Engine main shaft stability detection system and detection method thereof

Abstract

The invention relates to the field of intelligent detection, and discloses an engine main shaft stability detection system and a detection method thereof. In the process of detecting the stability of the engine main shaft, discrete wavelet transform and local time-frequency analysis are carried out on a vibration signal of the detected engine main shaft under a preset load condition, so that tiny vibration characteristics of the detected engine main shaft are excavated; according to the method, the key vibration mode of the engine main shaft is captured through dynamic aggregation of the local time-frequency characteristics of the vibration signals based on characteristic energy distribution, full-frequency-domain characteristic perception of the vibration signals is achieved, and whether abnormal vibration occurs on the engine main shaft under the preset load condition or not is judged based on the full-frequency-domain characteristic perception. In this way, the slight change of the stability of the detected engine main shaft in the operation process can be captured, and the accuracy of engine main shaft anomaly detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection, and more particularly to an engine main shaft stability detection system and a detection method thereof. Background Art

[0002] With the continuous improvement of industrial automation, higher requirements are placed on the operating status monitoring of mechanical equipment, especially engines. As the core component of many mechanical equipment, the stability of the main shaft of the engine directly affects the performance and safety of the entire system. Therefore, ensuring the stability of the main shaft operation is crucial to improving the working efficiency of the engine, extending its service life and ensuring safety.

[0003] In this regard, the invention patent with publication number CN110617967A discloses a universal engine main shaft stability detection device, which includes a detection mechanism, an engine fixing mechanism, a tensioning mechanism, and a platform mechanism. Through the synergistic effect of multiple mechanisms, the detection device can install engines of different sizes and detect whether the main shafts of engines of different sizes are running smoothly. Specifically, if the main shaft of the engine being tested is not running smoothly, it will contact the belt shaft friction wheel. After contact, the belt shaft friction wheel will drive the transmission roller to rotate, and then drive the display pulley to rotate through the belt. The rotation of the display pulley will further drive the rotation of the impeller. If the impeller does not rotate, it means that the main shaft of the engine being tested is running smoothly and meets the accuracy standard. In other words, the rotation of the impeller directly reflects the unstable state of the main shaft operation. By observing the speed and movement state of the impeller, the specific problems of the main shaft operation can be judged. This intuitive display method allows operators to easily identify whether the main shaft has unstable operation problems.

[0004] However, this method may not be sensitive enough to slight unsteadiness of the spindle. If the small fluctuations in the spindle during operation do not generate enough contact force to drive the friction wheel with the shaft to rotate, even if there is a certain degree of imbalance, it may not be effectively detected. In addition, there may be nonlinear power transmission characteristics in the entire transmission chain from the spindle to the impeller. This means that the speed and motion state of the impeller do not always accurately reflect the actual operation of the spindle, especially under different load conditions.

[0005] Therefore, an optimized engine main shaft stability detection system and detection method thereof are expected. Summary of the invention

[0006] The present invention provides an engine main shaft stability detection system and a detection method thereof, aiming to partially or completely solve the above technical problems. The technical solution of the present invention is as follows:

[0007] On the one hand, the present invention application provides an engine main shaft stability detection system, which includes: a signal acquisition module, which is used to acquire the vibration signal of the detected engine main shaft under predetermined load conditions collected by a vibration sensor; a signal processing module, which is used to perform discrete wavelet transform on the vibration signal to obtain a vibration two-dimensional time-frequency diagram; an image segmentation module, which is used to perform image segmentation on the vibration two-dimensional time-frequency diagram to obtain a set of vibration two-dimensional local time-frequency blocks; a vibration local time-frequency feature extraction module, which is used to extract the vibration local time-frequency features of each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks to obtain a set of vibration local time-frequency semantic coding feature vectors; a feature aggregation module, which is used to perform feature distribution field dynamic aggregation on the set of vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features; an abnormality detection module, which is used to determine whether the engine main shaft has abnormal vibration under predetermined load conditions based on the vibration full-frequency domain significant clustering prototype coding features.

[0008] In particular, the image segmentation module is used to: perform grid segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks.

[0009] In particular, the vibration local time-frequency feature extraction module is used to: input each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks into a vibration local time-frequency feature extractor based on a void convolutional neural network model to obtain a set of vibration local time-frequency semantic encoding feature vectors.

[0010] In particular, the static energy factor calculation unit is used to:

[0011] Calculating the mean and variance of the vibration local time-frequency semantic coding feature vector to obtain the vibration local time-frequency semantic mean and the vibration local time-frequency semantic variance;

[0012] Calculating the positional subtraction between the vibration local time-frequency semantic coding feature vector and the vibration local time-frequency semantic mean, and calculating the fourth power of each position of the feature vector after the subtraction to obtain the vibration local time-frequency semantic difference modulation vector;

[0013] Calculating the expected value of the vibration local time-frequency semantic difference modulation vector to obtain the vibration local time-frequency semantic expected value;

[0014] The vibration local time-frequency semantics expected value is divided by the square of the vibration local time-frequency semantics variance, and the obtained value is input into a sigmoid function to obtain the vibration local time-frequency semantics static energy factor.

