Reliability prediction and active fault warning method for complex equipment moving parts under variable working conditions

Abstract

The present invention discloses a method for predicting the reliability of complex equipment moving parts under variable working conditions and actively warning of faults, which belongs to the field of reliability prediction and fault warning of complex equipment. The method adopts fractional moment method and optimization correction method to obtain variable working condition reliability data of complex equipment moving parts; collects data such as the internal temperature, drive motor voltage and current, and load of complex equipment moving parts under different working conditions, constructs training and test data sets, and establishes a long short-term memory neural network model for predicting the reliability of complex equipment moving parts under different working condition parameters; monitors the operating condition data of complex equipment in real time and inputs it into the network to realize the reliability prediction of complex equipment moving parts, and issues an alarm when the predicted reliability of the moving parts is lower than the specified requirements, so as to realize active warning of faults. The present invention explores the reliability evolution mechanism of complex equipment moving parts under variable working conditions, realizes accurate reliability prediction and active warning of faults, and helps to ensure the safe and reliable operation of complex equipment.

Description

Technical Field

[0001] The present invention relates to the field of reliability prediction and fault early warning of complex equipment, and in particular to a method for reliability prediction and active fault early warning of moving parts of complex equipment under variable working conditions. Background Art

[0002] The reliability of moving parts is the key to ensure long-term trouble-free operation of complex equipment. Changes in operating parameters such as drive motor voltage and current, load, etc. can lead to performance failure, unreliable operation and other fault problems of moving parts of complex equipment. In order to ensure the reliability of the performance of moving parts of complex equipment and the safe and trouble-free operation of the whole equipment, a method for predicting the reliability of moving parts of complex equipment under variable operating conditions and active fault warning is urgently needed, so as to deeply study the evolution law of the reliability of moving parts of complex equipment under different operating conditions and predict possible faults in real time. Summary of the invention

[0003] In order to solve the shortcomings of the prior art and achieve the purpose of variable working condition reliability prediction and active fault warning of complex equipment moving parts, the present invention adopts the following technical solutions:

[0004] The variable working condition reliability prediction method of complex equipment moving parts includes the following steps:

[0005] Step 1: Collect performance data of moving parts of complex equipment under different working conditions to calculate fractional statistical moments of moving parts performance, and establish a probability model of moving parts performance function with unknown parameters, calculate analytical form fractional moments of moving parts performance according to the probability model, establish a correction criterion between the fractional statistical moments of the moving parts performance and the analytical form fractional moments, and use an optimization correction method to solve the optimal unknown parameters of the probability model to calculate the reliability of moving parts of complex equipment, and finally obtain the reliability data of moving parts at different times;

[0006] Step 2: construct a neural network model between different operating parameters and complex equipment reliability data, combine the operating parameter data and reliability data, establish a training data set and a test data set for training the neural network model;

[0007] Step 3: Based on the monitoring data, the reliability of moving parts of complex equipment is predicted in real time through the trained neural network model.

[0008] Furthermore, the step 1 comprises the following steps:

[0009] Step 1.1: Establish a reliability characterization model for complex equipment moving parts based on the performance requirements of complex equipment moving parts;

[0010] Step 1.2: Collect performance data of moving parts under different working conditions to calculate the fractional statistical moment of the performance of the moving parts;

[0011] Step 1.3: Use the logarithmic skew normal distribution to approximate the probability model of the performance function of the moving part, and obtain the analytical form fractional moment of the performance function of the moving part from the probability model;

[0012] Step 1.4: Establish a correction criterion between the performance fractional moments of the moving parts obtained from experimental data and the fractional moments in analytical form, and use the optimization model to solve the optimal unknown parameters of the probability model;

[0013] Step 1.5: Obtain a corrected probability model of the performance function of the moving component according to the optimal unknown parameters, and calculate the reliability of the moving component through the reliability characterization model of the moving component of the complex equipment based on the corrected probability model;

[0014] Step 1.6: Obtain the performance data of moving parts collected at different times, and based on the above steps, obtain the reliability data of moving parts of complex equipment.

[0015] Furthermore, the reliability characterization model of complex equipment moving parts established in step 1.1 is as follows:

[0016]

[0017] Among them, X represents the uncertainty parameters such as the internal temperature and load of the moving parts, represents the performance function of the moving parts, Indicates the performance threshold of the moving parts, represents the probability density function of the performance function of the moving part, Pr{·} represents the probability operation, and z represents the performance parameter of the moving part.

[0018] Furthermore, in step 1.5, the corrected moving component performance function probability model , reliability of moving parts:

[0019]

[0020] in, represents the optimal value of the unknown parameter.

