Elevator health condition assessment method based on fusion network model

Abstract

The invention discloses an elevator health condition assessment method based on a fusion network model, and belongs to the technical field of special equipment assessment. The method comprises the steps that multi-dimensional data of a to-be-tested elevator are collected and preprocessed; constructing a BP neural network model, taking a genetic algorithm GA as a strategy for optimizing a hyper-parameter, a weight and a threshold value of the BP neural network model to obtain a GA-BP neural network model, and obtaining a first evaluation result of the elevator health condition based on the GA-BP neural network model; a logic judgment model is constructed, and a second evaluation result of the elevator health condition is obtained; and fusing and grading evaluation results of the GA-BP neural network model and the logic judgment model, and taking a grading result as a final elevator health condition diagnosis result. According to the method, the learning and prediction capabilities of the neural network and the intuition and accuracy of logical judgment are combined, so that the accuracy and efficiency of elevator health condition evaluation are improved, and a powerful guarantee is provided for safe operation of the elevator.

Description

Technical Field

[0001] The present invention relates to the technical field of special equipment evaluation, and more particularly to an elevator health status evaluation method based on a fusion network model. Background Art

[0002] Elevators are important equipment in modern buildings, and their operating status directly affects the safety of passengers. In recent years, elevator safety accidents have occurred frequently. Traditional fault diagnosis methods rely on regular manual inspections and simple sensor alarms, and do not conduct multi-dimensional inspections of elevators. This is inefficient and not conducive to providing strong technical support for elevator maintenance and safe operation.

[0003] An analysis of closed accidents shows that the main causes of elevator accidents include inadequate safety management and maintenance, illegal operations or improper operation, failure of safety components or protective devices, untimely inspection, improper emergency rescue and other reasons. According to the above data, inadequate safety management, maintenance and untimely inspection account for a large proportion, while accidents caused by failure of safety components and protective devices are also high. If the health status and risks of the elevator can be evaluated and predicted before an accident occurs, unnecessary casualties will be reduced.

[0004] Therefore, how to provide an elevator health status assessment method based on a fusion network model that can solve the above problems is an issue that technical personnel in this field urgently need to solve. Summary of the invention

[0005] In view of this, the present invention provides an elevator health status assessment method based on a fusion network model, which is used to solve the technical problems existing in the above-mentioned prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for evaluating the health status of an elevator based on a fusion network model comprises the following steps:

[0008] S100: Collect multi-dimensional data of the elevator to be tested and perform pre-processing;

[0009] S200: constructing a BP neural network model, using a genetic algorithm GA as a strategy for optimizing hyperparameters, weights and thresholds of the BP neural network model, obtaining a GA-BP neural network model, and obtaining a first evaluation result of the health status of the elevator based on the GA-BP neural network model;

[0010] S300: constructing a logic judgment model to obtain a second evaluation result of the elevator health status;

[0011] S400: Fusing the first evaluation result and the second evaluation result, and grading the fusion result, and using the grading result as the final elevator health status diagnosis result.

[0012] Preferably, S100: collecting multi-dimensional data of the elevator to be tested and preprocessing it, includes:

[0013] S110: elevator attribute information of the elevator on standby for testing;

[0014] S120: Setting a sampling period, and collecting multi-dimensional data of the elevator to be tested based on the sampling period, wherein the multi-dimensional data includes: IoT monitoring information, maintenance information, and inspection information;

[0015] S130: Remove data records with missing values ​​and data that are obviously inconsistent with reality in the multidimensional data, and normalize the remaining data.

[0016] Preferably, the S130 includes:

[0017] Plot data lists according to multidimensional data;

[0018] Perform column indexing on the data in the data list, traverse the missing values ​​in the row, determine the data completeness, and determine the completeness of the data in the data list;

[0019] The data with a completeness greater than or equal to a first preset threshold is retained, and the data with a completeness less than the first preset threshold is removed;

[0020] Data records that are obviously inconsistent with reality are deleted from the data list by calculating a deviation value, wherein the deviation value is calculated based on an absolute value or a relative error. When the deviation value is greater than or equal to a second preset threshold, the data record is marked as inconsistent with reality and deleted;

[0021] The remaining data were normalized.

[0022] Preferably, the BP neural network model comprises: an input layer, a hidden layer and an output layer;

[0023] The number of nodes in the input layer is determined according to the number of features of the input data, the number of nodes in the output layer is 1, and the evaluation result of the health status of the elevator is output. The number of nodes in the hidden layer is obtained according to an empirical formula.

