Knowledge data dual-drive water transfer pump station unit health evaluation method

Abstract

The invention discloses a knowledge data dual-drive water transfer pump station unit health evaluation method, and the method comprises the steps: determining a plurality of evaluation criteria for the health of a water transfer pump station unit, and carrying out the hierarchical structure division of the structure of the water transfer pump station unit; based on the knowledge graph, constructing a water transfer pump station unit relation topological graph with different health indexes; adopting a three-scale analytic hierarchy process to obtain the weight of each health index; the second part comprises the steps of obtaining a unit health index set of a target water transfer pump station based on a knowledge graph, performing scoring after preprocessing, and obtaining a score of each index data; and obtaining a final unit health evaluation result of the target water transfer pump station according to the score of each piece of index data in combination with the corresponding weight. According to the method, health evaluation is carried out on the pump station unit under different working conditions by adopting a mode of combining the knowledge graph and the network hierarchy analysis, an evaluation index system is perfected, the mutual influence relationship among indexes is considered, and evaluation can be comprehensively and accurately carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of water transfer pump station unit evaluation, and more specifically to a knowledge and data dual-driven water transfer pump station unit health evaluation method. Background Art

[0002] The core equipment of a water transfer pump station is the water pump. The water pump unit is a complex electromechanical system, and its health status directly affects the normal operation of the entire pump station. In the context of the construction of intelligent pump stations, timely, comprehensive and objective evaluation of the operating status of the water pump unit is a key link in predictive maintenance. Through the accumulation of big data from the intelligent sensing system, combined with expert and experience knowledge, unit health evaluation can be carried out to timely grasp the operating status of the equipment, so as to take targeted preventive measures, improve work efficiency and reduce costs.

[0003] As a typical representative of large-scale engineering equipment, pump station units need to determine their operating conditions based on multiple factors such as flow, head, blade angle, frequency, active power and speed due to the complexity of their operating conditions. The operating status of the units varies greatly under different operating conditions. Most of the existing operating condition classification methods are based on a single or two indicators of power and head, which is not applicable to water transfer pump stations with complex operating conditions and affected by water volume.

[0004] As a typical representative of large-scale engineering equipment, pump station units are composed of multiple subsystems such as water pump units and main motor units. At present, the research methods for water pump health status assessment can be divided into typical assessment methods based on knowledge and experience (hierarchical analysis / expert system / fault tree, etc.) and data-driven (deep learning / grey clustering / neural network, etc.). Among them, traditional algorithms based on knowledge and experience rely too much on manual experience. For example, the hierarchical analysis method performs health assessment on the research object, ignoring the complex mutual influence relationship between various indicators that affect the health of the water pump unit; although the network hierarchy analysis method (ANP) considers the mutual influence relationship between various indicators within the same level, it has problems such as strong subjectivity and insufficient precision. The data-driven method has high requirements on data quality, takes a long time to train the model and requires a lot of computing resources, and the model has poor interpretability.

[0005] Therefore, how to combine knowledge and experience with data-driven methods to improve the accuracy of health assessment of water pump station units when operating conditions are complex and changeable and data samples are insufficient is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0006] In view of the above problems, the present invention provides a knowledge and data dual-driven health assessment method for water transfer pump station units to at least solve some of the technical problems mentioned in the above background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The present invention provides a knowledge and data dual-driven water transfer pump station unit health assessment method, comprising a first part and a second part;

[0009] The first part includes:

[0010] P1. Determine multiple evaluation criteria for the health of the water transfer pump station units, and divide the water transfer pump station unit structure into hierarchical structures based on the multiple evaluation criteria;

[0011] P2. Based on the knowledge graph and the divided hierarchical structure, a topological diagram of the relationship between water transfer pump station units with different health indicators is constructed;

[0012] P3, using the three-scale hierarchical analysis method to obtain the weight of each health indicator based on the multiple evaluation criteria;

[0013] The second part includes:

[0014] S1. Obtain the unit health indicator set of the target water transfer pump station based on the knowledge graph;

[0015] S2. Preprocessing the unit health indicator set;

[0016] S3, scoring each indicator data in the pre-processed unit health indicator set according to the multiple evaluation criteria to obtain a score for each indicator data;

[0017] S4. According to the score of each indicator data and the corresponding weight, the final unit health evaluation result of the target water transfer pump station is obtained.