[0015] In particular, the characteristic significant polymerization unit includes:

[0016] A vibration local time-frequency semantic clustering initial center extraction subunit is used to extract the vibration local time-frequency semantic encoding feature vector corresponding to the largest vibration local time-frequency semantic static energy factor in the set of vibration local time-frequency semantic static energy factors as the vibration local time-frequency semantic clustering initial center vector;

[0017] A vibration local time-frequency semantic dynamic aggregation energy factor calculation subunit is used to calculate the vibration local time-frequency semantic dynamic aggregation energy factor of each vibration local time-frequency semantic encoding feature vector in the set of vibration local time-frequency semantic encoding feature vectors based on the spatial span between each vibration local time-frequency semantic encoding feature vector in the set of vibration local time-frequency semantic encoding feature vectors and the initial center vector of the vibration local time-frequency semantic clustering and the static energy factor of each vibration local time-frequency semantic encoding feature vector and the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering to obtain a set of vibration local time-frequency semantic dynamic aggregation energy factors;

[0018] A normalization subunit, used for normalizing the set of vibration local time-frequency semantic dynamic aggregation energy factors to obtain a set of normalized vibration local time-frequency semantic dynamic aggregation energy factors;

[0019] A masking subunit, configured to perform masking processing on the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors to obtain a set of vibration local time-frequency semantic dynamic aggregation weight factors;

[0020] The vibration full-frequency domain significant clustering representation subunit is used to obtain the vibration full-frequency domain significant clustering prototype coding vector by weighting the set of the vibration local time-frequency semantic coding feature vectors based on the vibration local time-frequency semantic dynamic aggregation weight factor.

[0021] Furthermore, the vibration local time-frequency semantic dynamic aggregation energy factor calculation subunit is used to:

[0022] Calculate the product of the first weighting parameter, the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering, and the static energy factor of the vibration local time-frequency semantic coding feature vector to obtain a first vibration local time-frequency semantic dynamic aggregation energy factor;

[0023] Calculating the product of the square of the spatial span between the vibration local time-frequency semantic coding feature vector and the vibration local time-frequency semantic clustering initial center vector and a second weighting parameter to obtain a second vibration local time-frequency semantic dynamic aggregation energy factor;

[0024] The first vibration local time-frequency semantic dynamic aggregation energy factor and the second vibration local time-frequency semantic dynamic aggregation energy factor are divided to obtain the vibration local time-frequency semantic dynamic aggregation energy factor.

[0025] In particular, the normalization subunit is used to:

[0026] Calculate an exponential function value with a natural base e as the base and the negative number of each vibration local time-frequency semantic dynamic aggregation energy factor in the set of vibration local time-frequency semantic dynamic aggregation energy factors as the power to obtain a set of vibration local time-frequency semantic dynamic aggregation energy exponential factors;

[0027] The set of vibration local time-frequency semantic dynamic aggregation energy index factors and the constant one are added in position, and the inverse of the value obtained by the positional addition is calculated to obtain the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors.

[0028] In particular, the mask sub-unit is used to: in response to each normalized vibration local time-frequency semantic dynamic aggregation energy factor in the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors being greater than a predetermined threshold, set the normalized vibration local time-frequency semantic dynamic aggregation energy factors greater than the predetermined threshold to the original value, and set the rest to zero, so as to obtain the set of vibration local time-frequency semantic dynamic aggregation weight factors.

[0029] In particular, the anomaly detection module is used to:

[0030] The vibration full-frequency domain significant clustering prototype coding vector is input into a classifier-based detection module to obtain a detection result, and the detection result is used to indicate whether abnormal vibration occurs in the engine main shaft under a predetermined load condition.

[0031] In another aspect, the present invention provides an engine main shaft stability detection method, which adopts an engine main shaft stability detection system as described in any one of the first aspects above, comprising:

[0032] Acquire a vibration signal of a detected engine main shaft under a predetermined load condition collected by a vibration sensor; perform discrete wavelet transform on the vibration signal to obtain a vibration two-dimensional time-frequency diagram; perform image segmentation on the vibration two-dimensional time-frequency diagram to obtain a set of vibration two-dimensional local time-frequency blocks; extract the vibration local time-frequency features of each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks to obtain a set of vibration local time-frequency semantic coding feature vectors; perform feature distribution field dynamic aggregation on the set of vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features; based on the vibration full-frequency domain significant clustering prototype coding features, determine whether the engine main shaft has abnormal vibration under a predetermined load condition.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) In the present invention application, the vibration signal of the main shaft of the detected engine under predetermined load conditions is subjected to discrete wavelet transform and local time-frequency analysis to dig out the tiny vibration characteristics of the main shaft of the detected engine. The local time-frequency characteristics of the vibration signal are dynamically aggregated based on the characteristic energy distribution to capture the key vibration mode of the main shaft of the engine, and the full-frequency domain characteristic perception of the vibration signal is achieved. Based on this, it is judged whether the main shaft of the engine has abnormal vibration under predetermined load conditions. It is able to capture slight changes in the stability of the main shaft of the detected engine during operation, thereby improving the accuracy of abnormality detection.