[0024] Preferably, the empirical formula is:

[0025]

[0026] Among them, n1 is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, and a is a constant between 1 and 10.

[0027] Preferably, a genetic algorithm GA is used as a strategy to optimize the hyperparameters, weights and thresholds of the BP neural network model to obtain a GA-BP neural network model, including:

[0028] The initial population is established by real number coding. The initial population consists of four parts: hidden layer threshold, connection weight between hidden layer and output layer, connection weight between input layer and hidden layer, and output layer threshold, as a set of chromosomes.

[0029] Perform selection, crossover and mutation operations of genetic algorithm to generate the next generation population, calculate the individual fitness value of the new population, and take the minimum value as the optimal value;

[0030] The optimized connection weights and thresholds are introduced into the BP neural network, and the normalized data are used as training samples for repeated training.

[0031] Update the optimized connection weights, thresholds and network structure obtained through training to obtain a GA-BP neural network model;

[0032] The first evaluation result of the elevator health status is obtained based on the GA-BP neural network model.

[0033] Preferably, the expression for calculating the individual fitness value of the new population is:

[0034]

[0035] Where n is the number of nodes in the network output layer, y i is the actual output of the ith node; y i ′ is the expected output of the ith node.

[0036] Preferably, S300: constructing a logic judgment model to obtain a second evaluation result of the elevator health status includes:

[0037] Define logical relationships to determine the health status of elevators;

[0038] Among them, propositions of logical relations include:

[0039] The relationship between normal operation and fault conditions;

[0040] The relationship between maintenance operation and operation status;

[0041] The relationship between emergency electric operation and fault handling;

[0042] The relationship between docking operations and loading and unloading of cargo;

[0043] The relationship between fire operations and fire emergency response;

[0044] The defined logical questions are organized into a rule base, and a logical judgment model is obtained based on the rule base. Preferably, the logical judgment model is expressed as:

[0045] Exception = P00N AND P00M;

[0046] Among them, P00N and P00M represent the truth or falsity of any two propositions of logical relationship.

[0047] Preferably, S400 includes:

[0048] The binary values ​​output by the logic judgment model are converted into continuous values ​​in the same range as those output by the GA-BP neural network model;

[0049] Calculate the fusion result, the expression is:

[0050] F=W1*N output +W2*L output ;

[0051] Where F represents the final elevator health diagnosis result, N output represents the first evaluation result of the elevator health status output by the GA-BP neural network model, L output represents the second evaluation result of the elevator health status after the logic judgment model is converted, W1 represents the weight of the GA-BPGA-BP neural network model, and W2 represents the weight of the logic judgment model;

[0052] The fusion results are graded and used as the final elevator health status diagnosis result.

[0053] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method for evaluating the health status of an elevator based on a fusion network model. In the process of training the model using a BP neural network, the weights are repeatedly corrected and the model is continuously optimized, so that the model can fit the training data with arbitrary precision. The dual identification of the health status of the elevator based on the single-even time series model of the GA-BP neural network and the group-even space model based on logical judgment combines the learning and prediction capabilities of the neural network with the intuitiveness and accuracy of logical judgment, so as to improve the accuracy and efficiency of the elevator health status assessment and provide strong guarantee for the safe operation of the elevator. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] 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.

[0055] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0056] 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.

[0057] See also Figure 1 The embodiment of the present invention discloses a method for evaluating the health status of an elevator based on a fusion network model, comprising the following steps:

[0058] S100: Collect multi-dimensional data of the elevator to be tested and perform pre-processing;

[0059] S200: constructing a BP neural network model, using a genetic algorithm GA as a strategy for optimizing hyperparameters, weights and thresholds of the BP neural network model, obtaining a GA-BP neural network model, and obtaining a first evaluation result of the health status of the elevator based on the GA-BP neural network model;

[0060] S300: constructing a logic judgment model to obtain a second evaluation result of the elevator health status;

[0061] S400: Fusing the first evaluation result and the second evaluation result, and grading the fusion result, and using the grading result as the final elevator health status diagnosis result.

[0062] In a specific embodiment, S100: collecting multi-dimensional data of the elevator to be tested and preprocessing it, including:

[0063] S110: elevator attribute information of the elevator on standby for testing;

[0064] S120: Setting a sampling period, and collecting multi-dimensional data of the elevator to be tested based on the sampling period, wherein the multi-dimensional data includes: IoT monitoring information, maintenance information, and inspection information;

[0065] S130: Remove data records with missing values ​​and data that are obviously inconsistent with reality in the multidimensional data, and normalize the remaining data.