[0018] Furthermore, the step P1 specifically includes:

[0019] The operation stability criterion, equipment integrity criterion, external environment criterion and initial state criterion are used as multiple evaluation criteria for the health of water transfer pump station units;

[0020] The unit structure of the water transfer pump station is divided into target layer, system layer, criterion layer, primary indicator layer and secondary indicator layer; among which:

[0021] Taking the health status evaluation of the water pump unit as the target in the target layer;

[0022] The water pump and the motor are regarded as two devices in the system layer;

[0023] Each device corresponds to a plurality of evaluation criteria in the criterion layer;

[0024] Each evaluation criterion corresponds to one or more first-level indicators in the first-level indicator layer;

[0025] Each first-level indicator corresponds to a plurality of second-level indicators in the second-level indicator layer.

[0026] Furthermore, the step P2 specifically includes:

[0027] Nouns are extracted from the divided hierarchical structure as nodes of the knowledge graph. The relationships between different nodes are: part-component-component, part-component-equipment, part-connection-component, and part-connection-part.

[0028] Extracting equipment entities from historical water transfer pump station fault records and maintenance-related documents, and combining different equipment entities into triplets according to corresponding relationships; the triplets include fault-cause-fault, component-occurrence-fault, and fault-cause-fault symptom;

[0029] The combined triplet data is imported into the knowledge graph to form a topological diagram of the relationship between water pump station units with different health indicators.

[0030] Furthermore, the step P3 specifically includes:

[0031] According to the importance between every two evaluation criteria, the three-scale hierarchical analysis method is used to construct the weight of the evaluation criteria;

[0032] Under each evaluation criterion, a supermatrix is ​​constructed according to the importance between every two secondary indicators in the secondary indicator layer; and a weight matrix is ​​constructed according to the importance between every two primary indicators in the primary indicator layer; the supermatrix under each evaluation criterion is multiplied by the weight matrix to obtain a weighted supermatrix of each health indicator under each evaluation criterion;

[0033] The weighted supermatrix under each evaluation criterion is synthesized through the weight of the evaluation criterion to obtain the weight of each health indicator.

[0034] Furthermore, the unit health indicator set includes operating parameters and operation parameters;

[0035] The operating parameters include flow rate, head, blade angle, frequency, active power and speed;

[0036] The operating parameters include motor vibration, water pump vibration, motor noise, water pump noise, thrust bearing temperature and blade adjustment temperature.

[0037] Furthermore, the step S2 specifically includes:

[0038] For the operating condition parameters, determining a preset normal range of the operating condition parameters, deleting the operating condition parameters outside the preset normal range as abnormal values; and deleting the invalid values ​​that are empty in the operating condition parameters;

[0039] For the operating parameters, invalid data and null values ​​generated in the target water transfer pump station unit operating data due to sensor failure, transmission failure, and environmental noise are marked as abnormal, and then the operating parameters are normalized.

[0040] Furthermore, the step S3 specifically includes:

[0041] According to the preset working conditions, the pre-processed unit health index system is divided into index data affected by the working conditions and index data not affected by the working conditions;

[0042] For the index data affected by the working conditions, the index data affected by the working conditions are input into the constructed support vector machine multi-classification model to output a variety of water pump working conditions; the scores of each index data under different water pump working condition combinations are obtained through the operation stability criterion;

[0043] For indicator data that are not affected by working conditions, the corresponding scores are determined based on the preset thresholds of each indicator data using the equipment integrity criteria, external environment criteria, and initial state criteria.

