Human-machine collaborative health assessment method for fan transmission chain

Abstract

The invention belongs to the related technical field of intelligent operation and maintenance of wind power equipment, and discloses a fan transmission chain man-machine collaborative health assessment method, which comprises the following steps: simulating fault modes of a fan transmission chain, obtaining vibration signals in different fault modes, and training a constructed neural network model; obtaining an evaluation set of BtO and OtW about the risk degree of each fault mode; constructing a Bayesian hierarchical model, converting the risk weights of the fault modes and the evaluation set into a probability distribution form, and based on the Bayesian hierarchical model, determining an aggregation risk weight of each fault mode fused with risk opinions in the evaluation set; and evaluating the fault state membership degree of the fan transmission chain by using the trained neural network model, calculating a health degree index by fusing the fault state membership degree and the aggregation risk weight, and evaluating the health state of the fan transmission chain. Through combination of monitoring big data and operation and maintenance experience knowledge, accurate quantitative evaluation of the health state of the fan transmission chain can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field related to intelligent operation and maintenance of wind power equipment, and more specifically, relates to a human-machine collaborative health assessment method for a wind turbine transmission chain. Background Art

[0002] The operating environment of wind power equipment is complex and changeable. Long-term exposure to extreme harsh environments such as hurricanes, thunderstorms, and humidity makes its transmission chain components prone to fatigue, wear, fracture, corrosion and other damages of varying degrees and forms. The functional characteristics and operating efficiency gradually decline, seriously affecting its long-term stable and reliable operation capabilities. In addition, due to the accessibility restrictions of wind power equipment operation and maintenance, wind turbine maintenance work cannot be carried out as scheduled like conventional equipment. Limited accessibility conditions, harsh working environment, and quantitative maintenance resources make wind turbine transmission chain components typical multi-constrained limited access repair devices. With the in-depth advancement of Industry 4.0 and the continuous breakthroughs in emerging technologies such as multi-modal sensing, industrial big data, and artificial intelligence, intelligent operation and maintenance technology is considered to be a key entry point and breakthrough for improving the reliability, maintainability, and safety of wind turbine transmission chains.

[0003] According to the survey, most of the current wind turbine drive chain operation and maintenance technologies are driven by monitoring data and models under known ideal conditions. Due to the extremely harsh working environment of the wind turbine drive chain, complex and changeable operating conditions, uncertain damage and failure mechanisms, and a lack of fault samples, existing technologies are difficult to meet the needs of wind turbine drive chains for accurate health assessment. In recent years, with the rapid development of intelligent Internet of Things, new sensor and communication technologies and their application in wind power equipment, the operation monitoring data, working environment and health status monitoring data of wind turbine drive chains can be collected and aggregated in a timely manner. The collected monitoring data is transmitted back at a frequency of seconds, and the amount of new data added every day is as high as about 1.5TB.

[0004] Existing wind turbine transmission chain monitoring big data can effectively diagnose transmission chain failure modes and conduct transmission chain health assessment. However, existing assessment methods are usually qualitative judgments and cannot achieve accurate quantitative assessment of the health status of the transmission chain. Summary of the invention

[0005] In response to the above defects or improvement needs of the prior art, the present invention provides a human-machine collaborative health assessment method for a wind turbine transmission chain, which is used to solve the problem that the existing health assessment method for a wind turbine transmission chain is usually a qualitative judgment and cannot achieve accurate quantitative assessment of the health status of the transmission chain.

[0006] To achieve the above object, according to the present invention, a human-machine collaborative health assessment method for a wind turbine transmission chain is provided, comprising:

[0007] Offline training and data processing stage:

[0008] Determine and simulate the failure mode of the fan drive chain, obtain the vibration signal of the fan drive chain during operation under different failure modes, and use the vibration signal to train the constructed neural network model, and use the neural network model to evaluate the failure state membership of the fan drive chain;

[0009] Obtaining the BtO and OtW evaluation sets of the risk levels of each failure mode; constructing a Bayesian hierarchical model, converting the risk weight of the failure mode and the evaluation set into a probability distribution form, and determining the aggregated risk weight of each failure mode that incorporates all risk opinions in the evaluation set based on the Bayesian hierarchical model;

[0010] Online Assessment Phase:

[0011] The vibration signal during the operation of the fan transmission chain is input into the offline trained neural network model to obtain the fault state membership of the fan transmission chain; the health index is calculated by fusing the fault state membership with the aggregated risk weight based on the Ratio system method, and the health state of the fan transmission chain is evaluated based on the health index.

