A fault determination method, apparatus, device and storage medium

Abstract

This application discloses a fault determination method, apparatus, device, and storage medium, relating to the field of fault identification. The method includes: preprocessing an original signal to obtain preprocessed information, and performing a preset feature extraction operation on the preprocessed information to obtain feature data; inputting the feature data into a fault determination system for fault detection to obtain current fault information; and inputting the current fault information into a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results. This application uses a fault determination system to perform fault detection, and after obtaining the detection results, inputs the detection results into a manual fault evaluation interface for secondary evaluation. Based on the manual evaluation results, the target fault corresponding to the original signal is determined. Furthermore, it integrates an expert fault identification system, an AI fault identification system, and a manual fault identification system, improving the accuracy of fault identification through multiple fault identification judgments.

Description

Technical Field

[0001] This invention relates to the field of fault identification, and in particular to a fault determination method, apparatus, device, and storage medium. Background Technology

[0002] Rotating equipment has wide applications in industry, transportation, aviation, energy, and other fields. Typical rotating machinery includes steam turbines, gas turbines, centrifugal and axial compressors, fans, water pumps, water turbines, generators, and aircraft engines. Some large rotating equipment plays a decisive role in the safe and stable operation of systems. Therefore, timely and accurate identification of various abnormal states and faults in these rotating equipment is particularly important. Currently, most widely used fault identification systems for rotating equipment are based on signal analysis methods. These systems deploy vibration sensors, perform time-frequency domain analysis on vibration signals, and then build expert systems based on other equipment information such as temperature, current, and rotational speed to identify the equipment status. This approach is mainly based on an understanding of the equipment's mechanisms and the accumulation of industry knowledge, and has been developed over decades, making it relatively mature. However, it is also limited by theoretical development, with fault identification accuracy remaining at only around 80%, a level that is difficult to improve. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a fault determination method, apparatus, device, and storage medium that can improve the accuracy of fault identification. The specific solution is as follows:

[0004] In a first aspect, this application discloses a fault determination method applied to a comprehensive fault identification system, which includes an expert fault identification system, an AI fault identification system, and a manual fault identification system, comprising:

[0005] The original signal is preprocessed to obtain preprocessed information, and a preset feature extraction operation is performed on the preprocessed information to obtain feature data;

[0006] The feature data is input into the fault determination system for fault detection to obtain the current fault information;

[0007] The current fault information is input into the manual fault evaluation interface so that the target fault corresponding to the original signal can be determined based on the manual evaluation results.

[0008] Optionally, before performing preprocessing operations on the original signal to obtain preprocessed information, the method further includes:

[0009] Process variables for each device are collected through a high-speed acquisition area; wherein, the process variables include the device's vibration information, temperature information, current information, and rotational speed;

[0010] The process variable is identified as the original signal, and the original signal is sent to the server via a network switch.

[0011] Optionally, inputting the feature data into the fault determination system for fault detection to obtain current fault information includes:

[0012] The feature data is input into the fault determination system, and the target device corresponding to the feature data is determined.

[0013] Obtain the target condition set and target rule set corresponding to the target device;

[0014] The current fault information is determined based on the target condition set, the target rule set, and the feature data; wherein, the fault information includes the fault type and confidence level.

[0015] Optionally, determining the current fault information based on the target condition set, the target rule set, and the feature data includes:

[0016] Calculate the confidence level of each condition in the target condition set based on the feature data;

[0017] Determine the current rule set from all the target rule sets, and obtain the current condition set contained in the current rule set;

[0018] Obtain the current condition confidence set corresponding to the current condition set;

[0019] Calculate the fault confidence level corresponding to the current rule set based on the current condition confidence level set and the preset fault confidence level calculation formula.

[0020] The target fault confidence level is determined from all the fault confidence levels based on the preset fault confidence level judgment rules;

[0021] The target fault confidence level and the corresponding target fault type are determined as the current fault information.

[0022] Optionally, inputting the current fault information to the manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results includes:

[0023] The current fault information is input into the manual fault evaluation interface so that the accuracy of the current fault information can be determined according to the preset evaluation rules, so as to obtain the manual evaluation result;

[0024] If the manual evaluation result indicates that the current fault information is accurate, then the target fault is determined based on the current fault information;

[0025] If the manual evaluation result indicates that the current fault information is inaccurate, then manual fault information is obtained based on a preset manual determination operation, and the target fault is determined based on the manual fault information.

[0026] Optionally, before inputting the current fault information to the manual fault evaluation interface, the method further includes:

[0027] The feature data is input into the AI ​​fault identification system so that the corresponding AI output fault information can be calculated by the fault judgment model in the AI ​​fault identification system; wherein, the AI ​​output fault information includes the AI ​​output fault type and the AI ​​output fault credibility.