[0021] Furthermore, the fractional statistical moment of the performance of the moving parts in step 1.2 is as follows:

[0022]

[0023] in, It represents the fractional moment of the performance parameters of the moving parts, r represents the order of the fractional moment, The index value representing the order, Indicates the total number of preset orders, G iIt represents the performance data of the moving parts collected under the i-th working condition, and n represents the number of working conditions.

[0024] Furthermore, the logarithmic skew normal distribution in step 1.3 approximately characterizes the probability model of the performance function of the moving component, and the formula is as follows:

[0025]

[0026] Among them, r represents the order of the fractional moment, Represents positional parameters, represents the scale parameter, represents the shape parameter, R represents the real number domain, and They represent the probability density function and cumulative distribution function of the standard normal distribution, represents the unknown parameter vector;

[0027] The analytical form of the performance function of the moving part is the fractional moment, and the formula is as follows:

[0028]

[0029] in, represents the analytical form fractional moment corresponding to the probability model based on the performance parameter z of the moving part, Represents an exponential function.

[0030] Furthermore, the correction criterion Q in step 1.4 is as follows:

[0031]

[0032] in, Represents vectors of different orders, The index value representing the order, Indicates the total number of presets. Represents unknown parameters, represents the vector composed of different order fractional statistical moments, z represents the performance parameters of the moving parts, represents the analytical form of fractional moment, Represents a vector composed of fractional moments of different orders of analytical forms;

[0033] The formula of the optimization model is as follows:

[0034]

[0035] in, represents the optimal value of the unknown parameter, Indicates the unknown parameter corresponding to the minimum value operate.

[0036] Furthermore, the neural network model in step 2 is a long short-term memory neural network model, including an input layer, a long short-term memory unit as a hidden layer, a fully connected layer, and an output layer connected in sequence; whether the training is completed is judged by whether the loss function meets the requirements, if so, the training is completed, and a trained neural network model of a moving part of complex equipment is obtained, if not, the neural network model is continued to be trained, and the loss function of the neural network model training is as follows:

[0037]

[0038] in, represents the reliability data predicted by the neural network, represents the i-th reliability data in the test data set, It represents the Euclidean distance between two points, and m represents the data length.

[0039] Furthermore, the operating parameters include the internal temperature of the moving parts, the driving motor voltage and current, and the load, and the complex equipment reliability data includes the reliability of the moving parts at different times.

[0040] The active early warning method for complex equipment moving parts failure is based on the variable working condition reliability prediction method of complex equipment moving parts, and the reliability of complex equipment moving parts is predicted in real time. If the predicted reliability does not meet the specified requirements, an alarm is issued to conduct active early warning of the failure.

[0041] The advantages and beneficial effects of the present invention are:

[0042] (1) The present invention quantifies the uncertainty of the performance of the moving parts of complex equipment under variable working conditions through fractional moments, thereby fully capturing the statistical information of the uncertainty of the performance of the moving parts of complex equipment during operation;

[0043] (2) The present invention uses a parameterized probability model and an optimization correction method to achieve accurate evaluation of the reliability of moving parts of complex equipment under variable working conditions;

[0044] (3) The present invention integrates the real-time monitoring data and reliability data of moving parts during the operation of complex equipment to establish a neural network model for predicting the reliability of moving parts of complex equipment under different operating parameters. Combined with the real-time monitoring data of moving parts of complex equipment, the present invention realizes the reliability prediction and active fault warning of moving parts of complex equipment under variable operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flow chart of a method for predicting reliability of variable working conditions and actively warning of faults of moving parts of complex equipment in an embodiment of the present invention.

[0046] Figure 2It is an implementation diagram of reliability prediction and fault warning of moving parts of complex equipment in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0048] Reliability prediction and active fault warning methods for complex equipment moving parts under variable working conditions, such as Figure 1 As shown, the following steps are included:

[0049] Step 1: By collecting the performance data of the moving parts of complex equipment under different working conditions, calculating the fractional statistical moments of the performance function of the operating parts, establishing a probabilistic model of the performance function of the moving parts with unknown parameters and obtaining the analytical model of its fractional moments, establishing a correction criterion between the performance fractional moments of the moving parts obtained from the experimental data and the analytical fractional moments, using the optimization correction method to solve the unknown parameters of the probabilistic model, calculating the reliability of the moving parts of the complex equipment, and obtaining the reliability data of the moving parts at different times.

[0050] Specifically, the fractional order moment method and the optimization correction method are used to obtain the variable working condition reliability data of complex equipment, including the following steps:

[0051] Step 1.1: Establish the reliability characterization model P of complex equipment moving parts r , specifically:

[0052] (1)

[0053] Among them, X represents the uncertainty parameters such as the internal temperature and load of the moving parts, represents the high-speed crankshaft deformation function, Indicates the maximum deformation threshold of the high-speed crankshaft, represents the probability density function of the high-speed crankshaft deformation function, Pr{·} represents the probability operation, and z represents the high-speed crankshaft deformation.