[0012] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the neural network model is a multifunctional parallel network, and the multifunctional parallel network includes a counting self-attention module and a positioning self-attention module, wherein:

[0013] The counting self-attention module is used to calculate the number of faults, and the positioning self-attention module is used to calculate the fault probability distribution;

[0014] The multifunctional parallel network is used to combine the fault quantity calculated by the counting self-attention module with the fault probability distribution result calculated by the positioning self-attention module to obtain the fault state membership.

[0015] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the fault state membership is calculated according to the following formula:

[0016]

[0017] Where p is the sample number, the sample is a vibration signal, p = 1, 2, ..., P, P is the number of samples; f ( p ) is the time domain and frequency domain characteristics of the vibration signal; FSM (f ( p ) ) is the fault state membership; is the predicted number of failures; FPDj(f ( p )) is the failure probability distribution; θj is the model parameter under each failure mode category; j is the failure mode number; N is the maximum number among the failure mode numbers; L (p) Number labels for failure modes; is the calculation result of the localization self-attention module applicable to fault mode j in the layer before the output layer.

[0018] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the evaluation set is obtained through expert evaluation by an expert group, and obtaining the evaluation set specifically includes:

[0019] Obtaining evaluation information of the most serious failure mode and the least serious failure mode from any expert in the expert group;

[0020] Based on the evaluation information of any expert, the BtO vector and OtW vector of each fault mode risk degree corresponding to the evaluation information of any expert are obtained, and the BtO vectors and OtW vectors of all experts in the expert group are summarized to form the evaluation set.

[0021] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the risk weight of the fault mode and the evaluation set are converted into a probability distribution form in the Bayesian hierarchical model as follows:

[0022] The failure mode risk weights and the evaluation set are converted into a joint probability distribution form through the Bayesian hierarchical model; wherein the failure mode risk weights include a single risk weight of each failure mode determined based on the evaluation information of a single expert and an aggregate risk weight of each failure mode determined by integrating the opinions of all experts in the expert group.

[0023] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the aggregated risk weights of all expert opinions of the expert group for each failure mode determined in S2 specifically include:

[0024] Based on the evaluation information of a single expert in the expert group, a single risk weight of each failure mode determined based on the evaluation information of the single expert is obtained;

[0025] The vectors in the evaluation set and the single risk weight of each failure mode are converted into a multinomial distribution, and the single risk weight of each failure mode and the aggregate risk weight are converted into a Dirichlet distribution. The aggregate risk weight is determined by combining the Bayesian hierarchical model.

[0026] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the Bayesian hierarchical model is specifically:

[0027]

[0028] Among them, P(·) represents the probability distribution form; and are the BtO vector and OtW vector in the fault risk assessment set given by K experts, is the single fault risk weight vector derived from the K-bit expert evaluation information. is the single risk weight of failure mode j derived from the evaluation information of the kth expert, is the aggregate risk weight of failure mode j; and The failure risk BtO vector and OtW vector provided by the k-th expert; K is the total number of experts.

[0029] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the single risk weight of the fault mode j is determined based on the evaluation information of the kth expert. It is calculated by the following formula:

[0030]

[0031] in, and is the single risk weight of the most serious failure mode and the least serious failure mode determined according to the opinion of the kth expert; N is the maximum number in the failure mode number; The severity of the most serious failure mode provided by the kth expert relative to the jth failure mode, The severity of the jth failure mode relative to the least severe failure mode provided by the kth expert.

[0032] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the polynomial distribution is performed according to the following relationship:

[0033]

[0034] in, and The failure risk BtO vector and OtW vector provided by the k-th expert; is the single risk weight of failure mode j derived from the evaluation information of the kth expert; K is the total number of experts;

[0035] The Dirichlet distribution is performed according to the following relationship:

[0036]

[0037] in, is the aggregate risk weight of failure mode j; Dir(·) is a Dirichlet distribution; γ is a non-negative parameter with a GAMMA distribution.