[0028] Optionally, if the manual evaluation result indicates that the current fault information is accurate, then after determining the target fault based on the current fault information, the method further includes:

[0029] Calculate the difference between the confidence level of the target fault and the confidence level of the fault output by the AI;

[0030] Obtain the absolute value of the difference to get the target confidence difference;

[0031] The target confidence difference is compared with a preset model update deviation threshold.

[0032] If the target confidence difference is greater than the preset model update deviation threshold, the feature data is input into the fault judgment model so as to update the parameters to be updated in the fault judgment model and obtain the updated fault judgment model.

[0033] Secondly, this application discloses a fault determination device applied to a comprehensive fault identification system, which includes an expert fault identification system, an AI fault identification system, and a manual fault identification system, comprising:

[0034] The preprocessing module is used to perform preprocessing operations on the raw signal to obtain preprocessed information;

[0035] The feature extraction module is used to perform a preset feature extraction operation on the preprocessed information to obtain feature data;

[0036] The fault detection module is used to input the feature data into the fault determination system for fault detection in order to obtain the current fault information;

[0037] The fault determination module is used to input the current fault information to the manual fault evaluation interface so as to determine the target fault corresponding to the original signal based on the manual evaluation results.

[0038] Thirdly, this application discloses an electronic device, including:

[0039] Memory, used to store computer programs;

[0040] A processor is configured to execute the computer program to implement the steps of the fault determination method disclosed above.

[0041] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the fault determination method disclosed above.

[0042] As can be seen, this application provides a fault determination method, including: preprocessing the original signal to obtain preprocessed information, and performing a preset feature extraction operation on the preprocessed information to obtain feature data; inputting the feature data into a fault determination system for fault detection to obtain current fault information; and inputting the current fault information into a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results. Therefore, this application uses a fault determination system to perform fault detection, and after obtaining the detection results, inputs the detection results into a manual fault evaluation interface for secondary evaluation. Based on the manual evaluation results, the target fault corresponding to the original signal is determined. Through multiple judgments of fault identification, the accuracy of fault identification is improved. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0044] Figure 1 This is a flowchart of a fault determination method disclosed in this application;

[0045] Figure 2 This is a schematic diagram of the integrated fault identification system disclosed in this application;

[0046] Figure 3 This is a schematic diagram of a fault identification system architecture disclosed in this application;

[0047] Figure 4 This is a schematic diagram of an expert fault identification system architecture disclosed in this application;

[0048] Figure 5 This is a schematic diagram of the operation process of an expert fault identification system disclosed in this application;

[0049] Figure 6 This is a flowchart of a specific fault determination method disclosed in this application;

[0050] Figure 7 This is a schematic diagram of an ensemble learning model disclosed in this application;

[0051] Figure 8 A schematic diagram of the fault determination device provided in this application;

[0052] Figure 9 This application provides a structural diagram of an electronic device. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Currently, most widely used fault identification systems for rotating equipment are based on signal analysis methods. These systems deploy vibration sensors to perform time-frequency domain analysis of vibration signals, and then establish an expert fault identification system based on other equipment information such as temperature, current, and rotational speed to identify the equipment's condition. This approach is primarily based on an understanding of the equipment's mechanisms and the accumulation of industry knowledge, and has been developed over decades, making it relatively mature. However, it is also limited by theoretical advancements, with fault identification accuracy remaining at only around 80%, a level that is difficult to improve. Therefore, this application provides a fault determination method that can improve the accuracy of fault identification.

[0055] This invention discloses a fault determination method, see [link to relevant documentation]. Figure 1 As shown, this method is applied to a comprehensive fault identification system, which includes an expert fault identification system, an AI fault identification system, and a manual fault identification system. The method includes:

[0056] Step S11: Perform preprocessing operations on the original signal to obtain preprocessed information, and perform preset feature extraction operations on the preprocessed information to obtain feature data.

[0057] In this embodiment, the original signal is preprocessed to obtain preprocessed information, and a preset feature extraction operation is performed on the preprocessed information to obtain feature data. It is understood that before preprocessing the original signal to obtain preprocessed information, process variables of each device are collected through a high-speed acquisition area; wherein, the process variables include the device's vibration information, temperature information, current information, and rotational speed; the process variables are identified as the original signal, and the original signal is sent to the server through a network switch.

