Industrial equipment health degree evaluation method, system, equipment, medium and product

Abstract

The invention relates to the field of industrial equipment evaluation, and provides an industrial equipment health degree evaluation method and system, equipment, a medium and a product. The industrial equipment health degree evaluation method comprises the steps of obtaining health degree scoring index data of industrial equipment, and constructing a health degree scoring model; constructing a health degree scoring index data set of the industrial equipment, and calculating an initial weight of health degree scoring index data by adopting an extended entropy weight method; according to the initial weight of the health degree scoring index data and the health degree scoring index data of the industrial equipment, a health degree scoring model is adopted to calculate an initial health degree score; generating the adjustment amount of the preliminary comprehensive health degree score and the optimized health degree score by adopting the health degree score index data set of the industrial equipment, and carrying out iterative optimization on the weight to obtain an optimal weight; and obtaining a final health degree score according to the optimal weight and the weighted sum of the health degree score index data of the industrial equipment. And the accuracy of evaluating the health degree of the industrial equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment evaluation, and in particular, to a method, system, device, medium and product for evaluating the health of industrial equipment. Background Art

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] In the industrial field, the technology of equipment health state monitoring and evaluation (PHM) is the core means to ensure production safety and reduce operation and maintenance costs.

[0004] Traditional industrial equipment health state evaluation methods use the expert's experience and knowledge as the main basis for generating health score, and use artificial intelligence methods to generate a regression model based on a large amount of artificial data. However, this kind of artificial intelligence regression model is often one-time, that is, the generated model has no room for further improvement. That is, only by changing the data set, readjusting the hyperparameters or changing the artificial intelligence method used, can a model that is more in line with the user's cognition be generated compared with the existing model. Moreover, this method does not notice that for the same analysis object, the data given by different experts may be contradictory, such as the linear irrelevance between the data given by two experts. In principle, the interpretability is poor. This makes the regression model generated by the artificial intelligence model often have a large error with the subjective cognition of the operation and maintenance personnel, and the results obtained cannot convince the operation and maintenance personnel. Summary of the Invention

[0005] In order to solve the technical problems existing in the above background art, the present invention provides a method, system, device, medium and product for evaluating the health of industrial equipment. The present invention is particularly suitable for single-user industrial application scenarios, and only needs a small amount of data and only needs to adjust individual parameters to continuously optimize a health score model, so that the obtained health score result is more in line with the user's subjective cognition and improves the accuracy of industrial equipment health evaluation.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: The first aspect of the present invention provides a method for evaluating the health of industrial equipment.

[0007] A method for evaluating the health of industrial equipment includes: Obtaining the health score index data of industrial equipment and constructing a health score model; Constructing a health score index data set of industrial equipment and calculating the initial weight of the health score index data by using the extended entropy weight method; According to the initial weights of the health score index data and the health score index data of industrial equipment, a preliminary health score is calculated using a health score model; Using the health score index data set of industrial equipment, an adjustment amount of the preliminary comprehensive health score and an optimized health score are generated to iteratively optimize the weights and obtain the optimal weights; According to the optimal weights and the weighted sum of the health score index data of industrial equipment, the final health score is obtained.

[0008] Furthermore, the industrial equipment includes manufacturing equipment, energy industry equipment, chemical industry equipment, metallurgical industry equipment, and transportation industry equipment; if it is a five-axis linkage CNC machine tool in manufacturing equipment, the health score index data includes: spindle speed stability, spindle bearing temperature, hydraulic system oil viscosity, feed axis vibration intensity, and positioning accuracy; if it is a blast furnace blower in metallurgical industry equipment, the health score index data includes: fan efficiency, bearing temperature, vibration intensity, lubricating oil viscosity, outlet air pressure, motor current deviation rate, impeller dynamic balance deviation, and cooling water temperature.

[0009] Furthermore, the initial weights of the health score index data are obtained by multiplying the weights of the multi-level upper nodes of the health score index data.

[0010] Furthermore, the method for constructing the health score index data set of industrial equipment includes: Select Q pieces of health score index data of industrial equipment to construct a data matrix; Perform elementary row transformation on the data matrix to verify whether the rank of the data matrix is R, and determine which data are linearly related and which data are linearly independent; If the number of linearly independent data is equal to R, only one piece of data from each pair of linearly related multiple data in the data matrix is retained, and the rest are deleted; if the number of linearly independent data is greater than R, the linearly independent data exceeding R pieces are deleted, and only one piece of data from each pair of linearly related multiple data in the data matrix is retained, and the rest are deleted; if the number of linearly independent data is less than R, linearly independent data are supplemented to R pieces; Repeat the above steps to construct multiple data matrices to obtain the health score index data set of industrial equipment.

[0011] Furthermore, the method for calculating the initial weights of the health score index data using the extended entropy weight method includes: Standardize the health score index data of each industrial equipment to obtain extremely large indicators, extremely small indicators, intermediate indicators, and interval indicators, so as to construct a data group of the health score index data of each industrial equipment; Calculate the proportionality coefficient of each health - degree scoring index according to the data set of the health - degree scoring indexes of each industrial device; Calculate the information entropy of each health - degree scoring index according to the proportionality coefficient of each health - degree scoring index; Calculate the initial weight of the data of each health - degree scoring index according to the information entropy of each health - degree scoring index and the total number of health - degree scoring indexes;

[0012] Furthermore, the method of using the data set of the health - degree scoring indexes of industrial devices to generate the adjustment amount of the preliminary comprehensive health - degree score and the optimized health - degree score to iteratively optimize the weight to obtain the optimal weight includes: Generate the adjustment amount of the preliminary comprehensive health - degree score according to the preliminary comprehensive health - degree score and perform the iterative optimization process; During the iteration process, generate the preliminary comprehensive health - degree score of the current iteration number according to the weight of the previous iteration number and the data set of the health - degree scoring indexes of the industrial device of the current iteration number; generate the adjustment amount of the health - degree score of the current iteration number according to the preliminary comprehensive health - degree score of the current iteration number, and calculate the optimized comprehensive health - degree score of the current iteration number; furthermore, calculate the weight of the current iteration number according to the data set of the health - degree scoring indexes of the industrial device of the current iteration number and the calculated optimized comprehensive health - degree score; After the iteration is completed, select the optimal weight from the weights optimized in all iteration numbers according to the evaluation index.

