A method, system, device and medium for health assessment and early damage warning of transmission mechanism of aerospace ground measurement and control equipment

Abstract

The present invention discloses a method, system, device and medium for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment, which obtains monitoring data of a rotating machine of a transmission mechanism of aerospace ground measurement and control equipment; pre-processes the monitoring data of a rotating machine of aerospace ground measurement and control equipment to obtain standardized monitoring data; calculates a health index value according to the standardized monitoring data; compares the health index value with a warning threshold, and if the health index is greater than the warning threshold, an alarm is given, otherwise no alarm is given. The method of the present invention has certain stability and robustness, and can realize fault discrimination and early damage warning of typical components of rotating machinery (such as bearings, gears, blades, etc.), and provides theoretical guidance for equipment health assessment and predictive maintenance.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis of transmission mechanisms of aerospace ground measurement and control equipment, and specifically relates to a method, system, equipment and medium for health assessment and early damage warning of transmission mechanisms of aerospace ground measurement and control equipment. Background Art

[0002] Space ground measurement and control equipment is the key channel and information center for spacecraft to communicate with the ground after launch, and provides important tasks such as telemetry and command issuance when executing spacecraft launch missions. The high-density and normalized space activities have put forward higher requirements for measurement and control antennas. Bearings and gears, as core components of the transmission mechanism of space ground measurement and control equipment, are one of the components with the worst load and work in the transmission structure. Their operating safety and reliability directly affect the smoothness and sustainability of the entire ground measurement and control equipment.

[0003] The early fault symptoms of the transmission mechanism of the measurement and control equipment are not obvious during operation, the characteristic information is weak and is often drowned out by the strong noise of the mechanical equipment during operation. The traditional method that mainly relies on manual maintenance and fault analysis can no longer meet the operation and maintenance requirements of such complex communication systems. It is easy to lack maintenance and over-maintenance, and lacks predictive maintenance guidance. In this context, by conducting health assessment and fault warning on the transmission mechanism of the equipment, real-time monitoring and analysis of key components can be achieved, and the perception of early faults can be improved. Therefore, in-depth research on the health assessment and fault warning of the transmission mechanism of aerospace ground measurement and control equipment has broad and far-reaching application prospects.

[0004] In view of the large amount of normal or healthy state data generated during the operation of rotating machinery itself, some scholars have introduced the data in the healthy state into the construction of health indicators based on the concept of data-driven. In 2013, Qiu et al. first used the self-organizing map network to evaluate the performance degradation of rolling bearings based on the health state data training strategy. The minimum quantization error was constructed to achieve fault detection and quantitative evaluation of bearing degradation. In 2017, Guo et al. used recurrent neural network (RNN) to fuse statistical indicators extracted from the time domain and frequency domain to construct the health indicators of bearings. Various advanced health indicator construction methods continue to emerge, and unsupervised fusion technology based on deep learning algorithms has gradually become a research hotspot. How to construct reliable health indicators based on normal operation data to characterize the health evolution trend of rotating machinery and achieve early fault warning has certain practical significance.

[0005] The monitoring process of equipment in industrial production relies on health indicators that can characterize its health status and deterioration trend. However, due to the scarcity of abnormal data, it is a major challenge to construct health indicators only through normal data and use them for early damage warning. At the same time, the characteristics of the time trend in the bearing degradation process bring greater difficulties to the early damage warning of bearings. Summary of the invention

[0006] In order to solve how to overcome the problems in the prior art, the purpose of the present invention is to propose a method, system, equipment and medium for health assessment and early damage warning of the transmission mechanism of aerospace ground measurement and control equipment. The method can realize health status assessment and damage warning of the transmission mechanism of aerospace ground measurement and control equipment, and has certain stability and robustness.

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

[0008] A method for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment comprises the following steps:

[0009] Obtain monitoring data of rotating machinery of transmission mechanism of aerospace ground measurement and control equipment;

[0010] Preprocess the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data;

[0011] Calculate health indicator values ​​based on standardized monitoring data;

[0012] Compare the health indicator value with the warning threshold. If the health indicator is greater than the warning threshold, an alarm is triggered, otherwise no alarm is triggered.