[0058] Develop a comprehensive fault identification system for rotating equipment, such as Figure 2 As shown, this system protects a fault identification server, network switches, and rotating equipment. A high-speed acquisition module collects process variables such as vibration, temperature, current, and rotational speed from various rotating equipment and transmits them to the fault identification server via the network switch. The fault identification server simultaneously runs an expert fault identification system and an AI fault identification system, and also provides a manual fault identification entry point. The expert fault identification system and the AI ​​fault identification system establish fault identification models offline based on the equipment's operating mechanism and historical operating data. During online operation, multiple fault identification systems run simultaneously and assist each other. The online fault identification system architecture is as follows: Figure 3 As shown, preprocessing involves basic processing of the raw data, such as filtering, outlier removal, standardization, and time-frequency domain transformation. Feature extraction involves extracting time-frequency domain features for fault identification from the field data. Taking a steam turbine as an example, the features mainly include process parameters such as speed, load, thermal expansion, bearing temperature, lubricating oil temperature and pressure, excitation current, and unit efficiency, as well as vibration-related parameters such as amplitude, spectrum, and phase. Combining this data can identify approximately 80% of mechanical faults. These features are then input into both the expert fault identification system and the AI ​​fault identification system, meaning both systems receive the same feature data.

[0059] Step S12: Input the feature data into the fault determination system for fault detection to obtain the current fault information.

[0060] In this embodiment, the original signal is preprocessed to obtain preprocessed information, and a preset feature extraction operation is performed on the preprocessed information to obtain feature data. The feature data is then input into a fault determination system for fault detection to obtain current fault information. Specifically, the feature data is input into the fault determination system, and the target device corresponding to the feature data is determined; the target condition set and target rule set corresponding to the target device are obtained; the current fault information is determined based on the target condition set, the target rule set, and the feature data; wherein the fault information includes the fault type and confidence level.

[0061] It is understood that determining the current fault information based on the target condition set, the target rule set, and the feature data includes: calculating the condition confidence level corresponding to each condition in the target condition set based on the feature data; determining the current rule set from all the target rule sets and obtaining the current condition set contained in the current rule set; obtaining the current condition confidence level set corresponding to the current condition set; calculating the fault confidence level corresponding to the current rule set based on the current condition confidence level set and a preset fault confidence level calculation formula; determining the target fault confidence level from all the fault confidence levels based on a preset fault confidence level judgment rule; and determining the target fault confidence level and the corresponding target fault type as the current fault information.

[0062] Understandably, the expert fault identification system determines the system's operating status based on input features and built-in judgment rules, and when a fault occurs, it provides the fault type and its confidence level (percentage value). The architecture of the expert fault identification system is as follows: Figure 4 As shown, the fault identification expert database contains n types of fault-identifiable devices. Each type of device contains two sets: a fault identification condition set (i.e., a diagnostic condition set) and a fault identification rule set (i.e., a diagnostic rule set). Each fault identification rule corresponds to a type of fault. The fault identification rule set is a collection of fault identification rules that summarizes the fault types whose occurrence conditions are relatively clear. An example of a fault identification rule set is shown below:

[0063] R1: If the amplitude of the first harmonic in the vibration spectrum is large, and the amplitude remains basically unchanged when the rotation speed is constant, and the phase of the first harmonic remains basically unchanged when the rotation speed is constant, then the mass is unbalanced.

[0064] R2: If the rotational speed is close to the rated speed and the rotor vibration is large, then the shaft system is misaligned;

[0065] R3: If the first harmonic in the vibration spectrum is large, and the amplitude gradually changes when the rotation speed is constant, and the phase of the first harmonic gradually changes when the rotation speed is constant, then there is dynamic-static contact.

[0066] R4: If the first harmonic in the vibration spectrum is large, and the rotational speed changes little while the vibration increases rapidly, then resonance occurs.

[0067] The fault identification rule set is a collection of rules in the form of the following. The fault identification condition set is a set of these conditions. Because some rules have the same judgment conditions, all conditions are summarized into a set to facilitate rule calculation. If condition 1, And condition 2, And condition 3, ..., Then fault. Each condition is measured by a condition confidence level, which characterizes the probability of the condition occurring. Specifically, the confidence level is calculated based on the input features. For example, for the condition "the amplitude of the first harmonic in the vibration spectrum is relatively large", let the amplitude of the first harmonic be f. 1aLet the upper and lower limits for determining "large amplitude" be [f]. 1amin f 1amax The conditional confidence level is calculated as follows:

[0068] If f 1a <= f 1amin Then the conditional confidence level CT = 0;

[0069] If f 1a >=f 1amax Then the conditional confidence level CT = 1;

[0070] If f 1amin <f 1a <f 1amax Then the conditional credibility

[0071] As can be seen, conditional confidence is a percentage value between [0, 1]. Failure confidence represents the degree to which a failure is likely to occur. Failure confidence is the AND of the confidence levels of each condition under which the failure is likely to occur, i.e., Failure confidence = (Condition 1 confidence) AND (Condition 2 confidence) AND (Condition 3 confidence) AND ...