[0013] The second aspect of the present invention provides an industrial device health - degree evaluation system.

[0014] An industrial device health - degree evaluation system includes: A data acquisition module, which is configured to: acquire the health - degree scoring index data of industrial devices and construct a health - degree scoring model; An initial weight calculation module, which is configured to: construct a data set of the health - degree scoring indexes of industrial devices and calculate the initial weight of the health - degree scoring index data by using the extended entropy weight method; A health - degree score calculation module, which is configured to: calculate the preliminary health - degree score by using the health - degree scoring model according to the initial weight of the health - degree scoring index data and the health - degree scoring index data of industrial devices; A weight optimization module, which is configured to: use the data set of the health - degree scoring indexes of industrial devices to generate the adjustment amount of the preliminary comprehensive health - degree score and the optimized health - degree score to iteratively optimize the weight to obtain the optimal weight; An output module, which is configured to: obtain the final health - degree score according to the weighted sum of the optimal weight and the health - degree scoring index data of industrial devices.

[0015] The third aspect of the present invention provides a computer device, which includes: a processor adapted to execute a computer program; a computer-readable storage medium storing a computer program, which when executed by the processor, implements the steps in the industrial equipment health evaluation method described in the first aspect above.

[0016] The fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which is adapted to be loaded and executed by a processor to implement the steps in the industrial equipment health evaluation method described in the first aspect above.

[0017] The fifth aspect of the present invention provides a computer program product or a computer program.

[0018] The present invention provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the industrial equipment health evaluation method described in the first aspect above.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides an industrial equipment health evaluation method, system, device, medium and product, which obtains the health score index data of industrial equipment and constructs a health score model; constructs a health score index data set of industrial equipment, and calculates the initial weight of the health score index data by using the extended entropy weight method; according to the initial weight of the health score index data and the health score index data of industrial equipment, uses the health score model to calculate a preliminary health score; uses the health score index data set of industrial equipment to generate an adjustment amount for the preliminary comprehensive health score and an optimized health score, so as to iteratively optimize the weight to obtain an optimal weight; and obtains a final health score according to the weighted sum of the optimal weight and the health score index data of industrial equipment. The present invention is used to guide users to establish a health score model in an industrial scenario, and can provide an objective initial weight. If the user is not satisfied with the health evaluation result, based on the subjective evaluation method, through the interaction between the machine and the user, continuously optimize the health score model until the health score result meets the user's requirements and then stop, improving the accuracy of the industrial equipment health evaluation. And there needs to be a clear index to quantify the user's satisfaction with the optimized health evaluation model.

[0020] The present invention can greatly reduce the amount of training data. Only a dozen or so pieces of data are needed to optimize a result that satisfies the operation and maintenance personnel very much. While reducing the amount of data calculation, the accuracy of evaluation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0022] Figure 1 is a flowchart of a method for evaluating the health of industrial equipment shown in an embodiment of the present invention; Figure 2 is a flowchart of a method for optimizing the health weight shown in an embodiment of the present invention; Figure 3 is an example diagram of a tree model shown in an embodiment of the present invention; Figure 4 is an example diagram of a tree model and the weights of each index shown in an embodiment of the present invention; Figure 5 is an example diagram of a tree model of the health of a five-axis linkage numerically controlled machine tool in a certain factory shown in an embodiment of the present invention; Figure 6 is a flowchart of how to make a qualified data set shown in an embodiment of the present invention; Figure 7 is a flowchart of a method for optimizing weights shown in an embodiment of the present invention; Figure 8 is a structural diagram of an industrial equipment health evaluation system shown in an embodiment of the present invention; Figure 9 is a structural diagram of a computer device shown in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0025] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Figure 1 is the flowchart of the industrial equipment health assessment method shown in the embodiments of the present invention. Referring to Figure 1 , the method includes: Obtain the health assessment index data of industrial equipment and construct a health assessment model; Construct a health assessment index data set of industrial equipment and calculate the initial weights of the health assessment index data by using the extended entropy weight method; According to the initial weights of the health assessment index data and the health assessment index data of industrial equipment, use the health assessment model to calculate the preliminary health assessment score; Use the health assessment index data set of industrial equipment to generate the adjustment amount of the preliminary comprehensive health assessment score and the optimized health assessment score to iteratively optimize the weights and obtain the optimal weights; According to the optimal weights and the weighted sum of the health assessment index data of industrial equipment, obtain the final health assessment score.

[0027] The present invention is particularly suitable for the industrial application scenario facing a single user. Only a very small amount of data is required and only individual parameters need to be adjusted to continuously optimize on a health assessment model, so that the obtained health assessment result is more in line with the subjective cognition of users, improving the accuracy of industrial equipment health assessment.

[0028] The industrial equipment health assessment method proposed by the present invention is applicable to scenarios where the health status of equipment or systems needs to be evaluated in various industrial fields, mainly including the following fields: Manufacturing industry: Health assessment of key equipment such as numerically controlled machine tools, industrial robots, and machining centers.