[0013] Further, preprocessing the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data includes the following steps:

[0014] The monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment is divided into samples, and the samples are converted from the time domain to the frequency domain and standardized to obtain standardized monitoring data.

[0015] Further, calculating the health index value based on the standardized monitoring data includes the following steps:

[0016] Calculate reconstruction error based on standardized monitoring data;

[0017] The reconstruction error is used as the input of the LSTM network, and the parameters of the LSTM network are optimized by minimizing the prediction error. The LSTM network outputs the prediction residual, which is used as a health indicator.

[0018] The trend glitches of the health indicator are eliminated before comparing the health indicator value with the warning threshold.

[0019] Furthermore, the reconstruction error is calculated using an adversarial learning strategy based on the standardized monitoring data, including the following steps:

[0020] The reconstruction error Loss is calculated as follows:

[0021] Loss=BCE+D KL +MSE

[0022] In the formula, BCE is the binary cross entropy, D KL is the KL divergence, MSE is the mean square error;

[0023] The mean square error MSE is calculated as follows:

[0024]

[0025] Where N is the number of samples, X is i It is standardized monitoring data. is the value output by the generator.

[0026] Furthermore, the KL divergence D KL The calculation is as follows:

[0027]

[0028] Where N is the number of samples, To calculate the square of the variance of two sample distributions, Computes the square of the means of two sample distributions, To calculate the natural logarithm of the variance of two sample distributions;

[0029] The binary cross entropy BCE is calculated as follows:

[0030]

[0031] Where N is the number of samples, i represents the i-th sample, and y i represents the label of the i-th sample, p(y i ) is the probability that the discriminator output belongs to the label.

[0032] Furthermore, the calculation formula of the kernel density estimation f(x) of the health indicator is as follows:

[0033]

[0034] Where K(·) is the kernel function, h is the bandwidth parameter, n is the total number of samples, i is the observation number, x is the point to estimate the probability density, and x i is the ith observation value.

[0035] Furthermore, the warning threshold is calculated by the following formula:

[0036]

[0037] Among them, HI noris a sequence of health indicators under known normal conditions, Y is the warning threshold, E represents the mean, C is the coefficient, and Var represents the variance.

[0038] A health assessment and early damage warning system for transmission mechanisms of aerospace ground measurement and control equipment, comprising:

[0039] A monitoring data acquisition module is used to acquire monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment;

[0040] A preprocessing module is used to preprocess the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data;

[0041] A health index value calculation module is used to calculate the health index value based on the standardized monitoring data;

[0042] The comparison module is used to compare the health indicator value with the warning threshold. If the health indicator is greater than the warning threshold, an alarm is issued, otherwise no alarm is issued.

[0043] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the health assessment and early damage warning method of the transmission mechanism of aerospace ground measurement and control equipment is implemented.

[0044] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the aerospace ground measurement and control equipment transmission mechanism health assessment and early damage warning method is implemented.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] Compared with traditional methods, the health assessment and early damage warning method proposed in this invention does not rely on full life data, but only uses the data of normal operation of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to build health indicators and use them for health assessment and early damage warning. At the same time, the weak trend characteristics of the health indicators during the normal operation of the bearing are taken into account, ensuring early damage warning of the rotating machinery.

[0047] Furthermore, the generator and discriminator are trained based on the adversarial learning strategy to improve the reconstructibility of the signal, and the prediction residual based on the long short-term memory network is used as a health indicator to indicate the health status of the rotating machinery.

[0048] Furthermore, the adversarial learning strategy brings new training constraints, namely, the combination of binary cross entropy, KL divergence and mean square error loss function, so that the health index constructed by this method has a certain stability and robustness, which can realize the fault identification and early damage warning of typical components of rotating machinery (such as bearings, gears, blades, etc.) of the transmission mechanism of aerospace ground measurement and control equipment, and provide theoretical guidance for equipment health assessment and predictive maintenance.

[0049] Furthermore, the present invention utilizes a long short-term memory network to capture weak trend characteristics of normal data, and uses the predicted residual as a health indicator to further improve the accuracy of the health indicator.