[0044] Furthermore, the radial basis kernel function is used as the kernel function of the support vector machine multi-classification model.

[0045] Furthermore, (1) for the index data affected by the working conditions:

[0046] 1) Under the above mentioned operation stability criteria:

[0047] Determine the threshold intervals for different combinations of water pump working conditions; remove the index data outside the threshold interval under each combination of water pump working conditions as outliers, and score the index data within the threshold interval using a preset evaluation rule to obtain the scores of each index data under different combinations of water pump working conditions; the preset evaluation rule is expressed as:

[0048]

[0049] Among them, E represents the score of each indicator data; P represents the indicator data; is the healthy sample benchmark value; σ represents the standard deviation;

[0050] (2) For indicator data not affected by working conditions:

[0051] 1) Under the equipment integrity criteria and the external environment criteria:

[0052] According to the degree of damage to the parts, the parts are divided into no damage, slight damage and severe damage; and the corresponding index data are assigned values ​​of 0, 0.5 and 1 respectively;

[0053] 2) Under the initial state criteria:

[0054] Obtain the corrected running time score and factory age score for each indicator data; the corrected running time score includes the total running time score from the last overhaul to the present and the total running time score from the last minor overhaul to the present;

[0055] Total running time since the last overhaul score = 1-total running time since the last overhaul / national standard average overhaul time

[0056] Total running time since the last minor repair = 1-total running time since the last minor repair / national standard average minor repair time

[0057] Years of manufacture score = 1-years of manufacture / average lifespan.

[0058] Furthermore, in step S4, the product of the score of each indicator data and the corresponding weight is used as the final score of each indicator data; the final scores of all indicator data are added together to obtain the final unit health evaluation result of the target water transfer pump station.

[0059] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a knowledge and data dual-driven water pump station unit health evaluation method, which has the following beneficial effects:

[0060] 1. The present invention considers the relationship between different influencing factors based on the network hierarchy method ANP, and uses ANP to evaluate the health of the water pump unit, which is closer to reality.

[0061] 2. The traditional way of constructing a network hierarchy based on artificial experience is prone to missing some key information. The present invention, based on the knowledge graph, can extract key influencing factors and the correlation between factors based on a large number of samples and similar samples, and can further improve the network structure.

[0062] 3. When analyzing and scoring health indicators, the present invention takes into account the impact of different working conditions and uses a multi-classification support vector machine to classify working conditions. It can accurately divide operating intervals and parameter thresholds, making the evaluation results more realistic.

[0063] 4. The present invention adopts the knowledge graph algorithm to complete the knowledge graph according to the data accumulated on site, and the completed knowledge graph corresponds to a more complete ANP network.

[0064] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present invention.

[0065] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0067] Figure 1 A schematic flow chart of a knowledge and data dual-driven method for health assessment of water transfer pump station units provided in an embodiment of the present invention.

[0068] Figure 2 A schematic diagram of the topological relationship of the water transfer pump station units provided in an embodiment of the present invention.

[0069] Figure 3 A schematic diagram of the multi-condition classification process of a support vector machine provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figure 1 As shown, the embodiment of the present invention discloses a knowledge and data dual-driven water transfer pump station unit health evaluation method, including a first part and a second part; wherein:

[0072] The first part mentioned above includes:

[0073] P1. Determine multiple evaluation criteria for the health of water transfer pump station units, and divide the water transfer pump station unit structure into hierarchical structures based on the multiple evaluation criteria;

[0074] P2. Based on the knowledge graph and the divided hierarchical structure, a topological diagram of the relationship between water transfer pump station units with different health indicators is constructed;

[0075] P3, using the three-scale analytic hierarchy process to obtain the weight of each health indicator based on multiple evaluation criteria;

[0076] The above second part includes:

[0077] S1. Obtain the unit health indicator set of the target water transfer pump station based on the knowledge graph;

[0078] S2. Preprocessing the unit health indicator set;

[0079] S3, scoring each indicator data in the pre-processed unit health indicator set according to multiple evaluation criteria to obtain a score for each indicator data;

[0080] S4. According to the score of each indicator data and the corresponding weight, the final unit health evaluation result of the target water transfer pump station is obtained.