[0072] Taking rule R1 above as an example, if the confidence levels of the three conditions are 0.9, 0.8, and 0.75 respectively, then the confidence level of the quality imbalance fault is: CT = 0.9 ∩ 0.8 ∩ 0.75 = 0.75. The fault identification process of the overall expert fault identification system is as follows: Figure 5 As shown, after collecting feature data, the corresponding device is determined based on the feature data, according to, for example, the above. Figure 4 The correspondence between the device and the fault identification condition set and fault identification rule set shown determines the set to be used. Based on the determined set, the confidence level of each rule (fault) is calculated, and the confidence level of the fault with the highest probability is output.

[0073] The final fault identification result of the expert fault identification system is an OR operation of the confidence levels of each fault, that is, it outputs the confidence level of the fault with the highest probability. The specific formula is as follows:

[0074] CT = CT1∪CT2∪……∪CT n .

[0075] The system status is shown in Table 1 below for different CT values:

[0076] Table 1

[0077] NO Fault reliability Severity of the fault Equipment status description 1 CT<0.5 slight Long-term operation 2 0.5<=CT<0.65 Lighter There are potential risks, and monitoring needs to be strengthened. 3 0.65<=CT<0.8 heavier There is a current wind direction; shutdown and inspection may be necessary. 4 0.8<=CT serious There is a serious risk; please stop the machine and inspect it immediately.

[0078] The severity of the current fault and the current state of the equipment are determined based on the range of the calculated fault confidence level.

[0079] Step S13: Input the current fault information into the manual fault evaluation interface so as to determine the target fault corresponding to the original signal based on the manual evaluation results.

[0080] In this embodiment, the feature data is input to the fault determination system for fault detection to obtain current fault information. Then, the current fault information is input to a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results. Specifically, the current fault information is input to the manual fault evaluation interface to determine its accuracy according to preset evaluation rules, thus obtaining a manual evaluation result. If the manual evaluation result indicates that the current fault information is accurate, the target fault is determined based on the current fault information. If the manual evaluation result indicates that the current fault information is inaccurate, manual fault information is obtained based on a preset manual determination operation, and the target fault is determined based on this manual fault information.

[0081] As can be seen, this application provides a fault determination method, including: preprocessing the original signal to obtain preprocessed information, and performing a preset feature extraction operation on the preprocessed information to obtain feature data; inputting the feature data into a fault determination system for fault detection to obtain current fault information; and inputting the current fault information into a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results. Therefore, this application uses a fault determination system to perform fault detection, and after obtaining the detection results, inputs the detection results into a manual fault evaluation interface for secondary evaluation. Based on the manual evaluation results, the target fault corresponding to the original signal is determined. Through multiple judgments of fault identification, the accuracy of fault identification is improved.

[0082] See Figure 6 As shown, this embodiment of the invention discloses a fault determination method. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution.

[0083] Step S21: Perform preprocessing operations on the original signal to obtain preprocessed information, and perform preset feature extraction operations on the preprocessed information to obtain feature data.

[0084] Step S22: Input the feature data into the fault determination system for fault detection to obtain the current fault information.

[0085] Step S23: Input the current fault information into the manual fault evaluation interface so as to determine the target fault corresponding to the original signal based on the manual evaluation results.

[0086] In this embodiment, the current fault information is input to the manual fault evaluation interface so that the target fault corresponding to the original signal can be determined based on the manual evaluation results. Then, the feature data is input to the AI ​​fault recognition system so that the corresponding AI output fault information can be calculated by the fault judgment model in the AI ​​fault recognition system. The AI ​​output fault information includes the AI ​​output fault type and the AI ​​output fault credibility.

[0087] In recent years, with the rapid development of AI, the theory and application of machine learning and deep learning have made significant progress. AI algorithms mainly rely on training data and model structure, with relatively low dependence on underlying mechanisms. When the type, quantity, and quality of modeling data are all satisfactory, excellent training results can be achieved, and much research has been conducted in the field of equipment fault identification. However, in practical applications, there are few successful cases and large-scale applications. The main reason is the lack of fault label data, i.e., a lack of sufficient modeling data to train AI fault identification models, which greatly limits the application of AI in the field of fault identification.