[0029] Energy industry: Condition monitoring and health management of power generation sets, wind turbines, compressors, power transmission and transformation equipment, etc.

[0030] Chemical industry: Health analysis of equipment such as reaction kettles, pressure vessels, and pipeline systems to prevent accidents such as leakage and explosion.

[0031] Metallurgical industry: Health status assessment of blast furnaces, rolling mills, smelting equipment, etc. to ensure production continuity.

[0032] Transportation: Health monitoring of key components of rail transit vehicles (such as traction motors, gearboxes), aero-engines, etc.

[0033] For a five-axis simultaneous control CNC machine tool in manufacturing equipment, the health score index data includes: spindle speed stability, spindle bearing temperature, hydraulic system oil viscosity, feed axis vibration intensity, and positioning accuracy; for a blast furnace blower in metallurgical industry equipment, the health score index data includes: blower efficiency, bearing temperature, vibration intensity, lubricating oil viscosity, outlet air pressure, motor current deviation rate, impeller dynamic balance deviation, and cooling water temperature.

[0034] The following will take the specific application and data of a five-axis simultaneous control CNC machine tool as an example for illustration: Figure 2 is a flowchart of the health weight optimization method shown in an embodiment of the present invention. Referring to Figure 2 , the method includes: First, introduce how to establish a health analysis model based on the mechanism of equipment or system health analysis in the industrial field.

[0035] Mechanism-based modeling means that the user clearly knows the evaluation indicators and their hierarchical structures required for health scoring, and can establish a health scoring model according to the subordination relationship of indicators at different levels. In this case, the user should define and operate in the user operation module to complete the construction of the health scoring model. The health scoring model can be intuitively displayed to the user in the form of a tree. An example of the tree model is shown in Figure 3 as follows: In the calculation module, the health scoring model defined by the user is converted into a form that can be calculated by a computer for use in the weight optimization method described later. Next, introduce how to solve this model and perform corresponding mathematical derivations: To determine the weights of each indicator, several specific terms are defined as follows, and an example is given using a five-axis simultaneous control CNC machine tool in a certain factory: (1) Terminal indicator: refers to a health scoring indicator that cannot be further broken down into sub-indicators. In the tree structure of the model, it refers to the lowest-level indicator without branches. When modeling, the health score value of the terminal indicator is marked as , the total number of terminal indicators is R, and the sorting of each terminal indicator is r, . For example, in the model of Figure 4 , there are 5 terminal indicators in total, and the health scores of these terminal indicators are respectively , , , and . There are various ways to obtain the health score of terminal indicators. For example, vibration intensity is scored by dividing intervals according to international standards, temperature is referenced based on the normal range and fluctuations, performance is compared with design parameters to see the decline rate, and failure rate is counted based on the number of failures and repair duration. Without considering how to obtain the health score of a terminal indicator, assuming that the health scores of all terminal indicators have been obtained through preset existing calculation methods and collected data, the present invention only needs to use these scores for subsequent calculations.

[0036] (2) Node: It is a health score indicator that can be further broken down into sub-indicators. In the tree structure diagram of the model, it refers to a health score indicator with branches below. For example, Figure 4 in which indicator a can be considered a node because it can be further broken down into sub-indicators and .

[0037] (3) Branch: It is a sub-indicator disassembled at a certain node. In the tree structure diagram of the model, if a certain node can be further disassembled into sub-indicators, then the sub-indicator can be called the branch of that node. For example, Figure 4 in which the terminal indicators and are two branches of node indicator a.

[0038] (4) Sub-weight: It is the strength of the linear relationship between a node and its sub-indicators, and the sum of the weights of multiple sub-indicators under the same branch is 1. For example, Figure 4 in which two sub-indicators and are divided at node indicator a, then their sub-weights are respectively and , and and satisfy the relationship: . The subscript under the sub-weight represents the number of the previous node and the current branch, and the superscript represents the linear relationship between the layers in the tree structure.

[0039] For Figure 5 , taking the health analysis of a five-axis linkage numerically controlled machine tool in a certain factory as an example, based on the mechanism, the health state of the five-axis linkage numerically controlled machine tool can be evaluated from aspects such as the spindle system and the feed system, and each aspect contains several specific indicators, forming a clear hierarchical structure. The user defines and constructs the following tree-shaped health score model in the user operation module: Spindle system (node A), branches are spindle speed stability (terminal indicator a1), spindle bearing temperature (terminal indicator a2), hydraulic system oil viscosity (terminal indicator a3), and their sub-weights are respectively 0.2, 0.3, 0.5; The feed system (node B) branches into the vibration intensity of the feed axis (terminal index a4) and the positioning accuracy (terminal index a5), and their weights are respectively 0.5, 0.5. The sub - weight of node A is , and the sub - weight of node B . The above sub - weights satisfy the relationships: , and . In this model, the terminal indices are the indices without branches at the bottom layer of the tree - like structure, such as the spindle speed stability, the positioning accuracy of each axis, etc. Their health score values are a1, a2, a3, a4, a5 in this example, and the total number R = 5.

[0040] (5) Weight is the strength of the linear relationship between the terminal index or node and the comprehensive health in constructing a mathematical model. Define the weight corresponding to the terminal index as , for example Figure 4 in the terminal index has a weight of , and the weight of the node index a is .

