A method and system for evaluating performance degradation of an electro-hydraulic servo mechanism

Abstract

The application relates to a kind of electric static pressure servo mechanism performance degradation evaluation method and system, comprising: the multichannel sensing signal of electric static pressure servo mechanism is collected, and data set is divided into training data set and test data set;Respectively, initial feature set is extracted to training and test data set, and the initial feature set includes time domain feature, frequency domain feature and efficiency feature based on physical model;Robust feature insensitive to variable load is screened out in the initial feature set;After screening, the multidimensional robust feature is fused and reduced dimension, and low-dimensional sensitive feature set is obtained;Performance degradation evaluation model is established based on the low-dimensional sensitive feature set, and the evaluation grade result of electric static pressure servo mechanism is output by using the performance degradation evaluation model.The beneficial effects of the application are that: the current running state of electric static pressure servo system can be better understood by the depth feature extraction and fusion of multiple source sensing signals than traditional state evaluation.

Description

Technical Field

[0001] This invention belongs to the field of performance evaluation technology, and in particular relates to a method and system for evaluating the performance degradation of an electrostatic servo mechanism. Background Technology

[0002] In spacecraft flight control systems, the actuator engine (EHA) is a critical actuator, and its reliability directly affects flight safety. As space missions place increasingly higher demands on the safety and reliability of actuation systems, the performance degradation of the EHA has become a prominent issue. During long-term operation, wear and aging of key components can lead to internal leaks in the hydraulic system, resulting not only in energy loss but also affecting the dynamic response and control accuracy of the actuators. Internal leakage in the axial piston pump has become one of the main failure modes threatening spacecraft safety.

[0003] Currently, most research on EHA (Extreme Harshness) focuses on fault diagnosis, specifically how to quickly locate and identify obvious system faults. However, this post-hoc analysis approach does not fully consider the performance degradation trend and health evolution of EHA during long-term operation, and research on EHA performance degradation assessment is relatively limited. Furthermore, existing EHA performance degradation assessment methods are mostly modeled and analyzed under the assumption of constant load or ideal operating conditions, with some studies even completely ignoring the impact of load changes on system performance degradation. This approach leads to low accuracy of the assessment models when facing real, variable load environments.

[0004] Therefore, this paper proposes a performance degradation assessment method for electrostatic servo mechanisms (EHAs), which can provide a deeper understanding of the evolution trend of their health status and provide a scientific basis for engineers to formulate predictive maintenance strategies. This can effectively avoid sudden failures and improve the reliability and safety of the system, which is of great significance for ensuring the safe and reliable operation of EHAs and the high-end equipment they are in. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for evaluating the performance degradation of electrostatic servo mechanisms.

[0006] Firstly, a method for evaluating the performance degradation of an electrostatic servo mechanism is provided, including:

[0007] Step 1: Collect multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and test dataset;

[0008] Step 2: Extract initial feature sets from the training and test datasets respectively. The initial feature sets include time-domain features, frequency-domain features, and efficiency features based on the physical model.

[0009] Step 3: Select robust features that are insensitive to variable loads from the initial feature set;

[0010] Step 4: Fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set;

[0011] Step 5: Establish a performance degradation evaluation model based on the low-dimensional sensitive feature set, and use the performance degradation evaluation model to output the evaluation level result of the electrostatic servo mechanism.

[0012] Preferably, in step 1, the multi-channel sensing signals include motor speed, pressure difference across the plunger pump, oil temperature, and actuator displacement signal.

[0013] Preferably, step 2 includes:

[0014] Step 2.1: Extract time-domain features from the acquired multi-channel data, including: extracting features that can provide overall signal state information, energy changes, and impact based on the multi-channel data curves;

[0015] Step 2.2: Extract frequency domain features from the acquired multi-channel data, including: extracting features that provide information on the frequency components and energy distribution of the signal based on the multi-channel data curves;

[0016] Step 2.3: Extract efficiency features based on the physical model from the collected multi-channel data, including: calculating the volumetric efficiency index, which is defined by the ratio of actuator speed to motor speed.