[0088] Given the importance of large rotating equipment, users typically desire multiple fault identification methods for reference and accept the progressively improving accuracy of AI over time. Therefore, this invention proposes a fault identification model that combines an expert fault identification system with AI. The expert fault identification system and industry experts provide samples and labels for machine learning, continuously improving its fault identification accuracy over time. The machine learning fault identification results provide valuable references for traditional methods from a data analysis perspective. The two methods complement each other, thereby enhancing the overall effectiveness of the fault identification system.

[0089] Understandably, as mentioned above Figure 3 As shown, after feature extraction, the feature data is input into the AI ​​fault identification system, the expert fault identification system, and the manual fault identification system, respectively. At this point, the expert fault identification system calculates the fault credibility and outputs the target fault credibility. Simultaneously, the AI ​​fault identification system calculates the fault credibility based on the feature data to obtain the AI-output fault information. The manual fault identification system determines whether the target fault credibility is credible; if credible, it proceeds to the step of calculating the difference between the target fault credibility and the AI-output fault credibility. It should be noted that the manual fault identification system stores the feature data from each judgment and the current fault information output by the expert fault identification system as new fault samples in the fault database, so as to train the AI ​​fault identification system using the fault database.

[0090] Step S24: Calculate the difference between the target fault confidence level and the AI ​​output fault confidence level, and obtain the absolute value of the difference to obtain the target confidence level difference.

[0091] In this embodiment, after determining the target fault corresponding to the original signal based on the manual evaluation results, the difference between the confidence level of the target fault and the confidence level of the AI ​​output fault is calculated, and the absolute value of the difference is obtained to obtain the target confidence level difference.

[0092] Understandably, the manual fault identification entry point is an input interface reserved for industry experts, equipment maintenance personnel, etc., and mainly has two functions. First, it assesses the reliability of the fault output by the expert fault identification system, which can be evaluated by equipment maintenance personnel or industry experts. If the reliability is confirmed to be reliable and accurate, it can be used for automatic parameter updates of the AI ​​model. Second, it calculates the deviation between the fault reliability of the expert fault identification system and the fault reliability of the AI ​​fault identification system. If the following conditions are met, the parameters of the AI ​​model are updated.

[0093] |cT 专家系统 -CT AI |>α;Form 1;

[0094] cT 专家系统 To improve the reliability of the expert fault identification system, CT AI The reliability of the AI ​​system's fault assessment is represented by α, which is the model update bias threshold. If the fault identification results of the human expert fault identification system are unreliable, they will not be used.

[0095] While current expert fault identification systems can identify major fault types such as mass imbalance, rotor thermal bending, coupling misalignment, resonance, and dynamic-static rubbing, some faults remain difficult to automatically identify due to the complexity of some equipment mechanisms, diverse causes of failure, and difficulties in signal acquisition. These faults still require manual identification by industry experts based on various analytical data, such as oil film oscillation and fluid excitation. When such faults occur, industry experts must first assess and label them. The data is then entered into the fault database through the manual fault identification entry point, serving as sample data for subsequent periodic updates to the AI ​​model.

[0096] Step S25: Compare the target confidence difference with the preset model update deviation threshold. If the target confidence difference is greater than the preset model update deviation threshold, input the feature data into the fault judgment model to update the parameters to be updated in the fault judgment model, and obtain the updated fault judgment model.

[0097] In this embodiment, the difference between the target fault confidence level and the AI ​​output fault confidence level is calculated, and the absolute value of the difference is obtained. After obtaining the target confidence level difference, the target confidence level difference is compared with a preset model update deviation threshold. If the target confidence level difference is greater than the preset model update deviation threshold, the feature data is input into the fault judgment model so as to update the parameters to be updated in the fault judgment model and obtain the updated fault judgment model.

[0098] Understandably, the AI ​​fault identification system identifies faults based on input features and an offline-trained fault identification model, determines the system status, and provides the fault type and its confidence level (percentage data) when a fault occurs. The input features of the AI ​​fault identification system are the same as those of the expert fault identification system. Based on the fault samples accumulated by the expert fault identification system, AI modeling data is collected as shown in Table 2 below. Various AI models can be constructed based on this sample data.