[0038] According to the human-machine collaborative health assessment method for a wind turbine transmission chain provided by the present invention, the health index is calculated according to the following relationship:

[0039]

[0040] Among them, HI is the health index; DI is the damage index; AC is the amplification factor; Acc is the diagnostic accuracy of the neural network model; is the aggregate risk weight of fault mode j; FSMj is the jth element in the fault state membership; z is the Benefit criterion number; is the count of the number of faults predicted by the attention module.

[0041] In general, compared with the prior art, the above technical solution conceived by the present invention provides a human-machine collaborative health assessment method for a wind turbine transmission chain:

[0042] 1. Construct a neural network model to calculate the equipment fault state membership, then determine the aggregate risk weight of the fault mode based on the evaluation set obtained by the BWM evaluation method and the Bayesian hierarchical model, and then calculate the health index based on the Ratio system method to integrate the fault state membership and the aggregate risk weight, and conduct a health assessment on the health of the wind turbine transmission chain. This method combines monitoring big data with operation and maintenance experience and knowledge to achieve accurate quantitative assessment of the health status of the wind turbine transmission chain;

[0043] 2. By building a multifunctional parallel network, the number of faults in the wind turbine drive chain and the probability of each fault occurring can be determined, and the fault state membership of the output device can be combined to accurately quantify the evolution stage of the detected fault;

[0044] 3. Constructing the GCD method: First, based on the BWM method, K experts evaluate the risk level of each fault mode to obtain the evaluation set, and then construct a Bayesian hierarchical model, combining the use of multinomial distribution function and Dirichlet distribution function to finally determine the aggregated fault risk weight. The fault risk preference opinions of all experts can be comprehensively considered to give the optimal aggregated fault risk weight evaluation result;

[0045] 4. The extended Ratio system method proposed can consider the impact of the fault diagnosis model accuracy on the confidence of the results. On this basis, the fault state membership of the wind turbine transmission chain and the fault aggregation risk weight are combined to give the final health assessment result of the wind turbine transmission chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of the human-machine collaborative health assessment method for a fan transmission chain provided by the present invention;

[0047] Figure 2 is a schematic diagram of a human-machine collaborative health assessment method for a wind turbine transmission chain constructed according to a specific embodiment of the present invention;

[0048] Figure 3 yes Figure 2 Schematic diagram of the human-machine collaborative health assessment framework for the wind turbine transmission chain;

[0049] Figure 4 yes Figure 2 Versatile parallel network diagram in the human-machine collaborative health assessment method for the fan transmission chain;

[0050] Figure 5 It is the equipment health assessment result under various fault conditions constructed according to the specific embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0052] See also Figure 1 This embodiment provides a human-machine collaborative health assessment method for a wind turbine transmission chain, the method comprising:

[0053] Offline training and data processing stage:

[0054] Determine and simulate the failure mode of the fan drive chain, obtain the vibration signal of the fan drive chain during operation under different failure modes, and use the vibration signal to train the constructed neural network model, and use the neural network model to evaluate the failure state membership of the fan drive chain;

[0055] Obtain the BtO (Best-to-other) and OtW (Other-to-worst) evaluation sets of the risk level of each failure mode; construct a Bayesian hierarchical model, convert the risk weight of the failure mode and the evaluation set into a probability distribution form, and determine the aggregated risk weight of each failure mode that integrates all risk opinions in the evaluation set based on the Bayesian hierarchical model; the evaluation set can be evaluated and obtained based on the BWM method (Best to worst method) to evaluate the risk level of each failure mode, so as to form an evaluation set including BtO and OtW evaluation information; the evaluation set brings together the empirical knowledge related to wind turbine operation and maintenance, and the aggregated risk weight obtained based on the evaluation set is more objective and reliable;

[0056] Online Assessment Phase:

[0057] The vibration signal during the operation of the fan transmission chain is input into the offline trained neural network model to obtain the fault state membership of the fan transmission chain; the health index is calculated by fusing the fault state membership with the aggregated risk weight based on the Ratio system method, and the health state of the fan transmission chain is evaluated based on the health index.

[0058] In some specific embodiments, the neural network model is a multifunctional parallel network, which includes a counting self-attention module and a positioning self-attention module, wherein:

[0059] The counting self-attention module is used to calculate the number of faults, and the positioning self-attention module is used to calculate the fault probability distribution;

[0060] The multifunctional parallel network is used to combine the fault quantity calculated by the counting self-attention module with the fault probability distribution result calculated by the positioning self-attention module to obtain the fault state membership.