[0054] Step 1.2: Collect high-speed crankshaft deformation data under different working conditions , n=500 is the data volume, used to represent the number of working conditions;

[0055] Calculate the fractional statistical moment information of high-speed crankshaft deformation:

[0056] (2)

[0057] in, represents the fractional moment of the high-speed crankshaft deformation z, r represents the order of the fractional moment, The index value representing the order, Indicates the total number of preset orders, G i represents the high-speed crankshaft deformation data collected under the i-th working condition;

[0058] Step 1.3: Use logarithmic skew normal distribution to approximate the probability model of high-speed crankshaft deformation function:

[0059] (3)

[0060] in, Represents positional parameters, represents the scale parameter, represents the shape parameter, R represents the real number domain, and They represent the probability density function and cumulative distribution function of the standard normal distribution, represents the unknown parameter vector;

[0061] The analytical form fractional moment of the high-speed crankshaft deformation function is obtained from the probability model:

[0062] (4)

[0063] Step 1.4: Establish the correction criteria between the fractional moment of high-speed crankshaft deformation obtained from experimental data and the fractional moment of analytical form:

[0064] (5)

[0065] in, Represents vectors of different orders, Represents a vector composed of fractional moments of different orders;

[0066] The following optimization model is used to solve the parameters of the probability model :

[0067] (6)

[0068] Step 1.5: Based on the optimal values ​​of unknown parameters , and obtain the corrected high-speed crankshaft deformation function probability model , calculate the reliability of high-speed crankshaft:

[0069] (7)

[0070] Step 1.6: Obtain the high-speed crankshaft deformation data collected at different times, and obtain the reliability data of complex equipment according to the above steps and , m=300 is the data length.

[0071] Step 2: Combine the operating parameter data and reliability data to establish a training data set and a test data set, establish a long short-term memory neural network model between different operating parameters and high-speed crankshaft reliability, train the neural network model, and determine whether the loss function meets the requirements. If so, obtain the high-speed crankshaft neural network model, if not, return to continue training the neural network model.

[0072] Specifically, a long short-term memory neural network model is established with different operating parameters as input and the reliability of high-speed crankshafts of complex equipment as output, including the following steps:

[0073] Step 2.1: Collect data such as high-speed crankshaft temperature, drive motor voltage and current, load parameters, etc. of complex equipment under different working conditions to build a training data set And the test dataset , F represents the parameter data such as high-speed crankshaft temperature, drive motor voltage and current, load, etc., j=4 represents the parameter dimension; the training data set and the test data set are normalized;

[0074] Step 2.2: Determine the neural network model structure, including: input layer, hidden layer (long short-term memory unit), fully connected layer, output layer. The network structure is shown in Table 1:

[0075] Table 1 Neural network structure parameters

[0076] parameter Specific value Input layer feature dimension 4 Hidden layer input feature dimension 8 Number of hidden layers 3 Hidden layer output feature dimension 12 Fully connected layer input feature dimension 12 Fully connected layer output feature dimension 1

[0077] Step 2.3: Establish the neural network training loss function:

[0078] (8)

[0079] in, is the reliability data predicted by the neural network, represents the i-th reliability data in the test data set, Represents the Euclidean distance between two points;

[0080] Step 2.4: Offline train the long short-term memory neural network model using the training data set and the test data set to obtain the neural network model parameters.

[0081] Step 3: Implement proactive fault warning based on real-time prediction of reliability of moving parts of complex equipment using monitoring data.