[0017] Preferably, step 3 includes:

[0018] Step 3.1: Calculate the maximum information coefficient value between each feature in the initial feature set and the load change condition using the maximum information coefficient algorithm;

[0019] Step 3.2: Set a maximum information coefficient threshold, and remove features whose maximum information coefficient value is higher than the threshold to obtain a robust feature subset.

[0020] Preferably, in step 3.2, when setting the maximum information coefficient threshold, the Spearman rank correlation coefficient method is used to perform correlation analysis between each initial feature and the degree of leakage of the plunger pump, and sensitive features with large correlation coefficients are retained to determine the final threshold.

[0021] Preferably, step 3.2 further includes: normalizing the robust feature subset using the Min-Max normalization method.

[0022] Preferably, in step 4, a sparse autoencoder is used to fuse and reduce the dimensionality of the multidimensional robust features.

[0023] In a second aspect, a performance degradation evaluation system for an electrostatic servo mechanism is provided, for performing any of the methods described in the first aspect, including:

[0024] The acquisition module is used to acquire multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and experimental dataset.

[0025] The extraction module is used to extract an initial feature set from the acquired multi-channel data. The initial feature set includes time-domain features, frequency-domain features, and efficiency features based on a physical model.

[0026] A filtering module is used to filter out robust features that are insensitive to variable loads from the initial feature set;

[0027] The fusion and dimensionality reduction module is used to fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set.

[0028] A module is established to build a performance degradation evaluation model based on the low-dimensional sensitive feature set, and to output the evaluation level result of the electrostatic servo mechanism using the performance degradation evaluation model.

[0029] Thirdly, a computer storage medium is provided, wherein a computer program is stored therein; when the computer program is run on a computer, the computer causes the computer to perform any of the methods described in the first aspect.

[0030] Fourthly, an electronic device is provided, comprising:

[0031] Memory, used to store computer programs;

[0032] A processor for executing the computer program to implement the method as described in any of the first aspects.

[0033] The beneficial effects of this invention are:

[0034] 1. The present invention provides a method for evaluating the performance degradation of an electrostatic servo system. By extracting and fusing deep features from multi-source sensor signals, it can provide a better understanding of the current operating status of the electrostatic servo system than traditional state assessment.

[0035] 2. The efficiency feature based on the physical model introduced in this invention calculates the ratio of the actual output of the system to the theoretical input. It fundamentally utilizes the normalization concept to cancel out external load changes as a common influencing factor in mathematical operations. This makes the feature numerically decoupled from the load size and only sensitive to the actual performance degradation caused by internal leakage.

[0036] 3. The core advantage of the electrostatic servo system performance degradation assessment method provided by this invention lies in the construction of an assessment model capable of adapting to varying load conditions. This method effectively decouples the instantaneous impact of load fluctuations on performance indicators, thereby accurately extracting degradation characteristics that reflect the true health status of the system. This fundamentally solves the problem of inaccurate assessment results caused by changes in operating conditions in traditional methods, greatly improving the accuracy and reliability of the assessment. Attached Figure Description

[0037] Figure 1 This is a flowchart of a method for evaluating the performance degradation of an electrostatic servo system provided by the present invention;

[0038] Figure 2 This is a structural diagram of an electrostatic servo system performance degradation assessment system provided by the present invention. Detailed Implementation

[0039] The present invention will be further described below with reference to embodiments. The description of the embodiments below is only for the purpose of helping to understand the present invention. It should be noted that those skilled in the art can make several modifications to the present invention without departing from the principle of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0040] Example 1:

[0041] To address the problems of the prior art, Embodiment 1 of this application provides a method for evaluating the performance degradation of an electrostatic servo mechanism, comprising:

[0042] Step 1: Collect multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and test dataset.