[0081] Next, each of the above steps will be described in detail.

[0082] In the above steps P1 and P2, due to the complex internal structure of the water pump unit, the large number of parts and components, and the various parameters that affect the performance of the unit, it will be difficult to identify the influencing factors of the water pump unit as a whole, and it is likely to miss more important factors. Therefore, when identifying the influencing factors, first determine the four criteria of operating stability, equipment integrity, external environment and initial state as the importance comparison criteria, and then use the knowledge graph to mine the influencing factors under different evaluation criteria, and construct a relationship topology diagram of the water pump station units with different health indicators; specifically:

[0083] Firstly, based on multiple evaluation criteria, the hierarchical structure of the water transfer pump station unit structure is divided into target layer, system layer, criterion layer, primary indicator layer and secondary indicator layer; wherein: the health status evaluation of the water pump unit is taken as the target in the target layer; the water pump and the motor are taken as two devices in the system layer; each device corresponds to multiple evaluation criteria in the criterion layer; each evaluation criterion corresponds to one or more primary indicators in the primary indicator layer; each primary indicator corresponds to multiple secondary indicators in the secondary indicator layer;

[0084] Then, based on the knowledge graph, according to the divided hierarchical structure, a topological graph of the relationship between water pump station units with different health indicators is constructed; that is, nouns are extracted from the divided hierarchical structure as nodes of the knowledge graph, and the relationship between different nodes is: part-component-component, component-component-equipment, component-connection-component, and part-connection-part; the node triples are imported into the knowledge graph construction tool Neo4j to reduce the time cost of knowledge graph construction;

[0085] Equipment entities are extracted from the historical water transfer pump station fault records and maintenance related documents (such as fault records, maintenance manuals, and one-machine-one-file files), involving equipment type, fault type, equipment component type, fault representation, maintenance plan, fault cause, etc. After completing entity recognition and relationship extraction, different entities are combined into triples according to the corresponding relationship, including fault-cause-fault, component-occurrence-fault, and fault-cause-fault symptom; the combined triple data is imported into the knowledge graph construction tool Neo4j to supplement and improve the relationship between the knowledge graph indicator layer and the components; a relationship topology diagram of water transfer pump station units with different health indicators is formed, such as Figure 2 As shown in the figure, the bidirectional arrows indicate that the indicators within the two element groups affect each other, the unidirectional arrows indicate that the indicators within the tail element group affect the indicators within the head element group, and the circular arrows indicate that the indicators within the element group affect each other;

[0086] Specifically, the term frequency-inverse document frequency TF-IDF method can be used to define the algorithm for extracting node keywords in the knowledge graph, and perform feature extraction and representation on the data set;

[0087] The embodiment of the present invention utilizes the knowledge graph to improve the construction of the network layer, that is, automatically establishes the network layer based on the knowledge graph, reduces manual participation, and at the same time fully utilizes prior knowledge to improve the completeness of the evaluation index system.

[0088] In the above step P3, in the traditional network analysis method evaluation, the determination of the indicator weight is based on the 1-9 scale method, and the relative importance of the indicators is given by experts; when there are many indicators in the evaluation system, not only the amount of calculation is large, but also it is easy to cause poor consistency of the judgment matrix, and the calculated weight value has low reliability. The present invention adopts an improved (-1, 0, 1) three-scale method for importance judgment. This method has a self-adjusting function, does not require consistency testing, and can improve the quality of calculated weights; specifically:

[0089] (1) According to the importance between every two evaluation criteria, the three-scale improved hierarchical analysis method is used to construct the weight of the evaluation criteria; the specific steps include:

[0090] 1) The health status evaluation of the water pump unit is taken as the target layer, and the two main equipment, water pump and motor, are evaluated separately;

[0091] 2) The importance evaluation criteria for both the pump and the motor are operation stability, equipment integrity, external environment and initial state. These four evaluation criteria are represented as A1, A2, A3 and A4 respectively;

[0092] First, the (-1, 0, 1) three-scale hierarchical analysis method is used to construct an n*n dimensional comparison matrix according to the importance between every two evaluation criteria; where n represents a total of n evaluation criteria; the elements in the comparison matrix are expressed as:

[0093]

[0094] Among them, c ij Represents the element in the i-th row and j-th column of the comparison matrix; A i represents the evaluation criteria corresponding to the i-th row; A j represents the evaluation criteria corresponding to the jth column;

[0095] Then, the comparison matrix is ​​transformed into a consistency matrix B = (b ij )n*n; the eigenvector of the consistency matrix B is the weight of the evaluation criterion;

[0096] 3) Calculate the weighted supermatrix (i.e., the weight of each factor) of each health indicator under each evaluation criterion; specifically, the following steps are included:

[0097] ① Under each evaluation criterion, construct the supermatrix W according to the importance between every two secondary indicators in the secondary indicator layer;

[0098] ② According to the importance between every two first-level indicators in the first-level indicator layer, construct the weight matrix A;

[0099] ③ Multiply the supermatrix W under each evaluation criterion by the weight matrix A to obtain the weighted supermatrix of each health indicator under each evaluation criterion;

[0100] The equipment integrity criteria are now used as an example to illustrate the specific steps:

[0101] ① Construct the super matrix W: take the equipment integrity criterion as the main criterion, take one of the secondary indicators in the corresponding secondary indicator layer as the secondary criterion, and construct the judgment matrix by comparing the relative advantages with other secondary indicators according to the principle of the (-1, 0, 1) three-scaling method; get the normalized eigenvector (W i1 ,W i2 ,W i3 ,…,W iN ) T , which is the sorting vector of the network layer elements. Similarly, the vectors relative to other elements can also be calculated and recorded as W ij ,Right now:

[0102]

[0103] Among them, W ij is an element in the supermatrix W, j = 1, 2, ..., N; i = 1, 2, ..., N; based on Wij A supermatrix W of control elements can be constructed, in which the category ranking vectors irrelevant to the secondary index are 0; the supermatrix W is expressed as:

[0104]

[0105] ② Construct the weight matrix A: take the equipment integrity criterion as the main criterion, take one of the first-level indicators in the corresponding first-level indicator layer as the secondary criterion, and construct the judgment matrix by comparing the relative advantages with other first-level indicators according to the principle of the (-1, 0, 1) three-scaling method; the judgment matrix is ​​an element in the weight matrix A; based on this, the weight matrix A can be finally obtained; the category ranking vector in the weight matrix A that is not related to the first-level indicator is 0;

[0106]

[0107] ③ Multiply the supermatrix W under the equipment integrity criterion by the weight matrix A, that is, W*A, to obtain the weighted supermatrix of each health indicator under the equipment integrity criterion, expressed as:

[0108]

[0109] 4) The weights of the evaluation criteria obtained through the above calculations are used to synthesize the weighted supermatrix under each evaluation criterion (i.e., the single criterion weight is multiplied by the matrix block in the weighted supermatrix under the criterion) to obtain the weight of each health indicator and the weight ranking of each health indicator.