[0050] Furthermore, the present invention eliminates trend glitches of health indicators, thereby improving the effect of health assessment and the stability and accuracy of abnormal alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings and examples are intended to explain the present invention, and do not limit the scope of protection of the present invention.

[0052] Figure 1 The present invention proposes a general flow chart of the health assessment and early injury warning method;

[0053] Figure 2 A vibration waveform diagram of the full life data of a bearing according to an embodiment of the present invention;

[0054] Figure 3 It is a short-time Fourier time-frequency analysis diagram of the full life data of the bearing according to the embodiment of the present invention;

[0055] Figure 4 It is the probability density estimation of the health indicators under the normal state of the embodiment of the present invention;

[0056] Figure 5 is the cumulative probability density distribution of health indicators under normal conditions in the embodiment of the present invention;

[0057] Figure 6 This is a health indicator trend chart of an embodiment of the present invention.

[0058] Figure 7 It is a schematic diagram of the health assessment and early damage warning system of the transmission mechanism of aerospace ground measurement and control equipment of the present invention. DETAILED DESCRIPTION

[0059] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thoroughly and comprehensively understood.

[0060] A method for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment of the present invention firstly realizes dimensionality reduction, fusion and expression of high-dimensional data features based on a variational autoencoder; secondly, adversarial training of the generator and the discriminator is performed based on an adversarial learning strategy to improve the reconstructibility of the signal; finally, a long short-term memory network is used to capture the subtle change characteristics of the operating state of the rotating machinery, and the prediction residual is used as a health indicator to indicate the health state of the rotating machinery.

[0061] The method of the present invention specifically comprises the following steps:

[0062] Step 1: Collect the monitoring data sequence X of the normal state of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment, and divide it into samples M i , the sample M i Convert from time domain to frequency domain and perform normalization to obtain normalized data X(F).

[0063] The specific operation is to convert the monitoring data sequence X = {x1, x2, ..., x N}, divide the sample into non-overlapping windows with a local window size of L, and obtain sample M i :

[0064] M i =[m i ,…,m i+L ]

[0065] Among them, i represents the number of samples, and L is the window size.

[0066] Then use the following formula to calculate the sample M i All samples of are transformed into frequency domain to obtain the transformed samples X i ':

[0067]

[0068] Among them, ∫ represents the integral symbol, e is the base of the natural logarithm, f is the frequency, j is the imaginary unit, and t represents time.

[0069] For the transformed sample X i 'Perform maximum and minimum standardization to obtain standardized data X(F).

[0070]

[0071] Among them, min(X i ') is to calculate the converted sample X i ', max(X i ') is to calculate the converted sample X i ' is the maximum value in .

[0072] Step 2: Build and initialize the discriminative variational autoencoder model. The discriminative variational autoencoder model includes a variational autoencoder and a discriminator. The variational autoencoder includes a generator.

[0073] Through the variational autoencoder and discriminator, the adversarial learning strategy is used to obtain the reconstruction error Loss, and the dimensionality reduction fusion and expression of high-dimensional data features are achieved. Through this process, the model has a better reconstruction effect on the standardized data X(F) in step 1.

[0074] In step 2, an adversarial learning strategy is used, and the reconstructed error Loss is used as the training loss function to train the initialized discriminative variational autoencoder model until the loss function converges. The training is stopped to obtain the optimized model.

[0075] The optimal reconstruction error Loss is obtained through calculation of the optimized model;

[0076] Through this training process, the discriminator, generator, and encoder gradually optimize their parameters, making the discriminator better at distinguishing between real and generated samples, and the generator and encoder work together to produce samples similar to the real data distribution.

[0077] Among them, the reconstruction error Loss is calculated as follows:

[0078] Loss=BCE+D KL +MSE

[0079] In the formula, BCE is the binary cross entropy, D KL is the KL divergence and MSE is the mean square error.

[0080] The mean square error MSE is calculated as follows:

[0081]

[0082] Where N is the number of samples, X is i is the standardized data, is the value output by the generator.