[0043] In step 1, the multi-channel sensing signals include motor speed, pressure difference across the plunger pump, oil temperature, and actuator displacement signals. Furthermore, the training dataset is used to train the neural network performance degradation evaluation model for the electrostatic servo mechanism, and the experimental dataset is used for experimental verification.

[0044] Step 2: Extract initial feature sets from the training and test datasets respectively. The initial feature sets include time-domain features, frequency-domain features, and efficiency features based on the physical model.

[0045] Step 2 includes:

[0046] Step 2.1: Extract time-domain features from the acquired multi-channel data, including: extracting features that can provide overall signal state information, energy changes, and impact based on the multi-channel data curves.

[0047] Step 2.2: Extract frequency domain features from the acquired multi-channel data, including: extracting features that provide the frequency components and energy distribution of the signal based on the multi-channel data curves.

[0048] Step 2.3: Extract efficiency features based on the physical model from the collected multi-channel data, including: calculating the volumetric efficiency index, which is defined by the ratio of actuator speed to motor speed.

[0049] In step 2.3, this application first needs to introduce the physical nature of quantification leakage—volume loss. By obtaining the conversion efficiency between theoretical input and actual output, the influence of load changes is naturally eliminated, while retaining the sensitivity of characteristics to the degree of failure.

[0050] First, any change in the external load of a power system will inevitably cause a synchronous change in its control input and state output. In an electro-hydraulic servo actuator (EHA) mechanism, in order to overcome the increased external load, its closed-loop control system will inevitably instruct the motor to increase its speed (theoretical input increases), thereby driving the pump to output a larger flow rate, causing the actuator to generate a corresponding speed (actual output increases). In this dynamic balance, there is an inherent linear proportional relationship between the theoretical output and the actual output.

[0051] In hydraulic systems, the volumetric efficiency of a pump is defined as the ratio of the pump's actual output flow rate to its theoretical output flow rate. The formula for volumetric efficiency is as follows:

[0052]

[0053] In the formula, For volumetric efficiency, This is the actual output flow rate of the pump. This is the pump's theoretical output flow rate. For the pump's fixed displacement, The pump speed, The effective area of ​​the actuator piston. In actuator speed engineering, we are typically more concerned with the relative changes of characteristics than their absolute physical values. The effective area of ​​the actuator piston and the displacement of the pump are inherent physical constants of the system. Therefore, they can be combined into a single constant. .

[0054] To construct a simple, effective, and dimensionless volumetric efficiency index (VE), this application omits the coefficient k, and the volumetric efficiency index formula is as follows:

[0055]

[0056] In the formula, VE is the volumetric efficiency index, and x is the actuator displacement. n is the actuator speed, and n is the motor speed.

[0057] When the system is in good working order, and the external load changes: When the load increases, the controller instructs the motor speed to increase to provide more power. Consequently, the theoretical output flow rate of the pump increases, and the speed of the drive actuator also increases accordingly. Since there are no abnormal energy or volumetric losses within the system, the input (speed N) and output (speed v) change proportionally and in tandem. Therefore, the calculated volumetric efficiency index will remain essentially constant. Similarly, when the load decreases, both speed N and speed v will decrease proportionally, and the volumetric efficiency index will remain stable.

[0058] By constructing this ratio-based feature, changes in external load act as a common influencing factor on both theoretical and actual outputs, and their impact is effectively canceled out in the formula. This decouples the extracted efficiency feature numerically from the magnitude of the external load, thus achieving natural robustness to variable load conditions.

[0059] Secondly, the physical essence of internal leakage in a pump is the formation of an unintended flow channel between the high-pressure and low-pressure chambers, causing a portion of the high-pressure oil to leak directly without being used to drive the actuator. The magnitude of this leakage flow is directly related to the geometric dimensions of the leakage channel (i.e., the severity of the fault). When an internal leakage fault occurs in the system, the proportional relationship in the aforementioned volumetric efficiency formula is disrupted.