[0061] refer to Figure 2 The human-machine collaborative health assessment method for a wind turbine transmission chain provided in this embodiment mainly includes the following steps:

[0062] Step 1: Determine the typical fault mode of the wind turbine drive chain to be simulated, analyze the vibration response characteristics of the drive chain under different fault modes, and then place vibration sensors to collect vibration signals during the operation of the drive chain under different fault models.

[0063] Specifically, the vibration signal of the fan drive chain is collected by a NI-cDAQ-9174 / 9234 vibration sensor with a sampling frequency of 10240Hz. The sampling duration is set to 100s, and 1024000 data points are obtained in each fault mode. 510 samples of data points in each fault mode can be evenly extracted for training the neural network model. The fan drive chain is usually a gearbox, which mainly has gears, bearings, couplings and other components. The vibration sensor can be placed in the middle of the base of the fan equipment gearbox, which can better reflect the vibration of the gearbox.

[0064] Step 2: The vibration signals collected under each fault mode are used as samples to train the constructed multifunctional parallel network, and then the trained Versatile parallel network is used to evaluate the fault state membership of the wind turbine transmission chain.

[0065] Specifically, refer to Figure 3 and Figure 4, the constructed multifunctional parallel network adopts self-attention neural network. The Versatileparallel network includes a counter self-attention module and a locator self-attention module.

[0066] The calculation of the fault state membership degree comprises the following steps:

[0067] Label the number of faults for the vibration signal and extract its time domain and frequency domain features. The number of faults is used as the module output and the extracted features are used as input to train the Counter self-attention module. The number of faults is determined through the Locator self-attention module.

[0068] The vibration signal is labeled with a fault mode label, and its time domain and frequency domain features are extracted. The fault mode is used as the module output, and the extracted features are used as input to train the Locator self-attention module. The SoftMax activation function in the Locator self-attention module is modified to a SoftProb function, so that its output is a fault probability distribution;

[0069] Combine the number of faults calculated by the Counter self-attention module with the fault probability distribution result calculated by the Locator self-attention module to obtain the fault state membership.

[0070] Specifically, the Counter self-attention module is used to obtain the number of equipment faults. First, the vibration signal is extracted in the time domain and frequency domain. Then, the corresponding fault number label u is (p) ∈(1,...,U) time domain and frequency domain features f (p) (p=1,...,P) is input into the Counter self-attention module for model training, and U is the number of failure modes with the largest number of failures. The prediction result of the number of failures is:

[0071]

[0072] in, is the prediction result of the number of failures, W1 C and W2 C is the weight of the Counter self-attention module, b1 C and It is the bias of the Counter self-attention module.

[0073] The Locator self-attention module is used to obtain the fault probability distribution. p ∈(1,..NN,) time domain and frequency domain features f (p) (p=1,...,P) trains the Locator self-attention module, N is the maximum number in the fault mode number, and the fault category label can be the number of the fault mode. The fault category classification result is:

[0074]

[0075] in, is the failure mode classification result, W1 L and W2 L is the weight of the Locator self-attention module, b1 L and It is the bias of the Locator self-attention module.

[0076] In order to enable the model to identify multiple fault modes at the same time, modify the objective function of the Locator self-attention module:

[0077]

[0078] Where P is the total number of samples. If sample p belongs to fault mode j, E ij The value is 1 if yes, otherwise 0.

[0079] Then, the fault probability distribution FPD (f ( p ) )as follows:

[0080]

[0081] Among them, FPDj(f ( p ) ) is the failure probability distribution; j is the failure mode number; L ( p ) Label the failure modes; θ j is the model parameter under the jth fault category; is the calculation result of the localization self-attention module applicable to fault mode j in the layer before the output layer; N is the maximum number in the fault mode number; It is the result of normalization of the score of the j-th fault mode by the Softprob module.

[0082] The fault counting module is combined with the results of the fault probability distribution calculation module to finally obtain the fault state membership. The fault state membership is calculated according to the following formula:

[0083]

[0084] Where p is the sample number, the sample is a vibration signal, p = 1, 2, ..., P, P is the number of samples; f ( p ) is the time domain and frequency domain characteristics of the vibration signal; FSM (f ( p ) ) is the fault state membership; is the predicted number of failures.