[0066] Specifically, elevator attribute information includes basic information such as elevator model, specifications, manufacturer, installation time, and years of operation. IoT monitoring information includes real-time collection of elevator operating status, fault information, passenger information, operating environment data, etc. Maintenance information includes maintenance time, maintenance content, maintenance personnel, maintenance results, fault handling records, etc. Inspection and testing information includes regular or irregular inspection results, inspection reports, problems found during inspections, and solutions.

[0067] In a specific embodiment, the collected data may be collected from the following aspects:

[0068] The unit that manufactures (including remodels) elevators: the main safety components, bridge compartments, control cabinets, counterweights and mainframes, etc., the distribution of the shaft and other influencing factors. The user unit: elevator safety management personnel, use occasions, the environment below the pit, the environment inside the machine room, elevator insurance, the location of the alarm device, the management of rental housing, rescue channels, use registration information, management rules and regulations. The installation (including maintenance) unit: the installation unit's license level, the original manufacturer's installation, the supervision and inspection pass rate and other data. The maintenance unit: maintenance contract, contract filing, the level of the maintenance unit, original factory maintenance, regular inspection pass rate, and the cycle of advance inspection. Inspection agency: statutory inspection items, major issues, safety hazard investigation, safety assessment. Supervisory agency: elevator accidents, complaints, and safety risks and defects of elevators of the same model and type.

[0069] In a specific embodiment, S130 includes:

[0070] Plot data lists according to multidimensional data;

[0071] Perform column indexing on the data in the data list, traverse the missing values ​​in the row, determine the data completeness, and determine the completeness of the data in the data list;

[0072] The data with a completeness greater than or equal to a first preset threshold is retained, and the data with a completeness less than the first preset threshold is removed;

[0073] Data records that are obviously inconsistent with reality are deleted from the data list by calculating a deviation value, wherein the deviation value is calculated based on an absolute value or a relative error. When the deviation value is greater than or equal to a second preset threshold, the data record is marked as inconsistent with reality and deleted;

[0074] The remaining data were normalized.

[0075] Through the above technical solutions, the accuracy and completeness of the data are ensured, and the data is converted into a unified format to facilitate subsequent analysis and modeling.

[0076] In a specific embodiment, constructing a BP neural network model includes: an input layer, a hidden layer, and an output layer;

[0077] The number of nodes in the input layer is determined according to the number of features of the input data, the number of nodes in the output layer is 1, and the evaluation result of the health status of the elevator is output. The number of nodes in the hidden layer is obtained according to an empirical formula.

[0078] Specifically, the empirical formula is:

[0079]

[0080] Among them, n1 is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, and a is a constant between 1 and 10.

[0081] In a specific embodiment, an overly complex model may lead to overfitting problems. Therefore, when constructing a GA-BP neural network, it is necessary to select appropriate network structures and parameters to avoid overfitting. Therefore, the genetic algorithm GA is used as a strategy to optimize the hyperparameters, weights and thresholds of the BP neural network model to obtain a GA-BP neural network model, which specifically includes:

[0082] The initial population is established by real number coding. The initial population consists of four parts: hidden layer threshold, connection weight between hidden layer and output layer, connection weight between input layer and hidden layer, and output layer threshold, as a set of chromosomes.

[0083] Perform selection, crossover and mutation operations of genetic algorithm to generate the next generation population, calculate the individual fitness value of the new population, and take the minimum value as the optimal value;

[0084] The optimized connection weights and thresholds are introduced into the BP neural network, and the normalized data are used as training samples for repeated training.

[0085] Update the optimized connection weights, thresholds and network structure obtained through training to obtain a GA-BP neural network model;

[0086] The first evaluation result of the elevator health status is obtained based on the GA-BP neural network model.

[0087] In a specific embodiment, the expression for calculating the individual fitness value of the new population is:

[0088]

[0089] Where n is the number of nodes in the network output layer, y i is the actual output of the ith node; y i ′ is the expected output of the ith node.

[0090] By optimizing the initial weights and thresholds of the BP neural network through genetic algorithms, the training efficiency and prediction performance of the neural network can be improved.

