also Figure 3 In the embodiment of the present invention, the quality data asset value assessment model consists of an input layer, a hidden layer, an output layer, and a value calculation module. The specific parameters are set as follows: 5 neural network layers, 28 input layer neurons consisting of subdivision dimension indicators corresponding to feature dimensions, 9 feature dimensions as output layer neurons, the number of neurons corresponding to each layer in the 3 hidden layers is 32, 64, and 32 respectively, and the error target value is set to 10 -4 , using ReLU as the activation function, the upper limit of the number of training times is 1000.

[0066] After selecting the above parameters, use Python software to train the quality data asset value assessment model. The quality data asset value assessment model is trained internally, and weights and thresholds are appropriately adjusted during the training process. Finally, the evaluation results of these nine dimensions are entered into the following formula, which outputs the quality data asset value.

[0067]

[0068] Where, , QDAV represents quality data asset value, C represents complexity, N represents standardization, V represents correctness, D represents relevance, P represents completeness, R represents implementation, I represents integration; M represents maintenance, and O represents other degrees.

[0069] Figure 8 This is the distribution diagram of the importance of the input layer neuron features. The horizontal axis represents the name of the evaluation element, and the vertical axis represents the importance of the evaluation element. Table 1 lists the specific values ​​corresponding to the importance of the elements from large to small.

[0070] Table 1 Specific values ​​corresponding to the importance of elements

[0071]

[0072] Step 4: Optimize and adjust the quality data asset value assessment model

[0073] The quality data asset valuation model is further optimized using the optimal weights and thresholds obtained through evolutionary training using the snake optimization algorithm. This step involves adjusting the network structure, learning rate, and other hyperparameters to ensure that the optimization model meets accuracy requirements while maintaining good generalization capabilities. This process results in an accurate and reliable quality data asset valuation model.

[0074] The SO (Snake Optimizer) algorithm works as follows: Snake mating is influenced by two factors: ambient temperature and availability of food. If the temperature is low and food is plentiful, mating occurs; otherwise, the snakes simply search for food or consume any leftovers.

[0075] See also Figure 1 , the mathematical description of this process is as follows:

[0076] 1) Global search (no food)

[0077] If Q < threshold (threshold = 0.25), the snake searches for food by selecting any random location and updates its position in turn. The male snake position update formula is as follows:

[0078]

[0079] in, refers to the ith male position, Refers to the random position of the male, rand is a random number ranging from 0 to 1, Represents the male's ability to find food.

[0080] The female snake position update formula is as follows:

[0081]

[0082] in, refers to the ith female position, Refers to the random position of the female, rand is a random number ranging from 0 to 1, Represents the female's ability to find food.

[0083] 2) Local search (where there is food)

[0084] If Q > threshold and temperature > threshold (0.6), then the snake will only move towards the food, as shown in the following formula:

[0085]

[0086] in, is the position of the male or female individual, represents the optimal position of an individual, The value is 2. If Q>0.25 and temperature<threshold (0.6), then the snake will enter fighting mode or mating mode.

[0087] The formula for updating the position of the male snake in combat mode is as follows:

[0088]

[0089] in, refers to the ith male position, represents the individual in the best position in the female set, It represents the fighting ability of male snakes.

[0090] The formula for updating the position of the female snake in combat mode is as follows:

[0091]

[0092] in, represents the i-th female position, represents the individual in the best position in the male set, Represents the fighting ability of female snakes.

[0093] In the mating mode, if the snake egg hatches, the female with the worst ability is selected or replaced with the male. The formula is as follows:

[0094]

[0095]

[0096] in, Indicates the least capable individual in the male snake group. Indicates the least capable individual in a group of female snakes.

[0097] Step 5: Predict and evaluate the quality data asset value assessment model

[0098] By analyzing the results of the quality data in the manufacturing process in steps 3 and 4 and predicting and evaluating the quality data asset value assessment model, we can obtain the eigenvalues ​​of several dimensions. The value range of each eigenvalue is specified to be 0~10. These nine eigenvalues ​​provide the basis for analyzing the value of quality data assets, as shown in Table 2.

[0099] When using the quality data asset value assessment model to analyze the value of technical specification quality data assets, the utilization of quality data is better, and it can assist enterprises in analyzing the value of technical quality data assets in the quality management process.

[0100] Table 2 Model training results

[0101]

[0102] This paper combines the SO algorithm and BP neural network to propose a quality data asset valuation model based on the SO-BP neural network. This model can scientifically evaluate and estimate the value of an enterprise's quality data from multiple dimensions. By leveraging the excellent fault tolerance and nonlinear mapping capabilities of the BP neural network and fully integrating the snake optimization algorithm's ability to search a wide range of spaces and seek the global optimal solution, the two complement each other and form an organic integration. Compared with traditional artificial intelligence algorithms, this model significantly improves analysis efficiency and saves computing resources and time costs.

Claims

1. A method for quantitatively evaluating the value of quality data assets, characterized in that: include: Construct a quality data asset value assessment system: Quantify the value of quality data assets, analyze and identify key information in quality data asset management, determine technical specification value assessment factors from multiple dimensions, and construct a corresponding quality data asset value assessment system based on the analysis of technical specification assessment factors; Perform data preparation and preprocessing: Generate sample sets for training and testing of the quality data asset value assessment model based on the inspection data of parts in the manufacturing process collected by the inspection information system; In the data preparation stage, knowledge graph technology is used to integrate and capitalize quality data, and to build a knowledge graph data layer in the quality field; The snake optimization algorithm is used to build a quality data asset value assessment model based on SO-BP neural network and initialize the parameters; Optimize and adjust the quality data asset value assessment model; Predict and evaluate the quality data asset value assessment model; Multiple dimensions include: complexity, standardization, correctness, relevance, completeness, implementation, integration, maintainability and other dimensions. Other dimensions include development specification constraints and standard support. The quality data asset value assessment model uses the following value calculation formula to calculate the quality data asset value: , where , QDAV represents quality data asset value, C represents complexity, N represents standardization, V represents correctness, D represents relevance, P represents completeness, R represents implementation, I represents integration; M represents maintenance, and O represents other degrees.

2. The method according to claim 1, wherein The quality data in the production process includes the operating parameters of machine tools and equipment related to quality during the manufacturing process, online detection data, offline detection data after production is completed, and sensor data.

3. The method according to claim 1, wherein The snake optimization algorithm is used to build a quality data asset value assessment model based on BP neural network and perform parameter initialization, including: A back-propagation BP neural network is constructed, and the weights, thresholds and activation functions in the BP neural network are initialized; the objective function of the BP neural network is set as the mean square error (MSE) between the expected output and the actual output; the pre-processed data values ​​are numerically input to the BP neural network, and the BP neural network is evolutionarily trained using the snake optimization algorithm to obtain a quality data asset value assessment model, which consists of an input layer, a hidden layer, an output layer and a value calculation module.

4. The method according to claim 1, wherein When optimizing and adjusting the quality data asset value assessment model, the optimal weights and thresholds obtained by evolutionary training using the snake optimization algorithm are used to further optimize the quality data asset value assessment model, including adjusting the network structure, learning rate, and hyperparameters, to ensure that further optimization of the quality data asset value assessment model can maintain good generalization capabilities while meeting the accuracy requirements.

5. The method according to claim 1, wherein When predicting and evaluating the quality data asset value assessment model, the quality data in the manufacturing process is predicted and evaluated using the quality data asset value assessment model to obtain the eigenvalue of each dimension, and the value range of each eigenvalue is 0~10.

Images