[0099] Table 2

[0100] N0 Feature 1 Feature 2 …… Feature m Fault reliability 1 A1 B1 …… C1 CT1 2 A2 B2 …… C2 CT2 …… …… …… …… …… …… n An Bn …… Cn CTn

[0101] When building AI models, the lack of fault sample data, especially for expensive, large rotating equipment such as gas turbines, requires years of data accumulation. Deep learning, which demands a large amount of sample data, is generally unsuitable in such cases. Fault identification models often employ machine learning models. This solution uses an ensemble learning model for fault identification, constructing an ensemble learning model by weighting two typical weak learners: MLR (Multiple Linear Regression) and RBF (Radial Basis Function Network). The model structure is as follows: Figure 7 As shown, t1, t2......t n Let be the fault characteristic variables. The MLR model output is y1, and the RBF model output is y2. After weighting, the integrated output y is obtained, with weights w1 and w2 respectively. b is the output bias of the integrated model, so the integrated model output is:

[0102] y = w1y1 + w2y2 + b; Equation 2;

[0103] This model is used for AI fault identification. The coefficients of MLR, the parameters of RBF network, the weight variables, and the bias can all be obtained by training on an offline dataset. When performing fault identification online, the model is updated when the condition in Equation 1 is met. The updated parameters are the weights w1 and w2 and the bias b in Equation 2.

[0104] Model parameters are updated automatically.

[0105] Treating the parameters w1, w2, and b to be updated as state variables, transformation equation 2 is as follows:

[0106]

[0107] Where z = [y1 y2 1], β = [w1 w2 b] T ;

[0108] The state equations for constructing the fault identification model are as follows:

[0109] β k =β k-1 +q k-1 Formula 4;

[0110]

[0111] Based on output equation 3, the state equation is supplemented in equation 4. β represents the state variable, which is assumed to be an unbiased estimate; q represents the state bias, assumed to be a zero-mean white noise sequence; and e represents the measurement bias. The Kalman update can then transform it into:

[0112] β k|k-1 =β k-1|k-1 Formula 6;

[0113] P k|k-1 =P k-1|k-1 +Q; Equation 7;

[0114]

[0115]

[0116]

[0117] Where P is the covariance matrix of state β, and Q is the state noise q. k-1 The covariance matrix is ​​R, where R is the covariance matrix of the measurement noise. y represents the fault confidence level of the online expert fault identification system. When this value deviates from the AI's fault confidence level by more than a set threshold, a Kalman update is initiated, updating the model coefficients β according to equations 6 to 10. In equation 9, β... k|k This refers to the updated model parameters.

[0118] Understandably, this invention organically integrates expert fault identification systems with AI fault identification systems and manual fault identification. Due to the lack of sample data, the fault identification accuracy of existing AI systems cannot meet the requirements. Over time and with the accumulation of faults, expert and manual fault identification systems continuously provide sample data to the AI ​​fault identification system, and the fault identification accuracy of the AI ​​model will continuously improve. When the accumulated samples are large enough, the fault identification accuracy of the AI ​​model will surpass that of the expert fault identification system, thereby improving the practicality of AI fault identification. Its fault identification model can also be deployed in similar equipment. The AI ​​model parameters are automatically updated based on the fault credibility. That is, after the fault identification results of the expert fault identification system are manually confirmed, the weights of the AI ​​ensemble model can be automatically updated using the Kalman filter method, reducing the maintenance cost of the model during operation. For fault types that the expert fault identification system cannot cover, industry experts can judge and enter them into the fault database. After accumulating a certain amount of data, the AI ​​model can be retrained to further improve the fault identification accuracy of the AI ​​model.

[0119] Furthermore, there are various alternatives to ensemble learning models, such as SVM, decision trees, and random forests, which can all be used as components in the fault identification model ensemble. The choice of model depends on the specific data. However, regardless of the ensemble learning model used, as long as its output is a linearly weighted sum of multiple models, the Kalman model update scheme proposed in this paper can be adopted.

[0120] For details regarding steps S21 and S22, please refer to the corresponding content disclosed in the foregoing embodiments, which will not be repeated here.

[0121] As can be seen, this embodiment of the application obtains preprocessed information by preprocessing the original signal, and performs a preset feature extraction operation on the preprocessed information to obtain feature data; the feature data is input to a fault determination system for fault detection to obtain current fault information; the current fault information is input to a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results; the difference between the confidence level of the target fault and the confidence level of the fault output by the AI ​​is calculated; the absolute value of the difference is obtained to obtain the target confidence level difference; the target confidence level difference is compared with a preset model update deviation threshold; if the target confidence level difference is greater than the preset model update deviation threshold, the feature data is input to the fault judgment model to update the parameters to be updated in the fault judgment model, thereby obtaining an updated fault judgment model and improving the accuracy of fault identification.