[0041] (6) Weight path is the link formed by connecting the sub - weights between each layer from the comprehensive health to a certain terminal index. For example Figure 4 in the terminal index , its weight path is to and then to . L is the total number of layers of this model. For an irregular health score model, the number of layers of any weight path is not necessarily L, but only the maximum is L. It should be noted that the weight path of any terminal index conforms to the chain rule and can be described by a mathematical expression in the form of a mapping: . Among them represents the sub - weight between the -1th layer and the l th layer on the weight path of l . For example Figure 4 in which the total number of layers L is 4, taking and two different weights to illustrate the calculation method of the weight. Among them passes through 4 layers, but only passes through 3 layers. In Figure 4 :

[0042] This mapping relationship is reflected as: in the expression , and respectively correspond to Figure 4 the sub-weights on the weight path in . Similarly:

[0043] In the expression and respectively correspond to Figure 4 the sub-weights on the weight path in and .

[0044] By solving the weight values of the terminal indicators step by step, the weight values of the upper layer are deduced backwards until all sub-weights are obtained. The reasoning order needs to start from the terminal indicators. First, the terminal indicators of a certain weight path need to be found. Then, find the upper-level node corresponding to the terminal indicator (or branch), and then sum up the weights (or sub-weights) of all terminal indicators (or branches) under the same node, that is, the sub-weights of the upper-level node are obtained. Deduce upwards level by level until all sub-weights are obtained and then terminate.

[0045] For example Figure 4 in, in the first step, find the node in the third layer - indicator a and its branches in the fourth layer - indicator and , and their weight relationship satisfies the following formula:

[0046]

[0047]

[0048] Therefore, the weight of the node - indicator a is deduced, and the sub-weights of the fourth layer and . Further, for the node in the second layer - preliminary health degree A, since there is only one branch in its third layer - indicator a, so the sub-weights of the second layer and the third layer , and the weight of the node in the second layer . By adopting this method and generalizing it to other weight paths, all sub-weights in this model, as well as the weights of all nodes and terminal indicators, can be deduced.

[0049] For example, Figure 5 taking the health degree analysis of a five-axis linkage numerically controlled machine tool in a certain factory as an example, among which the weight path of indicator a1 relative to the comprehensive health degree of the five-axis linkage numerically controlled machine tool , 0.2, it can be deduced that the weight of index a1 relative to the comprehensive health of the five-axis linkage CNC machine tool . The weights of other indicators a2, a3, a4, and a5 can all be calculated by this method:

[0050]

[0051]

[0052]

[0053] Finally, after the model is constructed, only by knowing the total number R of terminal indicators, the health assessment model can be constructed to obtain the comprehensive health score :

[0054] For example Figure 5 in, the weight model based on mechanism modeling can be expressed as:

[0055] For example, Figure 5 the comprehensive health score of a five-axis linkage CNC machine tool in a certain factory in is:

[0056] This method based on mechanism modeling, through a clear index hierarchy and mathematical derivation, can accurately reflect the influence degree of each part of the equipment on the overall health, providing a scientific basis for the state monitoring and maintenance decision-making of the equipment. In practical industrial applications, according to different types of equipment and specific analysis requirements, the corresponding health score model can be flexibly defined and constructed, and the weights can be solved by the above method to achieve the accurate assessment of the equipment health.

[0057] According to the above mathematical derivation, such a principle is determined: only by knowing the weights of all terminal indicators, the sub-weights on all weight paths in the entire model and the weights of each indicator can be determined. Therefore, it is necessary to solve the weights of all terminal indicators by relying on the data set collected under real conditions. Given a model as:

[0058] Rewritten in matrix form as:

[0059] In this model, there are a total of R terminal indicators, then the weight Considered as variables, there are correspondingly R variables. Therefore, to find this weight matrix, R linearly independent data are needed. Each piece of data includes the health score of R terminal indicators, and they are combined into a data group matrix A of rank R. The complete form is:

[0060] Among them, represents the health score value of the r-th terminal indicator in the j-th data in the data group, ; is the comprehensive health score obtained based on the j-th data and the weight . Multiply both sides of the formula by an inverse matrix of A:

[0061] It is obtained that:

[0062] According to the above derivation, to determine a model weight, a set of data sets is required, where each piece of data contains the score values of each terminal indicator, and the matrix composed of this set of data sets has a rank of R. Therefore, when making the data sets, the collected data are grouped, and the Gaussian elimination method is used within each group to make N linearly independent data sets of rank R for backup. These N data sets are uniformly represented in the following format, where the superscript n represents the n-th data set among these N data sets, and the mathematical form is as follows:

[0063] In the formula, represents the n-th data group, ; represents the health score value of the r-th terminal indicator in the j-th data in the n-th data group, .

[0064] Taking Figure 5 a five-axis linkage numerically controlled machine tool in a certain automotive parts factory as an example, it is necessary to make a data set for its health model for subsequent optimization of the terminal indicator weights. The machine tool is sampled 10 times a day for 30 days, and a total of 10×30 = 300 pieces of data are obtained. For the convenience of understanding, the health scores directly based on the collected time series data are shown using each health score indicator as shown in Table 1. As described above, the process of converting the time series data into specific scores is omitted. Table 1 Example of Terminal Indicator Health Score Data (Partial)

[0065] Figure 6It is a flowchart showing how to create a qualified dataset in an embodiment of the present invention. Referring to Figure 6 , the specific method for creating the dataset is as follows: Step A: Select data from the existing data and generate a data group A containing Q (and Q ) pieces of data.

[0066] For example, randomly select Q = 10 (Q ≥ R = 5) from 300 pieces of data to form the initial version of the first data group , represents the health score value of the r-th terminal index in the j-th piece of data. In matrix form (for the convenience of readers' understanding, it is presented in the form of each piece of data instead of directly giving the matrix form. The following content is the same):

[0067] Step B: Verify whether the rank of these Q pieces of data is R through elementary row operations (row swapping, row multiplication, and row addition). At the same time, it can be determined which pieces of data are linearly related and which are linearly independent. If there are NLC pieces of linearly independent data among them.