[0060] At this point, the actual speed v of the actuator is no longer determined solely by the motor speed n, but by the effective flow rate after subtracting the leakage flow rate from the pump's theoretical flow rate. Therefore, the volumetric efficiency index will decrease as the degree of leakage increases.

[0061] Since the leakage rate is a monotonic function of the severity of the fault, and the decrease in efficiency indicators is also a monotonic function of the leakage rate, a clear and quantifiable monotonic mapping relationship is established between the extracted volumetric efficiency indicators and the severity of the fault. The magnitude of the decrease in eigenvalues ​​directly reflects the severity level of the leakage.

[0062] Step 3: Select robust features that are insensitive to variable loads from the initial feature set.

[0063] Step 4: Fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set.

[0064] Step 5: Establish a performance degradation evaluation model based on the low-dimensional sensitive feature set, and use the performance degradation evaluation model to output the evaluation level result of the electrostatic servo mechanism.

[0065] Example 2:

[0066] Based on Example 1, Example 2 of this application provides a more specific method for evaluating the performance degradation of electrostatic servo mechanisms, including:

[0067] Step 1: Collect multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and experimental dataset.

[0068] Step 2: Extract an initial feature set from the acquired multi-channel data. The initial feature set includes time-domain features, frequency-domain features, and efficiency features based on the physical model.

[0069] Step 3: Select robust features that are insensitive to variable loads from the initial feature set.

[0070] Step 3 includes:

[0071] Step 3.1: Calculate the maximum information coefficient value between each feature in the initial feature set and the load change conditions using the maximum information coefficient (MIC) algorithm to measure the correlation between the feature and the load.

[0072] In step 3.1, the higher the MIC value, the greater the influence of load on the feature.

[0073] The formula for calculating the maximum information coefficient is as follows:

[0074]

[0075] In the formula: x and y are two random variables, a and b are the number of grids in the x and y directions respectively, and B is the upper limit of the number of grids; I(X, Y) is the mutual information value, i.e.

[0076]

[0077] In the formula: p(x) and p(y) are the probability density functions of x and y respectively; p(x, y) is the joint probability density function.

[0078] Step 3.2: Set a maximum information coefficient threshold, remove features whose maximum information coefficient value is higher than the threshold, and retain the robust feature subset that is not sensitive to load changes but sensitive to fault evolution.

[0079] In step 3.2, when setting the maximum information coefficient threshold, Spearman's rank correlation coefficient method is used to perform correlation analysis between each initial feature and the degree of leakage of the plunger pump, and sensitive features with large correlation coefficients are retained to determine the final threshold.

[0080] For example, based on the calculation results, select Robust features are identified, and redundant features are removed.

[0081] Step 3.2 also includes: normalizing the robust feature subset using the Min-Max normalization method.

[0082] Specifically, the calculation formula is as follows:

[0083]

[0084] In the formula: Refers to data after normalization; The largest data point in the dataset; The smallest data point in the dataset; The data is to be standardized.

[0085] Step 4: Fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set.

[0086] In step 4, a sparse autoencoder (SAE) is used to fuse and reduce the dimensionality of the multidimensional robust features to obtain deep fused features that are sensitive to performance degradation.

[0087] Specifically, after selecting robust features, to further remove redundant information between features and extract the core information most sensitive to faults, an SAE deep learning network is used to process the robust feature set. Pre-defined training datasets are input into the feature fusion model to train and obtain a low-dimensional sensitive feature set.

[0088] The robust features mentioned above are fused using a sparse autoencoder. By fusing and comprehensively evaluating correlation, robustness, and monotonicity, corresponding performance indicators are obtained to quantitatively reflect the quality of the fused sensitive features.

[0089] The correlation metric is used to measure whether a feature can capture the performance degradation trend in an electrostatic servo system.