[0083] KL divergence D KL The calculation is as follows:

[0084]

[0085] Where N is the number of samples, To calculate the square of the variance of two sample distributions, Computes the square of the means of two sample distributions, Compute the natural logarithm of the variance of two sample distributions.

[0086] The binary cross entropy BCE is calculated as follows:

[0087]

[0088] Where N is the number of samples, i represents the i-th sample, and y i represents the label of the i-th sample (0 or 1), p(y i ) is the probability that the discriminator output belongs to a label (0 or 1).

[0089] Step 3: The optimal reconstruction error calculated in the above step 2 is used as the input of the long short-term memory network, and the parameters of the long short-term memory network are optimized by minimizing the prediction error, so that the prediction residual output by the long short-term memory network is used as the health indicator HI of the operating state of the rotating machinery;

[0090] Specifically, considering the time trend characteristics of the operating state of the transmission mechanism of the aerospace ground measurement and control equipment, a sliding time window with a window length of w is used to re-divide the reconstruction error Loss into a new subsequence Window by the following formula: i , as the input of the long short-term memory network.

[0091] Window i =[x i ,x i+1 ,...,x i+w-1 ]

[0092] Where w is the size of the sliding time window.

[0093] By minimizing the prediction error and optimizing the parameters of the LSTM network, the output of the LSTM network, i.e., the prediction residual, is obtained;

[0094]

[0095] in, is the predicted value, e t is the actual value, is the output value, e i is the input value and t is the time.

[0096] The predicted residual is then defined as the health index HI:

[0097]

[0098] in, is the predicted value (output value), e t is the actual value (input value).

[0099] Step 4: Perform kernel density estimation on the health indicators of the rotating machinery operating status in step 3 to obtain the probability density estimate f(x) and cumulative probability distribution of the health indicator HI. Determine the early warning threshold Y of rotating machinery damage based on data statistical analysis according to the probability density estimate f(x) and cumulative probability distribution of the health indicator HI.

[0100] The calculation formula of the health index HI kernel density estimation f(x) is as follows:

[0101]

[0102] Where K(·) is the kernel function, and Gaussian kernel function is used here. h is the bandwidth parameter, n is the total number of samples, i is the observation number, x is the point to estimate the probability density, and x i is the ith observation value.

[0103] The Gaussian kernel function expression is as follows:

[0104]

[0105] The warning threshold Y is calculated by the following formula:

[0106]

[0107] Among them, HI nor It represents the health indicator sequence under known normal state, and Y is the warning threshold.

[0108] Step 5: Use the exponentially weighted moving average method to correct the trend burrs of the health indicator and obtain the improved health indicator EWMA τ .

[0109] EWMA τ =α×obs t +(1-α)×EWMA t-1

[0110] Among them, EWMA t is the EWMA value at time t. α is the weight decay factor, ranging from 0 to 1. obs t is the observed value at time t. EWMA t-1 is the EWMA value at time t-1.

[0111] Step 6: During the monitoring process, the monitoring data of the rotating machinery of the aerospace ground measurement and control equipment transmission mechanism is collected again. test , and the standardized monitoring data X(F) is obtained by preprocessing in step 1 test, calculate the health index value and compare it with the warning threshold determined in step 4. If the calculated health index HI>Y, otherwise no alarm will be given, thus finally achieving health assessment and abnormal alarm.

[0112] See also Figure 7 According to an embodiment of the present invention, a health assessment and early damage warning system for aerospace ground measurement and control equipment transmission mechanism includes:

[0113] A monitoring data acquisition module is used to acquire monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment;

[0114] A preprocessing module is used to preprocess the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data;

[0115] A health index value calculation module is used to calculate the health index value based on the standardized monitoring data;

[0116] The comparison module is used to compare the health indicator value with the warning threshold. If the health indicator is greater than the warning threshold, an alarm is issued, otherwise no alarm is issued.

[0117] An electronic device according to an embodiment of the present invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment when executing the computer program.

[0118] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, and when the computer program is executed by a processor, the method for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment is implemented.

[0119] In order to illustrate the specific implementation of the present invention in detail, the health assessment and early damage warning of the PHM bearing full life data are taken as an example.