[0068] For example, the data group is as follows:

[0069] For each piece of data in, perform elimination through elementary row operations (row swapping, row multiplication, and row addition). The following are the specific steps of the elementary row operation process (column-by-column elimination) of this matrix: (1) Select the pivot row and perform elimination on the first column. Select row 1 (the row with the largest non-zero element in the first row) as the pivot row, and perform row operations on rows 2 to 10: row j = row j - (row j[column 1] / row 1[column 1]) × row 1. After the transformation, the column 1 of rows 2 to 10 becomes 0 or approximately 0. For example, after row 2 is transformed:

[0070] (2) Continue to select the pivot row and perform elimination on the second column. The value of column 2 in row 3 (90) is the largest, so it is selected as the new pivot row, and perform operations on rows 2, 4 to 10: row j = row j - (row j[column 2] / row 3[column 2]) × row 3. After row 2 is transformed, column 2 becomes 0, and column 2 of row 4 becomes: 70 - (70 / 90) × 90 = 0 (3) Continue to select the pivot row and perform elimination on the third column. The value of column 3 in row 4 (92) is the largest, and perform elimination: row j = row j - (row j[column 3] / row 4[column 3]) × row 4. Column 3 of row 5 becomes: 90 - (90 / 92) × 92 = 0 (4) Continue with the pivotal row selection and elimination for the fourth column. The value in row 5, column 4 (73) is non-zero. After elimination, the value in row 8, column 4 becomes: 67 - (67 / 73)×73 = 0 (5) Continue with the pivotal row selection and elimination for the fifth column. The value in row 8, column 5 (89) is non-zero. After elimination, the value in row 9, column 5 becomes: 92 - (92 / 89)×89 = 0 (6) Finally, after row transformation, the non-zero rows (linearly independent rows) are: row 1 (data set 1), row 3 (data set 3), row 4 (data set 4), row 5 (data set 5), row 8 (data set 8), row 9 (data set 9), = 6; the zero rows (linearly dependent rows) are rows 2, 6, 7, 10.

[0071] Step C: Perform the following operations based on the number of linearly independent data: (1) If , it means the current data set exactly meets the requirements. Only retain one piece of each pair of linearly dependent multiple data in the data set, and delete the rest. After completion, a qualified data set is obtained, and there are R pieces of data in this data set.

[0072] (2) If , it means some linearly independent data need to be removed. The number of linearly independent data to be removed is . At the same time, only retain one piece of each pair of linearly dependent multiple data in the data set, and delete the rest. After completing these two deletion operations, a qualified data set is obtained, and there are R pieces of data in this data set.

[0073] (3) If , linearly independent data need to be supplemented. The number of linearly independent data to be supplemented is at least pieces. After supplementation, return to step B.

[0074] Continuing the above, because of , so NLC - R = 1 should be deleted, that is, one linearly independent row, to form the final first data set , here assume that row 9 (data set 9) is deleted, and rows 1 (data set 1), 3 (data set 3), 4 (data set 4), 5 (data set 5), 8 (data set 8) are retained, then finally is as follows:

[0075] Step D: Repeat the operations in the above Steps A to C to continue producing the second group, the third group, and so on of data groups until the user believes that a sufficient number of data groups have been generated. Denote the number of data groups at this time as N, and these N data groups constitute the required data training set.

[0076] In the example, repeat the above steps to continue producing the second group, the third group of data sets, etc. from the remaining 290 pieces of data until the user obtains a satisfactory number of data sets.

[0077] The traditional entropy weight method can only normalize and determine the weights for extremely large type indicators and extremely small type indicators. The present invention proposes an extended entropy weight method for the situation where four main types of indicators may coexist in the health score model under real conditions. The four types of indicators are extremely large type indicators, extremely small type indicators, intermediate type indicators, and interval type indicators, and the weights are determined by an objective method. The introduction and examples are shown in Table 2.

[0078] Table 2 Indicator Types

[0079] For Figure 5 a five-axis linkage CNC machine tool in a certain automobile parts factory, it is necessary to determine the initial weights through the extended entropy weight method to provide an objective basis for subsequent optimization of the health indicator weights; Table 3 shows the indicator types of the five-axis linkage CNC machine tool, as shown in Table 3: Table 3 Indicator Types of Five-Axis Linkage CNC Machine Tool

[0080] Among them, the specific steps for determining the initial weights by the extended entropy weight method are as follows: Assume that in the first group of data there are R indicators and R pieces of data, where is the value in the j-th data under the r-th indicator, that is, the data in the j-th row and the r-th column, :

[0081] Step a: Standardize each indicator in turn. Let for the r-th terminal indicator, that is, when is the original data of data j under the r-th indicator, is the minimum value of data j, is the maximum value of data j, is the standardized data, then: Extremely large type indicator:

[0082] Extremely small type indicator:

[0083] Intermediate type index:

[0084] Among them, is the best value, and the maximum difference value in this data j . Interval type index: , where the best interval of is , and the maximum distance from the maximum or minimum value to the boundary of the best interval .