[0090]

[0091] In the formula, K represents the total number of sampling points, x i This represents the trend sequence of the acquired feature indicators, X=(x1,x2,...,x...). k Y = (y1, y2, ..., y) represents the sequence of acquired feature indicators. k ) represents the corresponding performance degradation sequence.

[0092] Quantitative correlation can measure the linear correlation between features and time, and thus determine the correlation between features and the degree of degradation of the electrostatic servo system.

[0093] Monotonicity refers to the consistency of the declining trend in the performance of mechanical equipment. Performance degradation is inevitable for any mechanical equipment during use. Therefore, the acquired characteristic indicators should exhibit a monotonic trend.

[0094]

[0095] In the formula, N represents the total number of fault levels. It is the feature mean corresponding to the j-th fault level. This represents a step function. The monotonicity of quantization features helps to remove features that exhibit sharp fluctuations and do not imply a degenerative trend.

[0096] Robustness refers to the tolerance of a feature measure to random noise and outliers.

[0097]

[0098] In the formula, K represents the total number of sampling points, x i This represents the trend sequence of the acquired feature indicators. This represents the trend term obtained by applying a moving average method to feature X. X=(x1,x2,...,x...) k ) represents the sequence of acquired feature indicators.

[0099] The values ​​of the three evaluation indicators are all within the range of [0,1] and are positively correlated with the characteristic indicator. To synthesize these three evaluation indicators, a linear weighting method as shown in the formula will be used as the method for selecting evaluation indicators.

[0100] ,

[0101] ,

[0102]

[0103] In the formula, E represents the linear weighted value of the three evaluation indicators. This indicates the attribute weight of the indicator. This represents the weight of the i-th attribute at the j-th level, i.e., the weight of the attribute at the total level. The percentage of the evaluation index is given by E, where i=j. E is positively correlated with the three evaluation indicators. The larger E is, the better the overall performance of the evaluation indicators, but it can also better reflect the performance degradation trend of the electrostatic servo system.

[0104] Each feature is fundamentally different and has different attribute weights. However, there are multiple methods for calculating these weights. Therefore, to avoid external interference, a weighted formula method is chosen to calculate the weights.

[0105]

[0106] In the formula, ω i1 =1, n is the number of attributes, i represents the i-th attribute, and j is the queue level.

[0107] Step 5: Establish a performance degradation evaluation model based on the low-dimensional sensitive feature set, and use the performance degradation evaluation model to output the evaluation level result of the electrostatic servo mechanism.

[0108] Specifically, firstly, multiple preset EHA fault levels are divided into three state levels: healthy, degraded, and faulty. Secondly, the training datasets corresponding to different fault levels are processed through the steps of initial feature extraction, robust feature selection, and deep fusion feature extraction to form training feature vectors with state level labels. Then, low-dimensional sensitive features are input to train the high-performance performance degradation evaluation model.

[0109] Initial features were extracted from the experimental dataset, robust features were initially screened, and features were fused and dimensionality reduced. Finally, the data were input into the performance degradation evaluation model of the electrostatic servo system. The accuracy of the evaluation model was verified by comparing the output health status with the actual health status.

[0110] It should be noted that the parts in this embodiment that are the same as or similar to those in Embodiment 1 can be referred to each other, and will not be repeated in this application.

[0111] Example 3:

[0112] Based on Embodiment 2, Embodiment 3 of this application provides a performance degradation evaluation system for an electrostatic servo mechanism, comprising:

[0113] The acquisition module is used to acquire multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and experimental dataset.

[0114] The extraction module is used to extract an initial feature set from the acquired multi-channel data. The initial feature set includes time-domain features, frequency-domain features, and efficiency features based on a physical model.

[0115] A filtering module is used to filter out robust features that are insensitive to variable loads from the initial feature set;

[0116] The fusion and dimensionality reduction module is used to fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set.