[0085] Step 3, reference Figure 3 In some specific embodiments, the evaluation set is obtained through evaluation by experts of an expert group. Specifically, a failure mode risk assessment expert group can be established to obtain evaluation information of the risk degree of each failure mode by experts of the expert group using the BWM method (Best to worst method), and form an evaluation set of failure mode risks Best-to-other and Other-to-worst; K experts can be invited to form an expert group for fault mode risk assessment of a wind turbine transmission chain. Experts of the expert group use the BWM method to evaluate the risk degree of each failure mode causing harm to the equipment based on their own knowledge and experience using natural semantics, and then convert it into a failure risk BtO / OtW evaluation set.

[0086] Specifically, obtaining the evaluation set includes:

[0087] Obtain the evaluation information of the most serious fault mode and the least serious fault mode by any expert in the expert group; that is, the K invited experts each determine the most serious fault and the least serious fault respectively;

[0088] Based on the evaluation information of any expert, the Best-to-Others (BtO) vector and Others-to-Worst (OtW) vector of each failure mode risk corresponding to the evaluation information of any expert are obtained, and the BtO vectors and OtW vectors of all experts in the expert group are summarized to form the evaluation set. That is, the Best to worst method is used to give the failure risk Best-to-Others (BtO) vector and Others-to-Worst (OtW) vector, and the two types of vectors of K experts are summarized to form the evaluation set.

[0089] In other embodiments, the evaluation set may be obtained in other forms, such as directly selecting relevant evaluation sets from existing research, or training an evaluation model based on existing data, obtaining the evaluation set through the evaluation model, or other artificial intelligence methods. The specific acquisition method is not limited.

[0090] Step 4: Based on the evaluation set given by the experts in the expert group, by constructing a Bayesian hierarchical model, the evaluation opinions of each expert and the risk weight of each failure mode are converted into a probability distribution, and the multinomial distribution function and the Dirichlet distribution function are comprehensively used to finally determine the aggregated failure risk weight.

[0091] The Bayesian hierarchical model converts the risk weight of the failure mode and the evaluation set into a probability distribution form as follows:

[0092] The Bayesian hierarchical model is used to transform the fault mode risk weight and the evaluation set into a joint probability distribution form; that is, the Bayesian hierarchical model is used to establish a functional relationship between the risk weight of the fault mode and the vector in the evaluation set in the form of probability distribution. The fault mode risk weight includes the single risk weight of each fault mode determined based on the evaluation information of a single expert and the aggregate risk weight of each fault mode determined by integrating the opinions of all experts in the expert group. In the joint probability distribution, the risk weight of the fault mode includes the single risk weight and the aggregate risk weight of each fault mode; the evaluation set includes the BtO vector and OtW vector of K experts.

[0093] Furthermore, the aggregate risk weights of all expert opinions of the expert group for each failure mode determined in S2 specifically include:

[0094] Based on the evaluation information of a single expert in the expert group, a single risk weight of each failure mode determined based on the evaluation information of the single expert is obtained; first, based on the evaluation information of the single expert, specifically the BtO vector and the OtW vector corresponding to the evaluation information of the single expert, the single risk weight of each failure mode corresponding to the evaluation information of the single expert can be calculated;

[0095] Then, the vectors in the evaluation set and the single risk weight of each failure mode are converted into a multinomial distribution, and the single risk weight of each failure mode and the aggregate risk weight are converted into a Dirichlet distribution, and the aggregate risk weight is determined by combining the Bayesian hierarchical model. After determining the single risk weight of each failure mode corresponding to the evaluation information of each expert, the unknown variable in the joint probability distribution determined by the Bayesian hierarchical model is only the aggregate risk weight, so the aggregate risk weight can be obtained by combining the multinomial distribution and the Dirichlet distribution.

[0096] Specifically, a Bayesian hierarchical model is constructed to convert the fault risk BtO / OtW evaluation set provided by K experts into a probability distribution form; the Bayesian hierarchical model is specifically as follows:

[0097]

[0098] Among them, P(·) represents the probability distribution form; and are the BtO vector and OtW vector in the fault risk assessment set given by K experts, is the single fault risk weight vector derived from the K-bit expert evaluation information. is the single risk weight of failure mode j derived from the evaluation information of the kth expert, is the aggregate risk weight of failure mode j; and The failure risk BtO vector and OtW vector provided by the k-th expert; K is the total number of experts.