[0085] Taking the first set of data produced as an example, taking the original data value of the first set of data (the first row) among them as an example to illustrate the process of standardizing each terminal index. Table 4 shows the first row standardization process of the first set of data of the five-axis linkage numerically controlled machine tool displayed as follows: Table 4 First row standardization process display of the first set of data of the five-axis linkage numerically controlled machine tool displayed for the first row

[0086] According to the steps described in the above table, sequentially complete the standardization of the original data of the 2nd, 3rd, 4th, and 5th groups of data in. After standardization, for the 5 groups of data, the data of each terminal index forms the following arrays (each column of several types of data): = [0.92, 0.85, 0.90, 0.88, 0.95] = [0.4167, 0.5333, 0.3, 0.5, 0.5833] = [0.3, 0.7, 0.5, 1.0, 0.4] = [0.9077, 0.9692, 0.9846, 0.8923, 0.9846] = [0.7, 0.75, 0.6, 0.8, 0.82] Step b: Calculate the proportionality coefficient of the samples under each index. is the proportionality coefficient of the jth data under the rth terminal index, is the standardized data:

[0087] Taking the standardized value of the terminal index Taking (the viscosity of hydraulic oil) as an example, the sum of the standardized values is: 0.3 + 0.7 + 0.5 + 1.0 + 0.4 = 2.9, and the proportionality coefficients in each group of data are , , , , .

[0088] Step c: Calculate the information entropy of each index in turn. The information entropy of the r-th index:

[0089] where is a constant used to standardize the entropy value. When , it is defined as .

[0090] Continuing with the example of the standardized value x3 of the terminal index (the viscosity of hydraulic oil), R = 5, r = 3, , calculate . Similarly, the information entropies of the 5 terminal indexes are calculated in turn as: , , , , .

[0091] Step d: Calculate the weight of each index in turn. Calculate the weight of the j-th index according to the information entropy:

[0092] Then the initial weight is .

[0093] Continuing with the above example, the total entropy value: = 0.98 + 0.85 + 0.918 + 0.89 + 0.87 = 4.508 The weights of each index: = (1 - 0.98) / (5 - 4.508) ≈ 0.0407 = (1 - 0.85) / (5 - 4.508) ≈ 0.305 = (1 - 0.918) / (5 - 4.508) ≈ 0.167 = (1 - 0.89) / (5 - 4.508) ≈ 0.224 = (1 - 0.87) / (5 - 4.508) ≈ 0.264 Initial weight 。

[0094] The mathematical principles of weight optimization include: (1) Using the prepared data set to optimize the weights. Let:

[0095]

[0096] In the formula, represents the weight coefficient generated after the nth iteration calculation. When n is 0, it means the initial weight. Denote the initial weight as . The initial weight is determined by the objective method, and the detailed method will be introduced later. . represents the health score value of the rth terminal index in the nth data group, ; represents the comprehensive health score value of the jth data in the nth data group, .

[0097] After the data set is made, select the first data group , and calculate the preliminary health score of the first group of data according to the initial weight:

[0098] Taking the result , calculated by a five-axis linkage CNC machine tool as an example, we get .

[0099] (2) Adjust the preliminary comprehensive health score obtained from the first data group . For a certain preliminary comprehensive health score , the adjustment amount is , and the optimized comprehensive health score is :

[0100] Denote the optimized comprehensive health score as , and the adjustment amount is denoted as , so it can be deduced that:

[0101] For example, the 4th data (spindle bearing temperature 62 points, vibration intensity 73 points) is abnormal, but the preliminary comprehensive health score of 79.00 points does not fully reflect the risk and is adjusted to 70.00 points. The adjustment amount , and other data are not adjusted, so the adjustment amount , the optimized comprehensive health score is .

[0102] (3) Using the method of subjective adjustment, the optimized comprehensive health score and the data set can be used to derive the new weights after the first iteration.

[0103] Because: , so multiply both sides of the equation by the inverse matrix of to obtain:

[0104]

[0105]

[0106]

[0107] In this way, the optimization from the initial weight to the weight after the first iteration is completed.

[0108] Continuing with the above example,

[0109] The weights after the first iteration: =

[0110] Such an iteration also has significance in industry. After adjustment, the comprehensive health score model significantly increases its attention to the spindle speed stability and the viscosity of the hydraulic oil, while reducing its dependence on the vibration intensity, which is more in line with the fault characteristics of the abnormal temperature of the spindle bearing in this case.

[0111] (4) Continuing to use the second set of data made and the weights after the first iteration, repeat the above steps to perform the second weight iteration as follows: First, calculate the preliminary comprehensive health score of the second set of data:

[0112] Continuing, under the weights after the first iteration, adjust the preliminary comprehensive health score matrix obtained from the second data set . For a certain preliminary comprehensive health score , the adjustment amount is , the optimized comprehensive health score is :

[0113] Denote the optimized comprehensive health score as , and the adjustment amount is denoted as , so it can be derived that:

[0114] Finally, using the optimized comprehensive health score and the data group a new weight array after the second iteration can be derived. In this way, the optimization from the weight in the first iteration to the weight after the second iteration is completed.

[0115] (5) By using an inductive method, the above process is generalized to the nth data group :

[0116]

[0117] where represents the preliminary comprehensive health score generated by the nth data group using the weight from the previous iteration. When n = 1, it refers to the score generated by the first data group using the initial weight ; . represents the optimized comprehensive health score generated based on the nth data group at the nth iteration, and the adjustment amount . Figure 7 is the flowchart of the weight optimization method shown in the embodiments of the present invention. Referring to Figure 7 , the algorithm for continuously iterating the weight W of the comprehensive health score model includes: Step 1: The user sets a maximum number of iterations e, and .

[0118] Step 2: In the nth iteration, multiply the nth data group by the weight generated in the (n - 1)th iteration to obtain the preliminary comprehensive health score . When n = 1, the initial weight used is .

[0119] Step 3: Use the subjective evaluation method to generate a preliminary comprehensive health score Adjustment amount and generate an optimized comprehensive health score .

[0120] Step 4: Use the nth data set and the optimized comprehensive health score , and use the formula to calculate the weight after the nth iteration .