[0110] In the above step S1, a set of unit health indicators of the target water transfer pump station is obtained based on the knowledge graph;

[0111] The unit health index set includes operating parameters and operation parameters; the operating parameters include flow, head, blade angle, frequency, active power and speed; the operation parameters include 21 parameters including motor vibration, water pump vibration, motor noise, water pump noise, thrust bearing temperature and blade adjustment temperature;

[0112] In the above step S2, the unit health indicator set is preprocessed; specifically, it includes:

[0113] For the operating condition parameters, determine the preset normal range of the operating condition parameters, delete the operating condition parameters outside the preset normal range as abnormal values; and delete the invalid values ​​that are empty in the operating condition parameters;

[0114] For operating parameters, invalid data and null values ​​generated by sensor failure, transmission failure, and environmental noise in the target water transfer pump station unit operating data are marked as abnormal, and then the operating parameters are normalized for better input into the model;

[0115] In the above step S3, each indicator data in the pre-processed unit health indicator set is scored according to multiple evaluation criteria to obtain a score for each indicator data; specifically, the score includes:

[0116] According to the preset working conditions, the pre-processed unit health index system is divided into index data affected by the working conditions and index data not affected by the working conditions;

[0117] (1) For index data affected by working conditions:

[0118] The index data affected by the working conditions are input into the constructed support vector machine multi-classification model to output a variety of water pump working conditions; the scores of each index data under different water pump working condition combinations are obtained through the operation stability criterion; specifically:

[0119] 1) Under the operation stability criterion: determine the threshold intervals of different pump working condition combinations; treat the index data outside the threshold interval under each pump working condition combination as an outlier, and the index data score of the outlier is zero; for the index data within the threshold interval, score them using a preset evaluation rule to obtain the score of each index data under different pump working condition combinations; the preset evaluation rule is expressed as:

[0120]

[0121] Among them, E represents the score of each indicator data; P represents the indicator data; is the healthy sample benchmark value; σ represents the standard deviation;

[0122] 2) In this step, the training data set is used to train the SVM classification model. For multi-value classification problems, multiple binary classifiers need to be constructed when applying SVM. In the embodiment of the present invention, a one-versus-rest method is used to construct multiple binary classifiers. Assuming that N binary classifiers are constructed, the method is as follows:

[0123] ①Build a binary classifier

[0124] For each category i (i=1,2,...,N), perform the following steps:

[0125] ② Divide the data set

[0126] Positive class: all samples with label i, marked as y i N =1

[0127] Negative class: all samples whose labels are not i are marked as y i N =-1

[0128] ③Training binary classifier

[0129] When predicting and classifying the test sample, the category label y in the output of N binary classifiers i N The largest category is the category of the test sample. The multi-condition classification process of the test sample is as follows: Figure 3 shown.

[0130] ④Determine the kernel function

[0131] The classification problem of the operating conditions of water pump units is complex and the various monitoring parameters are linearly inseparable. Therefore, the present invention selects the radial basis function (RBF) kernel function, which is very effective in solving such problems, as the kernel function of SVM. The key to the SVM classification algorithm is the selection of the penalty coefficient C and the parameter σ of the RBF kernel function. These two parameters determine the learning ability and prediction ability of SVM to a large extent. The present invention uses a grid search method to first select a parameter pair (C and σ), and then uses a cross-validation method to optimize the objective function until the optimal parameter pair is found to achieve the highest accuracy of cross-validation;

[0132] Compared with the traditional single-parameter classification method, the operating condition classification method based on the support vector machine of the present invention realizes multivariable classification, which can effectively describe the complex and changeable operating conditions, accurately divide the operating condition intervals, and create conditions for feature extraction under different operating conditions.

[0133] (2) For indicator data not affected by working conditions:

[0134] 1) Under the equipment integrity criteria and the external environment criteria:

[0135] According to the degree of damage to the parts, the parts are divided into no damage, slight damage and severe damage; and the corresponding index data are assigned values ​​of 0, 0.5 and 1 respectively;

[0136] 2) Under the initial state criteria:

[0137] According to the national standard (Technical Management Regulations for Pumping Stations GB / T 30948-2014), the main water pump and transmission device shall be overhauled every 3 to 5 years or 2500 to 15000 hours of operation, and shall be overhauled every 1 year or 1000 hours of operation; the main motor shall be overhauled every 3 to 8 years or 3000 to 20000 hours of operation, and shall be overhauled every 1 to 2 years or 2000 hours of operation; respectively expressed as: total running time score from the last overhaul to the present = 1-total running time from the present / national standard average overhaul time; total running time score from the last minor repair to the present = 1-total running time from the present / national standard average minor repair time

[0138] According to the average life of factory settings, the factory life evaluation value is calculated, that is, the factory life score = 1-factory life / average life.