[0082] Specifically, the operating data of complex equipment is monitored in real time and input into the established neural network model to realize the reliability prediction of the high-speed crankshaft of the complex equipment. If the predicted high-speed crankshaft reliability does not meet the specified requirements, an alarm is issued to realize active fault warning.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the reliability of complex equipment moving parts under variable working conditions, characterized by The steps include: Step 1: by collecting performance data of complex equipment moving parts under different working conditions, to calculate the moving parts performance fractional statistical moment, and establish a probability model of the moving parts performance function containing unknown parameters, calculate the analytical form fractional moment of the moving parts performance according to the probability model, establish a correction criterion between the moving parts performance fractional statistical moment and the analytical form fractional moment, and use the optimization correction method to solve the optimal unknown parameters of the probability model to calculate the reliability of complex equipment moving parts, and finally obtain the reliability data of the moving parts at different times; specifically including the following steps: Step 1.1: Establish a reliability characterization model for complex equipment moving parts based on the performance requirements of complex equipment moving parts; Step 1.2: Collect the performance data of the moving parts under different working conditions to calculate the fractional statistical moment of the performance of the moving parts; the formula for the fractional statistical moment of the performance of the moving parts is as follows: in, represents the fractional moment of the performance parameter of the moving part, r represents the order of the fractional moment, l=1,2,...,L represents the index value of the order, L represents the preset total order, G i represents the performance data of the moving parts collected under the i-th working condition, and n represents the number of working conditions; Step 1.3: Use the logarithmic skew normal distribution to approximate the probability model of the performance function of the moving parts, and obtain the analytical form fractional moment of the performance function of the moving parts from the probability model; the logarithmic skew normal distribution approximates the probability model of the performance function of the moving parts, and the formula is as follows: Where r represents the order of the fractional moment, a∈R represents the location parameter, b>0 represents the scale parameter, c∈R represents the shape parameter, R represents the real number domain, φ(·) and Φ(·) represent the probability density function and cumulative distribution function of the standard normal distribution, respectively, and Ξ=[a,b,c] represents the unknown parameters; The analytical form of the performance function of the moving part is the fractional moment, and the formula is as follows: in, represents the analytical form fractional moment corresponding to the probability model based on the performance parameter z of the moving part, and exp(·) represents an exponential function; Step 1.4: Establish a correction criterion between the performance fractional moment of the moving parts obtained from the experimental data and the analytical form fractional moment, and use the optimization model to solve the optimal unknown parameters of the probability model; the correction criterion Q formula is as follows: Where r = [r (1) ,r (2) ,...,r (l) ,...r (L) ] T represents vectors of different orders, l=1,2,...,L represents the index value of the order, L represents the preset total order, Ξ represents the unknown parameter, represents the vector composed of different order fractional statistical moments, z represents the performance parameters of the moving parts, represents the analytical form of fractional moment, Represents a vector composed of fractional moments of different orders of analytical forms; The formula of the optimization model is as follows: Among them, OC represents the optimal value of the unknown parameter, Indicates the unknown parameter Ξ operation corresponding to the minimum value; Step 1.5: Obtain a corrected probability model of the performance function of the moving component according to the optimal unknown parameters, and calculate the reliability of the moving component through the reliability characterization model of the moving component of the complex equipment based on the corrected probability model; Step 1.6: Obtain the performance data of the moving parts collected at different times, and based on the above steps, obtain the reliability data of the moving parts of the complex equipment; Step 2: construct a neural network model between different operating parameters and complex equipment reliability data, combine the operating parameter data and reliability data, establish a training data set and a test data set for training the neural network model; Step 3: Based on the monitoring data, the reliability of moving parts of complex equipment is predicted in real time through the trained neural network model.

2. The variable operating condition reliability prediction method for complex equipment moving parts according to claim 1 is characterized by: The reliability characterization model of complex equipment moving parts established in step 1.1 is as follows: Where X represents the uncertainty parameter of the moving part, G(X) represents the performance function of the moving part, represents the performance threshold of the moving parts, f X (X) represents the probability density function of the moving component performance function, Pr{·} represents the probability operation, and z represents the moving component performance parameter.

3. The variable operating condition reliability prediction method for complex equipment moving parts according to claim 2 is characterized by: In step 1.5, the corrected moving component performance function probability model f LSND (z;Ξ OC ), reliability of moving parts: Among them, OC represents the optimal value of the unknown parameter.

4. The variable operating condition reliability prediction method for complex equipment moving parts according to claim 1 is characterized by: The neural network model in step 2 is a long short-term memory neural network model, including an input layer connected in sequence, a long short-term memory unit as a hidden layer, a fully connected layer, and an output layer; whether the training is completed is determined by whether the loss function meets the requirements. If so, the training is completed to obtain a trained neural network model of complex equipment moving parts. If not, the neural network model is continued to be trained. The loss function of the neural network model training is as follows: in, represents the reliability data predicted by the neural network, represents the i-th reliability data in the test data set, ||·|| represents the Euclidean distance between two points, and m represents the data length.

5. The variable operating condition reliability prediction method for complex equipment moving parts according to claim 1 is characterized by: The operating parameters include the internal temperature of the moving parts, the voltage and current of the driving motor, and the load, and the reliability data of the complex equipment includes the reliability of the moving parts at different times.

6. An active early warning method for failure of moving parts of complex equipment, characterized by: According to the variable operating condition reliability prediction method for complex equipment moving parts according to any one of claims 1 to 5, the reliability of complex equipment moving parts is predicted in real time. If the predicted reliability does not meet the specified requirements, an alarm is issued to provide active fault warning.

Images