[0121] For , , , and The five parameters:

[0122]

[0123]

[0124]

[0125] Therefore, the formula is simplified to: , so the core of using this method is to find the adjustment amount of the nth data set Adjustment amount . For this purpose, the present invention proposes an evaluation adjustment method to generate the adjustment amount in Step 2 . This method is applicable to the situation where "the expert's experience is not sufficient to directly give an accurate health score, but can vaguely give an evaluation on the optimization direction of the health score". This method needs to be set by the user himself, and the setting items include the "optimization direction of the health score" and the "optimization ratio of the health score" corresponding to the "health score".

[0126] For example, taking the comprehensive health score of the above five-axis linkage CNC machine tool as an example, the "optimization ratio of health score" is set to three gears: "meeting the requirements", "score too large", and "score too small". For the "meeting the requirements" option, it means that the currently given score meets the requirements and no changes need to be made to the current score. For the "score too large" option, it means that the score should be lowered in the reviewer's mind. Therefore, the "optimization direction of health score" is to decrease, and the optimized comprehensive health score in the "optimization ratio of health score" is set to 95% of the preliminary comprehensive health score. Similarly, for the "score too small" option, it means that the score should be increased in the reviewer's mind. Therefore, the "optimization direction of health score" is to increase, and the optimized comprehensive health score in the "optimization ratio of health score" is set to 105% of the preliminary comprehensive health score. For the preliminary health score calculated above , if the scorers select "meeting the requirements", "meeting the requirements", "score too large", "score too small", and "meeting the requirements" for 5 of the comprehensive health scores respectively, then the adjusted comprehensive health score is , and the adjustment amount is .

[0127] It should be noted that the "optimization ratio of health score result" set by the user is similar to the concept of learning rate in machine learning and should be reasonably designed. If it is set improperly, it will have an adverse impact on the optimization effect of the weight. When the ratio is set too large, it may lead to never being able to obtain a satisfactory result for the user. In extreme cases, it may even cause the weight parameter update to be too large, ultimately resulting in the divergence of the model training process. When the ratio is set too small, the step size of the model parameter update will be very small, which will cause the model training process to be very slow and require more iteration times to obtain a satisfactory result for the user. To prevent the above two situations, an iteration number needs to be set. At the same time, to evaluate the optimization result, an evaluation index needs to be set.

[0128] Step 5: Determine whether the iteration is completed.

[0129] If , the iteration is not completed, n = n + 1, and return to Step 2 to continue the iteration.

[0130] Meanwhile, if , but the user is already satisfied with the optimization result to a certain extent and believes that the iteration can be stopped, the iteration can also be directly completed.

[0131] If , the iteration is completed.

[0132] After the iteration is completed, output the optimized weights obtained during each iteration process

[0133] Step 6: Select the optimal weights. Set an evaluation index to judge the quality of the optimized weights:

[0134] In the formula, is, at the nth iteration, the number of evaluated as satisfactory in . This index describes the degree of satisfaction with the comprehensive health score of each generated data group. The larger this index, the closer the estimated value is to the actual value, indicating a better result. According to the index result, set the optimal weight as .

[0135]

[0136] For example: in e = 3 iterations, for 12 pieces of data, the satisfactory numbers are 4, 3, and 5 respectively, then , , , because is the largest, so = .

[0137] Step 7: Export to the health score model and deploy it. Subsequently, this model can be deployed to a specific project through a calculation module. That is, the collected data can be used to perform health score calculations.

[0138] The above has introduced in detail the industrial equipment health evaluation method provided by the embodiments of the present invention in combination with Figure 1 , Figure 2 . Next, the industrial equipment health evaluation system provided by the embodiments of the present invention will be introduced with reference to the accompanying drawings.

[0139] Figure 8 is a schematic structural diagram of the industrial equipment health evaluation system shown in the embodiments of the present invention. Referring to Figure 8 , the system of the present invention includes: A data acquisition module, configured to: acquire health score index data of industrial equipment and construct a health score model; An initial weight calculation module, configured to: construct a dataset of health score index data of industrial equipment and calculate the initial weights of the health score index data using the extended entropy weight method; A health score calculation module, configured to: calculate a preliminary health score using the health score model based on the initial weights of the health score index data and the health score index data of industrial equipment; A weight optimization module, which is configured to: adopt a dataset of health score indicators of industrial equipment to generate an adjustment amount of a preliminary comprehensive health score and an optimized health score, so as to iteratively optimize the weights and obtain the optimal weights; An output module, which is configured to: obtain a final health score according to the weighted sum of the optimal weights and the health score indicator data of the industrial equipment.

[0140] According to the industrial equipment health evaluation system of the embodiment of the present invention, it can correspond to execute the method described in the embodiment of the present invention, and the above and other operations and / or functions of each module of the industrial equipment health evaluation system are respectively for realizing Figure 1 the corresponding processes of each method in, for the sake of brevity, will not be elaborated herein.

[0141] Refer to Figure 9 the structural diagram of the computer device shown in. The computer device includes a processor, a communication interface, and a computer-readable storage medium. Among them, the processor, the communication interface, and the computer-readable storage medium can be connected through a bus or other means. Among them, the communication interface is used to receive and send data. The computer-readable storage medium can be stored in the memory of the computer device. The computer-readable storage medium is used to store a computer program. The computer program includes program instructions. The processor is used to execute the program instructions stored in the computer-readable storage medium. The processor (or CPU (Central Processing Unit, central processor)) is the computing core and control core of the computer device, and is adapted to implement one or more instructions, specifically adapted to load and execute one or more instructions to implement the corresponding steps in the embodiment of the industrial equipment health evaluation method.