[0117] A module is established to build a performance degradation evaluation model based on the low-dimensional sensitive feature set, and to output the evaluation level result of the electrostatic servo mechanism using the performance degradation evaluation model.

[0118] It should be noted that the system provided in this embodiment is the corresponding system of the method provided in embodiment 2. Therefore, the parts that are the same as or similar to those in embodiment 2 in this embodiment can be referred to each other, and will not be described again in this application.

Claims

1. A method for evaluating the performance degradation of an electrostatic servo mechanism, characterized in that, include: Step 1: Collect multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and test dataset; Step 2: Extract initial feature sets from the training and test datasets respectively. The initial feature sets include time-domain features, frequency-domain features, and efficiency features based on the physical model. Step 3: Select robust features that are insensitive to variable loads from the initial feature set; Step 3 includes: Step 3.1: Calculate the maximum information coefficient value between each feature in the initial feature set and the load change condition using the maximum information coefficient algorithm; Step 3.2: Set a maximum information coefficient threshold, and remove features whose maximum information coefficient value is higher than the threshold to obtain a robust feature subset; Step 4: Fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set; Step 5: Establish a performance degradation evaluation model based on the low-dimensional sensitive feature set, and use the performance degradation evaluation model to output the evaluation level result of the electrostatic servo mechanism.

2. The method for evaluating the performance degradation of an electrostatic servo mechanism according to claim 1, characterized in that, In step 1, the multi-channel sensing signals include motor speed, pressure difference across the plunger pump, oil temperature, and actuator displacement signal.

3. The method for evaluating the performance degradation of an electrostatic servo mechanism according to claim 2, characterized in that, Step 2 includes: Step 2.1: Extract time-domain features from the acquired multi-channel data, including: extracting features that can provide overall signal state information, energy changes, and impact based on the multi-channel data curves; Step 2.2: Extract frequency domain features from the acquired multi-channel data, including: extracting features that provide information on the frequency components and energy distribution of the signal based on the multi-channel data curves; Step 2.3: Extract efficiency features based on the physical model from the collected multi-channel data, including: calculating the volumetric efficiency index, which is defined by the ratio of actuator speed to motor speed.

4. The method for evaluating the performance degradation of an electrostatic servo mechanism according to claim 3, characterized in that, In step 3.2, when setting the maximum information coefficient threshold, Spearman's rank correlation coefficient method is used to perform correlation analysis between each initial feature and the degree of leakage of the plunger pump, and sensitive features with large correlation coefficients are retained to determine the final threshold.

5. The method for evaluating the performance degradation of an electrostatic servo mechanism according to claim 4, characterized in that, Step 3.2 also includes: normalizing the robust feature subset using the Min-Max normalization method.

6. The method for evaluating the performance degradation of an electrostatic servo mechanism according to claim 5, characterized in that, In step 4, a sparse autoencoder is used to fuse and reduce the dimensionality of the multidimensional robust features.

7. A performance degradation evaluation system for an electrostatic servo mechanism, characterized in that, For performing the method according to any one of claims 1 to 6, comprising: The acquisition module is used to acquire multi-channel sensor signals from the electrostatic servo mechanism and divide the dataset into training dataset and experimental dataset. The extraction module is used to extract an initial feature set from the acquired multi-channel data. The initial feature set includes time-domain features, frequency-domain features, and efficiency features based on a physical model. A filtering module is used to filter out robust features that are insensitive to variable loads from the initial feature set; The fusion and dimensionality reduction module is used to fuse and reduce the dimensionality of the selected multidimensional robust features to obtain a low-dimensional sensitive feature set. A module is established to build a performance degradation evaluation model based on the low-dimensional sensitive feature set, and to output the evaluation level result of the electrostatic servo mechanism using the performance degradation evaluation model.

8. A computer storage medium, characterized in that, The computer storage medium stores a computer program; when the computer program is run on the computer, it causes the computer to perform the method described in any one of claims 1 to 6.

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

Images