[0120] The experimental data comes from the PRONOSTIA test bench, which consists of key components such as AC motor, reducer, torque meter, thermocouple, NI data acquisition card, voltage regulator, hydraulic cylinder, pressure sensor, coupling, and tested bearing. Speed ​​sensor and acceleration sensor are arranged to collect signals. The test bench is driven by AC motor, which drives the tested bearing to work after being decelerated by the reducer. The load acting on the tested bearing can be changed by adjusting the cylinder pressure. The test bench is used to study the fault detection, diagnosis and prediction methods of bearings.

[0121] The experimental conditions are a speed of 1800RPM and a radial load of 4000N. The sampling frequency of the vibration signal is 25.6kHz, and it is collected once every 10s. The length of each signal segment is 2560 sampling points. The experiment is stopped when the amplitude of the test bearing exceeds 20g. In the end, a total of 2803 sets of monitoring data are obtained. The time domain waveform of the full life data of the bearing is as follows Figure 2 As shown:

[0122] The specific steps of the method of the present invention are as follows:

[0123] 1) Collect the monitoring data X (the first 600 sets of monitoring data) of the bearing from the above-mentioned test bench, and obtain the monitoring data sequence as follows:

[0124] X={x1,x2,…,x N}

[0125] The samples are divided into non-overlapping windows with a local window size of L. The resulting sample description is as follows:

[0126] M i =[m i ,…,m i+L ]

[0127] In order to improve the model's sensitivity to faults, the following formula is used to calculate M i All samples of are converted into frequency domain to obtain X i ':

[0128]

[0129] Among them, ∫ represents the integral symbol, e is the base of the natural logarithm, f is the frequency, j is the imaginary unit, and t represents time.

[0130] X i 'Perform maximum and minimum standardization to obtain the standardized result X(F).

[0131]

[0132] Among them, min(X i ') is used to calculate X i ', max(X i ') is used to calculate X i ' is the maximum value in .

[0133] 2) Construct and initialize the discriminative variational autoencoder model. Through the variational autoencoder and discriminator modules, use the adversarial learning strategy to achieve dimensionality reduction, fusion and expression of high-dimensional data features. Through this process, the model has a better reconstruction effect on the data X(F) in step 1.

[0134] Among them, the mean square error MSE and KL divergence D are used KL , binary cross entropy BCE function as training loss function. Mean square error MSE, KL divergence D KL , the binary cross entropy BCE loss function is calculated as follows:

[0135]

[0136] Where N is the number of samples, X i is the value of the input sample, is the value output by the generator.

[0137]

[0138] Where N is the number of samples, To calculate the square of the variance of two sample distributions, Computes the square of the means of two sample distributions, Compute the natural logarithm of the variance of two sample distributions.

[0139]

[0140] Where N is the number of samples, i represents the i-th sample, and y i represents the label of the i-th sample (0 or 1), p(y i ) is the probability that the discriminator output belongs to a label (0 or 1).

[0141] Finally, the reconstruction error Loss is obtained:

[0142] Loss=BCE+D KL +MSE

[0143] 3) For the optimization network model constructed in step 2, the reconstruction error sequence Loss and its distribution of the standardized data X(F) of the experimental platform bearing in normal state are obtained. Considering the time trend characteristics of the degradation process of rotating machinery, the reconstruction error sequence is re-divided into new subsequences along the time dimension according to the step size. The subsequence is expressed as follows:

[0144] Window i =[x i ,x i+1 ,...,x i+w-1 ]

[0145] Wherein, w is the sliding time window size, and w=20 is taken.

[0146] Based on the long short-term memory network, a time series prediction model is constructed for the reconstruction error Loss in step 2. The long short-term memory network model parameters are optimized by minimizing the prediction error, and finally the health index HI of the rotating machinery operating status is obtained.

[0147] The LSTM model parameters are optimized by minimizing the prediction error, i.e.

[0148]

[0149] The prediction residuals are defined as the health indicator, which is defined as follows:

[0150]

[0151] in, is the output value (prediction value), e i is the input value (real value).