[0035] (2) In the present invention application, a bidirectional latent variable motif is constructed based on a short mesoscale sequence relative to a probability statistic value to perform mesoscale bidirectional migration, and a posteriori recovery is performed based on sequence constraints on the interactive information, thereby improving the convergence effect of the vibration full-frequency domain significant clustering prototype coding vector in the probability density domain and improving the accuracy of the detection result obtained by inputting a detection module based on a classifier. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and other purposes, features and advantages of the present invention will become more apparent by describing the embodiments of the present invention in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation on the present invention. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0037] Figure 1 A block diagram of an engine main shaft stability detection system according to an embodiment of the present invention;

[0038] Figure 2 It is a schematic diagram of the principle of the engine main shaft stability detection system according to an embodiment of the present invention;

[0039] Figure 3 A block diagram of a feature aggregation module in an engine main shaft stability detection system according to an embodiment of the present application;

[0040] Figure 4 The present invention is a flowchart of a method for detecting engine main shaft stability according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] Below, the exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0042] As shown in the present application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular, but also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list, and the method or device may also include other steps or elements.

[0043] Although the present invention application has made various references to certain modules in the system according to the embodiment of the present invention application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0044] The present invention application uses a flow chart to illustrate the operations performed by the system of the embodiment of the present invention application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0045] Below, the exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0046] On the one hand, if Figure 1 and Figure 2 As shown, the present invention provides an engine main shaft stability detection system 300, including: a signal acquisition module 310, used to obtain the vibration signal of the detected engine main shaft under predetermined load conditions collected by a vibration sensor; a signal processing module 320, used to perform discrete wavelet transform on the vibration signal to obtain a vibration two-dimensional time-frequency diagram; an image segmentation module 330, used to perform image segmentation on the vibration two-dimensional time-frequency diagram to obtain a set of vibration two-dimensional local time-frequency blocks; a vibration local time-frequency feature extraction module 340, used to extract the vibration local time-frequency features of each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks to obtain a set of vibration local time-frequency semantic coding feature vectors; a feature aggregation module 350, used to perform feature distribution field dynamic aggregation on the set of vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features; an abnormality detection module 360, used to determine whether the engine main shaft has abnormal vibration under predetermined load conditions based on the vibration full-frequency domain significant clustering prototype coding features.

[0047] In particular, the signal acquisition module 310 is used to obtain the vibration signal of the detected engine main shaft under predetermined load conditions collected by the vibration sensor. It should be understood that in the engine main shaft stability detection, the vibration signal is a key data source, which directly reflects the state of the main shaft during operation. The amplitude, frequency, phase and spectrum characteristics of the vibration signal are closely related to the stability of the main shaft. By performing multi-stage processing and analysis on the vibration signal, the slight abnormal vibration of the main shaft can be effectively captured and the accuracy of the detection can be improved. The vibration signal under the predetermined load conditions can more truly reflect the performance of the main shaft in actual work, and provide reliable data support for stability detection.

[0048] In particular, the signal processing module 320 is used to perform discrete wavelet transform on the vibration signal to obtain a two-dimensional vibration time-frequency diagram. It should be understood that there is a lot of noise in the engine operating environment, and the vibration sensor itself may introduce noise, which will interfere with the accuracy and reliability of the signal. In addition, the vibration signal collected by the vibration sensor is a time domain signal, which only provides time sequence information, while the vibration of the engine main shaft may contain multiple frequency components. Directly processing the time domain signal cannot extract the multi-scale characteristics of the vibration signal in different frequency domains, and cannot fully reflect the complexity of the vibration. Therefore, in the technical solution of the present invention, the vibration signal is subjected to discrete wavelet transform to obtain a two-dimensional vibration time-frequency diagram. In particular, the discrete wavelet transform has good denoising ability and can effectively suppress the noise in the vibration signal to retain useful vibration characteristics. Here, the time domain signal is converted into a time-frequency domain signal by discrete wavelet transform to decompose the vibration signal into sub-signals of multiple scales, each sub-signal reflects the vibration characteristics in different frequency ranges, and the generation of a two-dimensional time-frequency diagram can more comprehensively describe the characteristics of the vibration signal, thereby capturing the local characteristics of the vibration signal at different time scales, which is helpful to identify small abnormal vibrations.