[0122] See Figure 8As shown in the illustration, this application also discloses a fault determination device applied to a comprehensive fault identification system. The comprehensive fault identification system includes an expert fault identification system, an AI fault identification system, and a manual fault identification system, comprising:

[0123] Preprocessing module 11 is used to perform preprocessing operations on the original signal to obtain preprocessed information;

[0124] The feature extraction module 12 is used to perform a preset feature extraction operation on the preprocessed information to obtain feature data;

[0125] The fault detection module 13 is used to input the feature data into the fault determination system for fault detection in order to obtain the current fault information;

[0126] The fault determination module 14 is used to input the current fault information to the manual fault evaluation interface so as to determine the target fault corresponding to the original signal based on the manual evaluation results.

[0127] As can be seen, this application includes: performing preprocessing operations on the original signal to obtain preprocessed information, and performing a preset feature extraction operation on the preprocessed information to obtain feature data; inputting the feature data into a fault determination system for fault detection to obtain current fault information; and inputting the current fault information into a manual fault evaluation interface to determine the target fault corresponding to the original signal based on the manual evaluation results. Therefore, this application uses a fault determination system to perform fault detection, obtains the detection results, inputs the results into a manual fault evaluation interface for secondary evaluation, and determines the target fault corresponding to the original signal based on the manual evaluation results. By performing multiple judgments on fault identification, the accuracy of fault identification is improved.

[0128] In some specific embodiments, the preprocessing module 11 specifically includes:

[0129] The process variable acquisition unit is used to acquire process variables of each device through a high-speed acquisition area; wherein, the process variables include vibration information, temperature information, current information, and rotational speed of the device;

[0130] The original signal transmission unit is used to determine the process variable as the original signal and transmit the original signal to the server through a network switch;

[0131] The unit is used to perform preprocessing operations on the original signal to obtain preprocessed information.

[0132] In some specific embodiments, the feature extraction module 12 specifically includes:

[0133] The feature extraction unit is used to perform a preset feature extraction operation on the preprocessed information to obtain feature data.

[0134] In some specific embodiments, the fault detection module 13 specifically includes:

[0135] The target device determination unit is used to input the feature data into the fault determination system and determine the target device corresponding to the feature data;

[0136] The set acquisition unit is used to acquire the target condition set and target rule set corresponding to the target device;

[0137] A condition confidence calculation unit is used to calculate the condition confidence of each condition in the target condition set based on the feature data.

[0138] The current condition set acquisition unit is used to determine the current rule set from all the target rule sets and acquire the current condition set contained in the current rule set;

[0139] The current condition confidence set acquisition unit is used to acquire the current condition confidence set corresponding to the current condition set;

[0140] The fault confidence calculation unit is used to calculate the fault confidence corresponding to the current rule set based on the current condition confidence set and the preset fault confidence calculation formula.

[0141] The target fault credibility determination unit is used to determine the target fault credibility from all the fault credibility based on a preset fault credibility judgment rule;

[0142] The current fault information determination unit is used to determine the target fault confidence level and the corresponding target fault type as the current fault information; wherein, the fault information includes the fault type and confidence level.

[0143] In some specific embodiments, the fault determination module 14 specifically includes:

[0144] The AI ​​output fault information calculation unit is used to input the feature data into the AI ​​fault recognition system so that the corresponding AI output fault information can be calculated by the fault judgment model in the AI ​​fault recognition system; wherein, the AI ​​output fault information includes the AI ​​output fault type and the AI ​​output fault credibility.

[0145] The manual evaluation result acquisition unit is used to input the current fault information into the manual fault evaluation interface so as to determine whether the current fault information is accurate according to the preset evaluation rules, so as to obtain the manual evaluation result;

[0146] The first target fault determination unit is used to determine the target fault based on the current fault information if the manual evaluation result indicates that the current fault information is accurate.

[0147] The manual fault information acquisition unit is used to acquire manual fault information based on a preset manual determination operation if the manual evaluation result indicates that the current fault information is inaccurate.

[0148] The second target fault determination unit is used to determine the target fault based on the artificial fault information;

[0149] The difference calculation unit is used to calculate the difference between the confidence level of the target fault and the confidence level of the fault output by the AI.

[0150] A target credibility difference determination unit is used to obtain the absolute value of the difference in order to obtain the target credibility difference.

[0151] A numerical comparison unit is used to compare the target confidence difference with a preset model update deviation threshold.

[0152] The model update unit is used to input the feature data into the fault judgment model if the target confidence difference is greater than the preset model update deviation threshold, so as to update the parameters to be updated of the fault judgment model and obtain the updated fault judgment model.

[0153] Furthermore, embodiments of this application also provide an electronic device. Figure 9 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0154] Figure 9 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the fault determination method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0155] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0156] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0157] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the fault determination method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program capable of performing other specific tasks.