[0152] 4) Perform kernel density estimation on the health index distribution of the bearing in normal state in step 3 to obtain the probability density estimation and cumulative probability distribution of the health index, and determine the early warning threshold of its damage based on mathematical statistical analysis.

[0153]

[0154] Among them, K(·) is the kernel function. Here we use the Gaussian kernel function, and the calculation formula is as follows:

[0155]

[0156] h is the bandwidth parameter, which controls the width of the kernel and affects the degree of smoothing. n is the total number of samples. i is the i-th group of samples.

[0157] Based on this step, the health indicators of the normal state of the bearing in the experimental data (600 groups of samples) are used to obtain the overall distribution of the health indicators under normal conditions, the probability density distribution and the cumulative probability distribution of the obtained health indicators. Figure 4 and Figure 5 As shown, the alarm threshold is calculated based on the 3σ principle. Figure 4 The vertical dashed line in the figure represents the position of the alarm threshold.

[0158]

[0159] Among them, HI nor Represents the indicator sequence under known normal conditions. When C=3, 99.73% of the health indicator values ​​under this condition are within this threshold range.

[0160] 5) Due to background noise, abnormal noise and random fluctuations, the overall health index shows a monotonous increase, while the local performance is fluctuation and oscillation, and even deviates from the overall trend. This phenomenon is called trend burr phenomenon. In order to better improve and enhance the evaluation effect of health indicators, the exponential weighted moving average is used to correct the trend burr of health indicators. The corrected health indicator is the improved health indicator, and its calculation method is as follows:

[0161] EWMA τ =α×obs t +(1-α)×EWMA t-1

[0162] Among them, EWMA t is the EWMA value at time t. α is the weight decay factor, ranging from 0 to 1. obs t is the observed value at time t. EWMA t-1 is the EWMA value at time t-1.

[0163] 6) The full life cycle data collected in the experiment (a total of 2803 sets of data) are used as test samples to verify the effectiveness of the proposed method. Similarly, the test samples use the data processing process in step 1 to obtain standardized monitoring data X(F) test The network model built in steps 2 and 3 is used to directly calculate the health index HI value (one HI value is calculated for each set of data), and the health index is improved by the method in step 5, and finally a HI sequence of length 2803 is formed. The HI is compared with the warning threshold determined in step 4. If the health index is greater than the warning threshold, an alarm is triggered, otherwise no alarm is triggered.

[0164] The changing trends of health indicators in the PHM dataset are as follows: Figure 6 As shown, the red dotted line represents the warning threshold. It can be seen that in the early stage of bearing operation, that is, in the healthy state, the health index always remains in a relatively stable range. With the running time, the bearing health index exceeds the set threshold at around 1400 samples. Moreover, when the bearing begins to deteriorate, the value of the health index increases significantly with the intensification of vibration, which can show that the proposed method can correctly perform health assessment and early damage warning for rotating machinery.

[0165] The time-frequency analysis of PHM data is performed, and the obtained time-frequency diagram is as follows: Figure 3 As shown in the figure, from the perspective of the entire frequency band, after the vertical dotted line, which is close to the warning position of the proposed method, many frequency bands appear more obvious and show a brighter trend, which can be regarded as early damage. From the perspective of time and frequency, the effectiveness of the warning of the proposed method is confirmed.

[0166] The present invention can perform health assessment of the operating status of rotating machinery only through the data of the normal state of rotating machinery, introduce adversarial learning strategy, and consider the time trend characteristics. From the experimental results, the health index constructed by this method has a stronger sensitivity to early damage, has certain stability and robustness, and can realize fault identification and early damage warning of typical components of rotating machinery (such as bearings, gears, blades, etc.), providing theoretical guidance for equipment health assessment and predictive maintenance.

[0167] The above description is only for the best embodiment of the present invention, but it should not be understood as limiting the claims. The present invention is not limited to the above embodiments, and its specific structure is allowed to be changed. However, all changes made within the protection scope of the independent claims of the present invention are within the protection scope of the present invention.