[0139] In the above step S4, the final unit health evaluation result of the target water transfer pump station is obtained according to the score of each indicator data and the corresponding weight.

[0140] In summary, in order to change the current status of the planned maintenance mode of pump station units and improve operation and maintenance efficiency, the embodiment of the present invention provides a knowledge and data dual-driven health evaluation method for water diversion pump station units. The method combines knowledge graph + improved ANP to perform health evaluation on pump station units under different working conditions, improves the evaluation index system, and considers the mutual influence relationship between indicators. It can make a comprehensive and accurate evaluation, so as to timely grasp the operating status of the unit, discover problems early, and provide support for management decisions, so as to formulate more scientific and reasonable operation and maintenance strategies and plans, and improve operation and maintenance efficiency and quality.

[0141] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0142] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A knowledge and data dual-driven water transfer pump station unit health assessment method, characterized in that: It includes the first part and the second part; The first part includes: P1. Determine multiple evaluation criteria for the health of the water transfer pump station units, and divide the water transfer pump station unit structure into hierarchical structures based on the multiple evaluation criteria; P2. Based on the knowledge graph and the divided hierarchical structure, a topological diagram of the relationship between water transfer pump station units with different health indicators is constructed; P3, using the three-scale hierarchical analysis method to obtain the weight of each health indicator based on the multiple evaluation criteria; The second part includes: S1. Obtain the unit health indicator set of the target water transfer pump station based on the knowledge graph; S2. Preprocessing the unit health indicator set; S3, scoring each indicator data in the pre-processed unit health indicator set according to the multiple evaluation criteria to obtain a score for each indicator data; S4. According to the score of each indicator data and the corresponding weight, the final unit health evaluation result of the target water transfer pump station is obtained.

2. According to the knowledge and data dual-driven health evaluation method of water transfer pump station units in claim 1, it is characterized in that: The step P1 specifically includes: The operation stability criterion, equipment integrity criterion, external environment criterion and initial state criterion are used as multiple evaluation criteria for the health of water transfer pump station units; The unit structure of the water transfer pump station is divided into target layer, system layer, criterion layer, primary indicator layer and secondary indicator layer; among which: Taking the health status evaluation of the water pump unit as the target in the target layer; The water pump and the motor are regarded as two devices in the system layer; Each device corresponds to a plurality of evaluation criteria in the criterion layer; Each evaluation criterion corresponds to one or more first-level indicators in the first-level indicator layer; Each first-level indicator corresponds to a plurality of second-level indicators in the second-level indicator layer.

3. According to the knowledge and data dual-driven health evaluation method of water transfer pump station units in claim 2, it is characterized in that: The step P2 specifically includes: Nouns are extracted from the divided hierarchical structure as nodes of the knowledge graph. The relationships between different nodes are: part-component-component, part-component-equipment, part-connection-component, and part-connection-part. Extracting equipment entities from historical water transfer pump station fault records and maintenance-related documents, and combining different equipment entities into triplets according to corresponding relationships; the triplets include fault-cause-fault, component-occurrence-fault, and fault-cause-fault symptom; The combined triplet data is imported into the knowledge graph to form a topological diagram of the relationship between water transfer pump station units with different health indicators.