[0158] Furthermore, embodiments of this application also disclose a storage medium storing a computer program, which, when loaded and executed by a processor, implements the fault determination method steps disclosed in any of the foregoing embodiments.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0160] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0161] The above provides a detailed description of the fault determination method, apparatus, device, and storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fault determination method, characterized in that, Applied to a comprehensive fault identification system, which includes an expert fault identification system, an AI fault identification system, and a manual fault identification system, including: The original signal is preprocessed to obtain preprocessed information, and a preset feature extraction operation is performed on the preprocessed information to obtain feature data; The feature data is input into the expert fault identification system for fault detection to obtain current fault information; The feature data is input into the AI ​​fault identification system so that the corresponding AI output fault information can be calculated by the fault judgment model in the AI ​​fault identification system; wherein, the AI ​​output fault information includes the AI ​​output fault type and the AI ​​output fault confidence level; The current fault information is input to the manual fault evaluation interface so that the target fault corresponding to the original signal can be determined based on the manual evaluation results, including: The current fault information is input into the manual fault evaluation interface so that the accuracy of the current fault information can be determined according to the preset evaluation rules, so as to obtain the manual evaluation result. If the manual evaluation result indicates that the current fault information is accurate, then the target fault is determined based on the current fault information; If the manual evaluation result indicates that the current fault information is inaccurate, then manual fault information is obtained based on a preset manual determination operation, and the target fault is determined based on the manual fault information. Calculate the difference between the confidence level of the target fault corresponding to the target fault and the confidence level of the fault output by the AI; Obtain the absolute value of the difference to get the target confidence difference; The target confidence difference is compared with a preset model update deviation threshold. If the target confidence difference is greater than the preset model update deviation threshold, the feature data is input into the fault judgment model so that the parameters to be updated of the fault judgment model are updated according to the Kalman filter method to obtain the updated fault judgment model; wherein, the parameters to be updated include the output weights of the multiple linear regression and the output weights and bias b of the radial basis function network; The expert fault identification system determines the system's operating status based on input features and built-in judgment rules, and provides the fault type and target fault confidence level when a fault occurs. The AI ​​fault identification system is an ensemble learning model constructed based on weighted multiple linear regression and radial basis function networks. The output of the multiple linear regression is... The output of the radial basis function network is The weighted average output is then obtained. The weights are respectively , Let b be the bias of the ensemble model output, then the output of the ensemble learning model is: .

2. The fault determination method according to claim 1, characterized in that, Before performing preprocessing operations on the original signal to obtain preprocessed information, the method further includes: Process variables for each device are collected through a high-speed acquisition area; wherein, the process variables include the device's vibration information, temperature information, current information, and rotational speed; The process variable is identified as the original signal, and the original signal is sent to the server via a network switch.

3. The fault determination method according to claim 1, characterized in that, The step of inputting the feature data into the expert fault identification system for fault detection to obtain current fault information includes: The feature data is input into the expert fault identification system to determine the target device corresponding to the feature data; Obtain the target condition set and target rule set corresponding to the target device; The current fault information is determined based on the target condition set, the target rule set, and the feature data; wherein, the fault information includes the fault type and confidence level.

4. The fault determination method according to claim 3, characterized in that, Determining the current fault information based on the target condition set, the target rule set, and the feature data includes: Calculate the confidence level of each condition in the target condition set based on the feature data; Determine the current rule set from all the target rule sets, and obtain the current condition set contained in the current rule set; Obtain the current condition confidence set corresponding to the current condition set; Calculate the fault confidence level corresponding to the current rule set based on the current condition confidence level set and the preset fault confidence level calculation formula. The target fault confidence level is determined from all the fault confidence levels based on the preset fault confidence level judgment rules; The target fault confidence level and the corresponding target fault type are determined as the current fault information.

5. A fault determination device, characterized in that, Applied to a comprehensive fault identification system, which includes an expert fault identification system, an AI fault identification system, and a manual fault identification system, the fault determination method based on any one of claims 1 to 4 includes: The preprocessing module is used to perform preprocessing operations on the raw signal to obtain preprocessed information; The feature extraction module is used to perform a preset feature extraction operation on the preprocessed information to obtain feature data; The fault detection module is used to input the feature data into the expert fault identification system for fault detection in order to obtain the current fault information; The fault determination module is used to input the current fault information to the manual fault evaluation interface so as to determine the target fault corresponding to the original signal based on the manual evaluation results.

6. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the fault determination method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, Used to store a computer program; wherein, when the computer program is executed by a processor, it implements the fault determination method as described in any one of claims 1 to 4.

Images