[0091] In a specific embodiment, if there are few data samples of certain fault types, the model may have poor recognition ability for these fault types. Therefore, in S300, a logic judgment model is constructed to obtain a second evaluation result of the elevator health status, and a logical judgment is performed based on the real-time collected elevator operation data and the defined logical relationship to draw a conclusion on whether the elevator is abnormal. Specifically, it includes:

[0092] Define logical relationships to determine the health status of elevators;

[0093] Among them, propositions of logical relations include:

[0094] The relationship between normal operation and fault conditions;

[0095] The relationship between maintenance operation and operation status;

[0096] The relationship between emergency electric operation and fault handling;

[0097] The relationship between docking operations and loading and unloading of cargo;

[0098] The relationship between fire operations and fire emergency response;

[0099] As shown in Table 1, the defined logical questions are organized into a rule base, and based on the rule base, a logical judgment model is obtained. The following rule base can be established to judge the health status of the elevator.

[0100] Table 1 Rule base statistics

[0101]

[0102]

[0103] Rule base application: embed the rule base in the elevator health status assessment, and automatically trigger corresponding response actions by real-time monitoring of the elevator's multi-dimensional information.

[0104] Rule base update: With the continuous advancement of elevator technology and the diversification of fault types, the rule base needs to be updated and improved regularly to ensure its accuracy and effectiveness.

[0105] Through the above steps, a complete set of elevator health status judgment logic and rule library can be established, which is conducive to the evaluation of elevator health status and provides strong support for the safe operation and maintenance of elevators.

[0106] In a specific embodiment, the logic judgment model expression is:

[0107] Exception = P00N AND P00M;

[0108] Among them, P00N and P00M represent the truth or falsity of any two propositions of logical relationship.

[0109] Specifically, the logical AND operation is used, and the logical AND operation is described by 0 when there is 0 and all 1s when there are 1s. P00N and P00M are the true or false states of the propositions of any two logical relations of P001-P007 in the above table, that is, the logical value of 0 or 1 (1 means normal, 0 means abnormal). For example, P00NP is P001=1, and P00M is P003=0. Therefore, the elevator is running normally, but slight noise or vibration is detected, 1AND0=0, which means that the health of the elevator is abnormal and it needs to stop running and switch to emergency electric operation or maintenance mode.

[0110] The logic judgment model in this embodiment is a dual space model of logic judgment, which is actually a combination of a series of logic rules and judgments, and is used to judge whether the elevator is healthy according to multiple state parameters of the elevator. These logic rules can be mathematically represented by logical expressions and Boolean algebra, so as to facilitate understanding and calculation. In practical applications, the construction of logical expressions and the values ​​of parameters need to be determined according to the specific elevator system and health diagnosis requirements.

[0111] Specifically, the GA-BP neural network model outputs a probability vector and selects the category with the highest probability as N output .

[0112] The logical judgment model directly outputs the category label (assuming it is 0 or 1)

[0113] In a specific embodiment, S400 includes:

[0114] The binary value output by the logical judgment model is converted into a continuous value of the same range as the output by the GA-BP neural network model, for example by mapping 0 to a low score (such as 0) and 1 to a high score (such as 1, or more commonly, a floating point number indicating high confidence, such as 0.9).

[0115] The fusion result is calculated as the final elevator health diagnosis result, and the expression is:

[0116] F=W1*N output +W2*L output ;

[0117] Where F represents the final elevator health diagnosis result, N output represents the first evaluation result of the elevator health status output by the GA-BP neural network model, L outputrepresents the second evaluation result of the elevator health status after the logic judgment model is converted, W1 represents the weight of the GA-BPGA-BP neural network model, and W2 represents the weight of the logic judgment model;

[0118] The fusion results are graded and used as the final elevator health status diagnosis result.

[0119] In a specific embodiment, the grading includes: scoring the elevator into five major levels: one, two, three, four, and five, and judging whether there are problems with the health status of the elevator throughout its life cycle and predicting the existing risks through the score levels.

[0120] Through the above steps, the present invention combines the learning and prediction capabilities of neural networks with the intuitiveness and precision of logical judgment to improve the accuracy and efficiency of elevator abnormality diagnosis.

[0121] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0122] 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 method for evaluating the health status of an elevator based on a fusion network model, characterized in that: The following steps are involved: S100: Collect multi-dimensional data of the elevator to be tested and perform pre-processing; S200: constructing a BP neural network model, using a genetic algorithm GA as a strategy for optimizing hyperparameters, weights and thresholds of the BP neural network model, obtaining a GA-BP neural network model, and obtaining a first evaluation result of the health status of the elevator based on the GA-BP neural network model; S300: constructing a logic judgment model to obtain a second evaluation result of the elevator health status; S400: Fusing the first evaluation result and the second evaluation result, and grading the fusion result, and using the grading result as the final elevator health status diagnosis result.