[0099] The single risk weight of failure mode j determined based on the evaluation information of the kth expert It is calculated by the following formula:

[0100]

[0101] in, and is the single risk weight of the most serious failure mode and the least serious failure mode determined according to the opinion of the kth expert; N is the maximum number in the failure mode number; The severity of the most serious failure mode provided by the kth expert relative to the jth failure mode, The severity of the jth failure mode relative to the least severe failure mode provided by the kth expert.

[0102] The polynomial distribution is performed according to the following relationship:

[0103]

[0104] in, and The failure risk BtO vector and OtW vector provided by the k-th expert; is the single risk weight of failure mode j derived from the evaluation information of the kth expert; K is the total number of experts;

[0105] The Dirichlet distribution is performed according to the following relationship:

[0106]

[0107] in, is the aggregate risk weight of failure mode j; Dir(·) is a Dirichlet distribution; γ is a non-negative parameter with a GAMMA distribution.

[0108] This embodiment proposes a group confidence decision (GCD) method, which includes first obtaining an evaluation set by evaluating the risk degree of each fault mode by K experts based on the BWM method, then constructing a Bayesian hierarchical model, combining the use of a multinomial distribution function and a Dirichlet distribution function, and finally determining the aggregated fault risk weight. The proposed group confidence decision method can integrate the opinions of K experts and give a comprehensive and reliable fault risk weight assessment result.

[0109] Step 5: The extended Ratio system method is used to integrate the fault state membership and the aggregated risk weight to calculate the equipment health index, thereby making an overall comprehensive quantitative assessment of the health of the wind turbine transmission chain. After obtaining the state membership and the aggregated risk weight, the equipment health index is analyzed and obtained through the extended Ratio system method. The health index is calculated according to the following relationship:

[0110]

[0111] Among them, HI is the health index; DI is the damage index; AC is the amplification factor; Acc is the diagnostic accuracy of the neural network model, and the diagnostic accuracy is the ratio of the number of correctly diagnosed samples of the network to the total number of diagnosed samples; is the aggregate risk weight of fault mode j; FSMj is the jth element in the fault state membership; z is the Benefit criterion number; is the number of faults predicted by the counting self-attention module. When the network diagnosis accuracy is low, the amplification factor is introduced to increase the damage index to reflect the impact of the uncertainty of the fault risk on the health status of the wind turbine.

[0112] In order to further explain the embodiments of the present invention in detail, in some specific embodiments, wind power gearbox fault simulation experimental data is used to verify the method. The wind power gearbox fault simulation test bench consists of two ABB mv1008-225 motors (1.2kW), a gearbox, a flywheel, a set of data acquisition instruments and a computer. One of them is used as a prime mover to drive a multi-stage gearbox, and the other is used as an asynchronous generator to simulate various resistance torques. In the experiment, the input speed of the device was set to 1400 rpm, and the speeds of the two meshing gear sets in the gearbox were 1184 rpm and 840 rpm respectively.

[0113] The health of the wind turbine gearbox needs to be evaluated under fifteen fault modes, namely, one health state (H1), six single fault conditions (IFS1-IFS6), and eight compound fault conditions (CFS1-CFS8). The specific fault types and corresponding labels are shown in Table 1. In this specific embodiment, the six single fault conditions are numbered, that is, N=6, and the number of the compound fault condition can be a combination of the single fault condition numbers. The six common single fault conditions involved in this experiment include broken teeth, gear shedding, gear cracks, loose couplings, bearing rolling element wear, and bearing outer ring wear.

[0114] Table 1

[0115]

[0116]

[0117] The gearbox vibration signal is collected by a NI-cDAQ-9174 / 9234 vibration sensor with a sampling frequency of 10240Hz. The sampling duration is set to 100s, which means that 1024000 data points (i.e. one sensor × 100s × 10240Hz) are obtained in each fault state or fault mode. 510 samples of data points in each fault state are selected, and a total of 7650 samples are used as the data set. After being evenly mixed, the samples are divided into 60% training set, 10% validation set and 30% test set.

[0118] In order to verify the effectiveness of the Versatile parallel fault state membership evaluation network, based on the same data set mentioned above, this paper adopts 8 methods, namely 8 different networks for comparative analysis, which are recorded as method-1 to method-8, as described below.