[0168] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

Claims

1. A method for health assessment and early damage warning of a transmission mechanism of aerospace ground measurement and control equipment, characterized in that: The following steps are involved: Obtain monitoring data of rotating machinery of transmission mechanism of aerospace ground measurement and control equipment; Preprocess the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data; Calculate health indicator values ​​based on standardized monitoring data; Compare the health indicator value with the warning threshold. If the health indicator is greater than the warning threshold, an alarm is triggered. Otherwise, no alarm is triggered. Calculating health indicator values ​​based on standardized monitoring data includes the following steps: Calculate reconstruction error based on standardized monitoring data; The reconstruction error is used as the input of the LSTM network, and the parameters of the LSTM network are optimized by minimizing the prediction error. The LSTM network outputs the prediction residual, which is used as a health indicator. The trend glitches of the health indicator are eliminated before comparing the health indicator value with the warning threshold.

2. The method for health assessment and early damage warning of aerospace ground measurement and control equipment transmission mechanism according to claim 1 is characterized in that: Preprocessing the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data includes the following steps: The monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment is divided into samples, and the samples are converted from the time domain to the frequency domain and standardized to obtain standardized monitoring data.

3. The method for health assessment and early damage warning of aerospace ground measurement and control equipment transmission mechanism according to claim 1 is characterized in that: The reconstruction error is calculated using an adversarial learning strategy based on standardized monitoring data, including the following steps: The reconstruction error Loss is calculated as follows: Loss=BCE+D KL +MSE In the formula, BCE is the binary cross entropy, D KL is the KL divergence, MSE is the mean square error; The mean square error MSE is calculated as follows: Where N is the number of samples, X i It is standardized monitoring data. is the value output by the generator.

4. The method for health assessment and early damage warning of aerospace ground measurement and control equipment transmission mechanism according to claim 3 is characterized in that: KL divergence D KL The calculation is as follows: Where N is the number of samples, To calculate the square of the variance of two sample distributions, Computes the square of the means of two sample distributions, To calculate the natural logarithm of the variance of two sample distributions; The binary cross entropy BCE is calculated as follows: Where N is the number of samples, i represents the i-th sample, and y i represents the label of the i-th sample, p(y i ) is the probability that the discriminator output belongs to the label.

5. The method for health assessment and early damage warning of aerospace ground measurement and control equipment transmission mechanism according to claim 1 is characterized in that: The calculation formula of the kernel density estimate f(x) of the health indicator is as follows: Where K(·) is the kernel function, h is the bandwidth parameter, n is the total number of samples, i is the observation number, x is the point to estimate the probability density, and x i is the ith observation value.

6. The method for health assessment and early damage warning of aerospace ground measurement and control equipment transmission mechanism according to claim 1 is characterized in that: The warning threshold is calculated by the following formula: Among them, HI nor is a sequence of health indicators under known normal conditions, Y is the warning threshold, E represents the mean, C is the coefficient, and Var represents the variance.

7. A health assessment and early damage warning system for transmission mechanisms of aerospace ground measurement and control equipment, characterized in that: include: A monitoring data acquisition module is used to acquire monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment; A preprocessing module is used to preprocess the monitoring data of the rotating machinery of the transmission mechanism of the aerospace ground measurement and control equipment to obtain standardized monitoring data; A health index value calculation module is used to calculate the health index value based on the standardized monitoring data; The comparison module is used to compare the health indicator value with the warning threshold. If the health indicator is greater than the warning threshold, an alarm is triggered, otherwise no alarm is triggered. Calculating health indicator values ​​based on standardized monitoring data includes the following steps: Calculate reconstruction error based on standardized monitoring data; The reconstruction error is used as the input of the LSTM network, and the parameters of the LSTM network are optimized by minimizing the prediction error. The LSTM network outputs the prediction residual, which is used as a health indicator. The trend glitches of the health indicator are eliminated before comparing the health indicator value with the warning threshold.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for health assessment and early damage warning of the transmission mechanism of aerospace ground measurement and control equipment as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the method for health assessment and early damage warning of the transmission mechanism of aerospace ground measurement and control equipment as described in any one of claims 1 to 6 is implemented.

Images