4. According to the knowledge and data dual-driven water transfer pump station unit health assessment method of claim 3, it is characterized in that: The step P3 specifically includes: According to the importance between every two evaluation criteria, the three-scale hierarchical analysis method is used to construct the weight of the evaluation criteria; Under each evaluation criterion, a supermatrix is ​​constructed according to the importance between every two secondary indicators in the secondary indicator layer; and a weight matrix is ​​constructed according to the importance between every two primary indicators in the primary indicator layer; the supermatrix under each evaluation criterion is multiplied by the weight matrix to obtain a weighted supermatrix of each health indicator under each evaluation criterion; The weighted supermatrix under each evaluation criterion is synthesized through the weight of the evaluation criterion to obtain the weight of each health indicator.

5. According to the knowledge and data dual-driven water transfer pump station unit health assessment method of claim 1, it is characterized in that: The unit health indicator set includes operating parameters and operation parameters; The operating parameters include flow rate, head, blade angle, frequency, active power and speed; The operating parameters include motor vibration, water pump vibration, motor noise, water pump noise, thrust bearing temperature and blade adjustment temperature.

6. According to the knowledge and data dual-driven water transfer pump station unit health assessment method of claim 5, it is characterized in that: The step S2 specifically includes: For the operating condition parameters, determining a preset normal range of the operating condition parameters, deleting the operating condition parameters outside the preset normal range as abnormal values; and deleting the invalid values ​​that are empty in the operating condition parameters; For the operating parameters, invalid data and null values ​​generated in the target water transfer pump station unit operating data due to sensor failure, transmission failure, and environmental noise are marked as abnormal, and then the operating parameters are normalized.

7. The knowledge and data dual-driven water transfer pump station unit health assessment method according to claim 3 is characterized in that: The step S3 specifically includes: According to the preset working conditions, the pre-processed unit health index system is divided into index data affected by the working conditions and index data not affected by the working conditions; For the index data affected by the working conditions, the index data affected by the working conditions are input into the constructed support vector machine multi-classification model to output a variety of water pump working conditions; the scores of each index data under different water pump working condition combinations are obtained through the operation stability criterion; For indicator data that are not affected by working conditions, the corresponding scores are determined based on the preset thresholds of each indicator data using the equipment integrity criteria, external environment criteria, and initial state criteria.

8. The knowledge and data dual-driven water transfer pump station unit health assessment method according to claim 7 is characterized in that: The radial basis kernel function is used as the kernel function of the support vector machine multi-classification model.

9. The knowledge-data dual-driven water transfer pump station unit health assessment method according to claim 7 is characterized by: (1) For index data affected by working conditions: 1) Under the above mentioned operation stability criteria: Determine the threshold intervals for different combinations of water pump working conditions; remove the index data outside the threshold interval under each combination of water pump working conditions as outliers, and score the index data within the threshold interval using a preset evaluation rule to obtain the scores of each index data under different combinations of water pump working conditions; the preset evaluation rule is expressed as: Among them, E represents the score of each indicator data; P represents the indicator data; is the healthy sample benchmark value; σ represents the standard deviation; (2) For indicator data not affected by working conditions: 1) Under the equipment integrity criteria and the external environment criteria: According to the degree of damage to the parts, the parts are divided into no damage, slight damage and severe damage; and the corresponding index data are assigned values ​​of 0, 0.5 and 1 respectively; 2) Under the initial state criteria: Obtain the corrected running time score and factory age score for each indicator data; the corrected running time score includes the total running time score from the last overhaul to the present and the total running time score from the last minor overhaul to the present; Total running time since the last overhaul score = 1-total running time since the last overhaul / national standard average overhaul time Total running time since the last minor repair = 1-total running time since the last minor repair / national standard average minor repair time Years of manufacture score = 1-years of manufacture / average lifespan.

10. The knowledge and data dual-driven water transfer pump station unit health assessment method according to claim 1 is characterized in that: In step S4, the product of the score of each indicator data and the corresponding weight is used as the final score of each indicator data; the final scores of all indicator data are added together to obtain the final unit health evaluation result of the target water transfer pump station.

Images