2. The elevator health status assessment method based on the fusion network model according to claim 1 is characterized in that: S100: Collect multi-dimensional data of the elevator to be tested and perform pre-processing, including: S110: elevator attribute information of the elevator on standby for testing; S120: Setting a sampling period, and collecting multi-dimensional data of the elevator to be tested based on the sampling period, wherein the multi-dimensional data includes: IoT monitoring information, maintenance information, and inspection information; S130: Remove data records with missing values ​​and data that are obviously inconsistent with reality in the multidimensional data, and normalize the remaining data.

3. The elevator health status assessment method based on the fusion network model according to claim 2 is characterized in that: The S130 includes: Plot data lists according to multidimensional data; Perform column indexing on the data in the data list, traverse the missing values ​​in the row, determine the data completeness, and determine the completeness of the data in the data list; The data with a completeness greater than or equal to a first preset threshold is retained, and the data with a completeness less than the first preset threshold is removed; Data records that are obviously inconsistent with reality are deleted from the data list by calculating a deviation value, wherein the deviation value is calculated based on an absolute value or a relative error. When the deviation value is greater than or equal to a second preset threshold, the data record is marked as inconsistent with reality and deleted; The remaining data were normalized.

4. The elevator health status assessment method based on the fusion network model according to claim 1 is characterized in that: The BP neural network model comprises: an input layer, a hidden layer and an output layer; The number of nodes in the input layer is determined according to the number of features of the input data, the number of nodes in the output layer is 1, and the evaluation result of the health status of the elevator is output. The number of nodes in the hidden layer is obtained according to an empirical formula.

5. The elevator health status assessment method based on the fusion network model according to claim 4 is characterized in that: The empirical formula is: Among them, n1 is the number of hidden layer nodes, n is the number of input layer nodes, m is the number of output layer nodes, and a is a constant between 1 and 10.

6. The elevator health status assessment method based on the fusion network model according to claim 1 is characterized in that: The genetic algorithm GA is used as a strategy to optimize the hyperparameters, weights and thresholds of the BP neural network model, and the GA-BP neural network model is obtained, including: The initial population is established by real number coding. The initial population consists of four parts: hidden layer threshold, connection weight between hidden layer and output layer, connection weight between input layer and hidden layer, and output layer threshold, as a set of chromosomes. Perform selection, crossover and mutation operations of genetic algorithm to generate the next generation population, calculate the individual fitness value of the new population, and take the minimum value as the optimal value; The optimized connection weights and thresholds are introduced into the BP neural network, and the normalized data are used as training samples for repeated training. Update the optimized connection weights, thresholds and network structure obtained through training to obtain a GA-BP neural network model; The first evaluation result of the elevator health status is obtained based on the GA-BP neural network model.

7. The elevator health status assessment method based on the fusion network model according to claim 6 is characterized in that: The expression for calculating the individual fitness value of the new population is: In the formula, n is the number of nodes in the network output layer, y i is the actual output of the ith node; y′ i is the expected output of the ith node.

8. The elevator health status assessment method based on the fusion network model according to claim 1 is characterized in that: S3 00: Build a logical judgment model to obtain the second evaluation result of the elevator health status, including: Define logical relationships to determine the health status of elevators; Among them, propositions of logical relations include: The relationship between normal operation and fault conditions; The relationship between maintenance operation and operation status; The relationship between emergency electric operation and fault handling; The relationship between docking operations and loading and unloading of cargo; The relationship between fire operations and fire emergency response; The defined logical questions are organized into a rule base, and based on the rule base, a logical judgment model is obtained.

9. The elevator health status assessment method based on the fusion network model according to claim 8 is characterized in that: The logical judgment model expression is: Exception = P00N AND P00M; Among them, P00N and P00M represent the truth or falsity of any two propositions of logical relationship.

10. The elevator health status assessment method based on the fusion network model according to claim 1 is characterized in that: S400, including: The binary values ​​output by the logic judgment model are converted into continuous values ​​in the same range as those output by the GA-BP neural network model; Calculate the fusion result, the expression is: F=W1*N output +W2*L output ; Where F represents the final elevator health diagnosis result, N output represents the first evaluation result of the elevator health status output by the GA-BP neural network model, L output represents the second evaluation result of the elevator health status after the logic judgment model is converted, W1 represents the weight of the GA-BPGA-BP neural network model, and W2 represents the weight of the logic judgment model; The fusion results are graded and used as the final elevator health status diagnosis result.

Images