[0119] Method-1: pairwise probabilistic multi-label classification (PPMLC)

[0120] Method-2: multilabel decision tree (MDT)

[0121] Method-3: gradient boosting decision tree (GBDT)

[0122] Method-4: random k labelsets (RAKEL)

[0123] Method-5: dual-extreme learning machine (Dual-ELM)

[0124] Method-6: multi-label radial basis function (ML-RBF)

[0125] Method-7: multi-label convolutional neural network with wavelettransform (WT-MLCNN)

[0126] Method-8: Multifunctional parallel network proposed by the present invention

[0127] To eliminate the randomness of the comparison results, thirty repeated tests were performed. The training and test diagnostic accuracy of the eight methods are recorded in Table 2. From the table, it can be seen that the diagnostic accuracy of the network of the present invention is significantly higher than that of the other seven methods. This further proves that the present invention has strong creativity and applicability and can be applied to actual industry.

[0128] Table 2

[0129]

[0130]

[0131] Table 3 shows the number of faults and the fault probability distribution obtained when the test set is input into the Versatile parallel network. It can be seen from Table 3 that most of the samples corresponding to the 15 fault conditions can be correctly distinguished. Combining the number of faults predicted by the Counter self-attention module and the fault probability distribution calculated by the Locator self-attention module, the equipment health index can be accurately evaluated.

[0132] Table 3

[0133]

[0134]

[0135] The risk weight of each single fault situation is determined by GCD. CL is the most serious fault with a fault risk weight as high as 0.3138, followed by LT (0.3109), BT (0.1447), CT (0.1332) and BRW (0.0974).

[0136] By expanding the Ratio system method, the fault state membership and fault risk weight are combined to obtain the overall health of the equipment under different fault conditions, such as Figure 5 shown.

[0137] The method of this embodiment collects the vibration signal generated by the wind turbine transmission chain, constructs a Versatile parallel network and determines the degree of membership of the equipment fault state; invites experts in the field of intelligent operation and maintenance to form an expert group for wind turbine transmission chain fault risk assessment. The experts in the expert group use the Group confidence decision (GCD) method based on the Best to Worst Method (BWM) to assess the fault risk, construct a Bayesian hierarchical model, convert the evaluation opinions of each expert into a probability distribution, and combine the polynomial distribution function and the Dirichlet distribution function to determine the aggregated fault risk weight; define the equipment health index formula through the extended Ratio system, calculate the equipment health index, and thus realize the health assessment of the wind turbine transmission chain. This embodiment solves the problem that the existing wind turbine health assessment method cannot comprehensively consider the opinions of all experts in the expert group, and achieves the purpose of comprehensive health assessment.

[0138] The human-machine collaborative health assessment method for the wind turbine transmission chain provided by the present invention uses a Versatile parallel network to calculate the equipment fault state membership, then determines the fault risk weight through the GCD model, and then calculates the health index based on the extended Ratio system method by integrating the fault state membership and the fault risk weight, and performs a health assessment on the health of the wind turbine transmission chain. The impact of different fault conditions on equipment behavior is measured at a finer granularity, and then the current health status of the equipment can be accurately reflected, providing guidance for subsequent maintenance and post-event monitoring strategies, improving maintenance efficiency, and having better flexibility.

[0139] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A human-machine collaborative health assessment method for a wind turbine transmission chain, characterized in that: include: Offline training and data processing stage: Determine and simulate the failure mode of the fan drive chain, obtain the vibration signal of the fan drive chain during operation under different failure modes, and use the vibration signal to train the constructed neural network model, and use the neural network model to evaluate the failure state membership of the fan drive chain; Obtaining the BtO and OtW evaluation sets of the risk levels of each failure mode; constructing a Bayesian hierarchical model, converting the risk weight of the failure mode and the evaluation set into a probability distribution form, and determining the aggregated risk weight of each failure mode that incorporates all risk opinions in the evaluation set based on the Bayesian hierarchical model; Online Assessment Phase: The vibration signal of the fan transmission chain during operation is input into the offline trained neural network model to obtain the fault state membership of the fan transmission chain; The health index is calculated by fusing the fault state membership and the aggregated risk weight based on the Ratio system method, and the health state of the wind turbine transmission chain is evaluated based on the health index.

2. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 1, characterized in that: The neural network model is a multifunctional parallel network, which includes a counting self-attention module and a positioning self-attention module, wherein: The counting self-attention module is used to calculate the number of faults, and the positioning self-attention module is used to calculate the fault probability distribution; The multifunctional parallel network is used to combine the fault quantity calculated by the counting self-attention module with the fault probability distribution result calculated by the positioning self-attention module to obtain the fault state membership.

3. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 2, characterized in that: The fault state membership is calculated according to the following formula: Where p is the sample number, the sample is a vibration signal, p = 1, 2, ..., P, P is the number of samples; f ( p ) is the time domain and frequency domain characteristics of the vibration signal; FSM (f ( p ) ) is the fault state membership; is the predicted number of failures; FPDj(f ( p ) ) is the failure probability distribution; θj is the model parameter under each failure mode category; j is the failure mode number; N is the maximum number among the failure mode numbers; L (p) Number labels for failure modes; is the calculation result of the localization self-attention module applicable to fault mode j in the layer before the output layer.

4. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 1, characterized in that: The evaluation set is obtained through evaluation by an expert group, and obtaining the evaluation set specifically includes: Obtaining evaluation information of the most serious failure mode and the least serious failure mode from any expert in the expert group; Based on the evaluation information of any expert, the BtO vector and OtW vector of each fault mode risk degree corresponding to the evaluation information of any expert are obtained, and the BtO vectors and OtW vectors of all experts in the expert group are summarized to form the evaluation set.

5. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 4, characterized in that: In the Bayesian hierarchical model, the risk weight of the failure mode and the evaluation set are converted into a probability distribution form as follows: The failure mode risk weights and the evaluation set are converted into a joint probability distribution form through the Bayesian hierarchical model; wherein the failure mode risk weights include a single risk weight of each failure mode determined based on the evaluation information of a single expert and an aggregate risk weight of each failure mode determined by integrating the opinions of all experts in the expert group.

6. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 5, characterized in that: The aggregate risk weights of all expert opinions of the expert group for each failure mode determined in S2 specifically include: Based on the evaluation information of a single expert in the expert group, a single risk weight of each failure mode determined based on the evaluation information of the single expert is obtained; The vectors in the evaluation set and the single risk weight of each failure mode are converted into a multinomial distribution, and the single risk weight of each failure mode and the aggregate risk weight are converted into a Dirichlet distribution. The aggregate risk weight is determined by combining the Bayesian hierarchical model.

7. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 5, characterized in that: The Bayesian hierarchical model is specifically: Among them, P(·) represents the probability distribution form; and are the BtO vector and OtW vector in the fault risk assessment set given by K experts, is the single fault risk weight vector derived from the K-bit expert evaluation information. is the single risk weight of failure mode j derived from the evaluation information of the kth expert, is the aggregate risk weight of failure mode j; and The failure risk BtO vector and OtW vector provided by the k-th expert; K is the total number of experts.

8. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 6, characterized in that: The single risk weight of failure mode j determined based on the evaluation information of the kth expert It is calculated by the following formula: in, and is the single risk weight of the most serious failure mode and the least serious failure mode determined according to the opinion of the kth expert; N is the maximum number in the failure mode number; The severity of the most serious failure mode provided by the kth expert relative to the jth failure mode, The severity of the jth failure mode relative to the least severe failure mode provided by the kth expert.

9. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 6, characterized in that: The polynomial distribution is performed according to the following relationship: in, and The failure risk BtO vector and OtW vector provided by the k-th expert; is the single risk weight of failure mode j derived from the evaluation information of the kth expert; K is the total number of experts; The Dirichlet distribution is performed according to the following relationship: in, is the aggregate risk weight of failure mode j; Dir(·) is a Dirichlet distribution; γ is a non-negative parameter with a GAMMA distribution.

10. The human-machine collaborative health assessment method for a wind turbine transmission chain according to claim 2, characterized in that: The health index is calculated according to the following relationship: Among them, HI is the health index; DI is the damage index; AC is the amplification factor; Acc is the diagnostic accuracy of the neural network model; is the aggregate risk weight of fault mode j; FSMj is the jth element in the fault state membership; z is the Benefit criterion number; is the count of the number of faults predicted by the attention module.

Images