[0142] This embodiment provides a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in the computer device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and, of course, the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and this storage space stores the processing system of the computer device. And, in this storage space, one or more instructions suitable for being loaded and executed by the processor are also stored. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk memory; optionally, it can also be at least one computer-readable storage medium located far from the aforementioned processor.

[0143] In one embodiment, one or more instructions are stored in the computer-readable storage medium; the one or more instructions stored in the computer-readable storage medium are loaded and executed by a processor to implement the corresponding steps in the embodiment of the above-mentioned industrial equipment health evaluation method.

[0144] This embodiment provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the corresponding steps in the embodiment of the above-mentioned industrial equipment health evaluation method.

[0145] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of an embodiment implemented in hardware, an embodiment implemented in software, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0146] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An industrial equipment health evaluation method, characterized in that, Including: Obtain the health score index data of industrial equipment and construct a health score model; Construct a health score index data set of industrial equipment and calculate the initial weights of the health score index data using the extended entropy weight method; According to the initial weights of the health score index data and the health score index data of industrial equipment, use the health score model to calculate the preliminary health score; Use the health score index data set of industrial equipment to generate the adjustment amount of the preliminary health score to iteratively optimize the initial weights and obtain the optimal weights; Obtain the final health score based on the weighted sum of the optimal weights and the health score index data of industrial equipment.

2. The industrial equipment health evaluation method according to claim 1, wherein The industrial equipment includes manufacturing equipment, energy industry equipment, chemical industry equipment, metallurgical industry equipment, and transportation industry equipment; If it is a five-axis linkage CNC machine tool in manufacturing equipment, the health score index data includes: spindle speed stability, spindle bearing temperature, hydraulic system oil viscosity, feed axis vibration intensity, and positioning accuracy; if it is a blast furnace blower in metallurgical industry equipment, the health score index data includes: fan efficiency, bearing temperature, vibration intensity, lubricating oil viscosity, outlet air pressure, motor current deviation rate, impeller dynamic balance deviation, and cooling water temperature.

3. The industrial equipment health evaluation method according to claim 1, characterized in that, The initial weights of the health score index data are obtained by multiplying the weights of the multi-level upper nodes of the health score index data.

4. The industrial equipment health evaluation method according to claim 1, wherein The method for constructing the health score index data set of industrial equipment includes: Select Q pieces of health score index data of industrial equipment to construct a data matrix; Perform elementary row transformation on the data matrix to verify whether the rank of the data matrix is R, and determine which data are linearly related and which data are linearly independent; If the number of linearly independent data is equal to R, only retain one piece of data from each pair of linearly related multiple pieces of data in the data matrix, and delete the rest; if the number of linearly independent data is greater than R, delete the linearly independent data exceeding R pieces, and only retain one piece of data from each pair of linearly related multiple pieces of data in the data matrix, and delete the rest; if the number of linearly independent data is less than R, supplement linearly independent data to R pieces; Repeat the above steps to construct multiple data matrices to obtain the health score index data set of industrial equipment.

5. The industrial equipment health evaluation method according to claim 1, wherein The method for calculating the initial weights of the health score index data using the extended entropy weight method includes: Standardize the health score index data of each industrial equipment to obtain extremely large indicators, extremely small indicators, intermediate indicators, and interval indicators to construct a data group of the health score index data of each industrial equipment; Calculate the proportionality coefficient of each health score index according to the data group of the health score index data of each industrial equipment; Calculate the information entropy of each health score index according to the proportionality coefficient of each health score index; Calculate the initial weights of each health score index data according to the information entropy of each health score index and the total number of health score indexes.

6. The industrial equipment health evaluation method according to claim 1, wherein Using the dataset of the health score indicators of industrial equipment to generate the adjustment amount of the preliminary comprehensive health score and the optimized health score, and iteratively optimizing the weights to obtain the optimal weights; the method includes: Generating the adjustment amount of the preliminary comprehensive health score according to the preliminary comprehensive health score, and performing the iterative optimization process; During the iteration process, generating the preliminary comprehensive health score of the current iteration according to the weights of the previous iteration number and the dataset of the health score indicators of the industrial equipment of the current iteration number; generating the adjustment amount of the health score of the current iteration number according to the preliminary comprehensive health score of the current iteration number, and calculating the optimized comprehensive health score of the current iteration number; furthermore, calculating the weights of the current iteration number according to the dataset of the health score indicators of the industrial equipment of the current iteration number and the calculated optimized comprehensive health score; After the iteration is completed, selecting the optimal weight from the weights optimized by all iteration numbers according to the evaluation indicators.

7. An industrial equipment health evaluation system, characterized in that, Including: A data acquisition module, which is configured to: acquire the health score indicator data of industrial equipment and construct a health score model; An initial weight calculation module, which is configured to: construct a dataset of the health score indicators of industrial equipment and calculate the initial weights of the health score indicator data by using the extended entropy weight method; A health score calculation module, which is configured to: calculate the preliminary health score by using the health score model according to the initial weights of the health score indicator data and the health score indicator data of industrial equipment; A weight optimization module, which is configured to: use the dataset of the health score indicators of industrial equipment to generate the adjustment amount of the preliminary comprehensive health score and the optimized health score, and iteratively optimize the weights to obtain the optimal weights; An output module, which is configured to: obtain the final health score according to the weighted sum of the optimal weights and the health score indicator data of industrial equipment.

8. A computer device, characterized in that A processor, adapted to execute a computer program; A computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by the processor, the steps in the industrial equipment health evaluation method according to any one of claims 1-6 are implemented.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is adapted to be loaded and executed by the processor to implement the steps in the industrial equipment health evaluation method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program, and when the computer program is executed by the processor, the steps in the industrial equipment health evaluation method according to any one of claims 1-6 are implemented.

Images