[0049] In particular, the image segmentation module 330 is used to perform image segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks.

[0050] In the technical solution of the present invention application, it is considered that the vibration two-dimensional time-frequency diagram contains the vibration information of the engine main shaft at different times and frequencies. Therefore, the vibration two-dimensional time-frequency diagram is grid-cut to divide the vibration two-dimensional time-frequency diagram into multiple local areas, and each area represents a local time-frequency block. Different local time-frequency blocks can reflect vibration characteristics in different time scales and frequency ranges. In this way, the vibration characteristics of each local area can be analyzed more carefully, and tiny abnormal vibrations can be captured to more accurately locate the abnormal area in the vibration signal and improve the accuracy of abnormality detection.

[0051] In particular, the vibration local time-frequency feature extraction module 340 is used to extract the vibration local time-frequency features of each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks to obtain a set of vibration local time-frequency semantic coding feature vectors.

[0052] Since each local time-frequency block represents vibration information within a specific time and frequency range, different local time-frequency blocks contain different vibration features. By processing each vibration two-dimensional local time-frequency block in the set of the vibration two-dimensional local time-frequency block, the vibration features of each local area can be analyzed more carefully and tiny abnormal vibrations can be captured. Compared with the traditional convolutional neural network (CNN), the hole convolution can maintain a high spatial resolution without increasing the number of parameters. This is very important for capturing fine-grained features in vibration signals. Therefore, in the technical solution of the present invention, each vibration two-dimensional local time-frequency block in the set of the vibration two-dimensional local time-frequency block is input into a vibration local time-frequency feature extractor based on a hole convolutional neural network model, so as to mine the vibration features in each vibration two-dimensional local time-frequency block, so as to obtain a set of vibration local time-frequency semantic encoding feature vectors, so as to more comprehensively understand the complexity of the vibration signal.

[0053] In particular, the feature aggregation module 350 is used to perform dynamic feature distribution field aggregation on the set of vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features.

[0054] It should be understood that each vibration local time-frequency semantic coding feature vector in the set of vibration local time-frequency semantic coding feature vectors respectively contains the vibration local time-frequency semantic features of different frequency components extracted from the original vibration signal, and the vibration local time-frequency semantic features of each component serve as a vibration local time-frequency semantic particle. When participating in the subsequent vibration local time-frequency feature semantic aggregation process to obtain a more sufficient and accurate vibration full-frequency domain significant representation, these different vibration local time-frequency semantic particles will have different importance and contribution to the subsequent detection tasks.

[0055] Based on this, in the technical solution of the present invention, the set of vibration local time-frequency semantic coding feature vectors is further dynamically aggregated to obtain a vibration full-frequency domain significant aggregation representation. In particular, here, the dynamic aggregation process is modeled based on the interaction between each vibration local time-frequency semantic particle through a concept similar to a force field. This force field not only takes into account the direct spatial relationship between each vibration local time-frequency semantic (such as Euclidean distance), but also combines the properties of each vibration local time-frequency semantic itself (such as static energy factor) and the overall semantic structure of the entire vibration local time-frequency semantic set. In this way, the importance of each vibration local time-frequency semantic feature in the vibration local time-frequency feature semantic aggregation process can be dynamically adjusted, and the information can be integrated into a more compact and representative vibration full-frequency domain significant aggregation representation according to a specific aggregation strategy and control task, that is, a vibration full-frequency domain significant clustering prototype encoding feature.

[0056] In a specific example of the present invention, Figure 3 As shown, the feature aggregation module 350 includes: a static energy factor calculation unit 351, which is used to calculate the vibration local time-frequency semantic static energy factor of each vibration local time-frequency semantic coding feature vector in the set of vibration local time-frequency semantic coding feature vectors to obtain a set of vibration local time-frequency semantic static energy factors; a feature significant aggregation unit 352, which is used to perform feature significant aggregation based on the imitation field on the set of vibration local time-frequency semantic coding feature vectors based on the set of vibration local time-frequency semantic static energy factors to obtain the vibration full-frequency domain significant clustering prototype coding feature.

[0057] Specifically, the static energy factor calculation unit 351 is used to calculate the vibration local time-frequency semantic static energy factor of each vibration local time-frequency semantic coding feature vector in the set of vibration local time-frequency semantic coding feature vectors to obtain a set of vibration local time-frequency semantic static energy factors.

[0058] It should be understood that for each vibration local time-frequency semantic encoding feature vector in the set of the vibration local time-frequency semantic encoding feature vector, its stability or importance in the feature space is evaluated based on a predefined metric function, and a numerical representation is assigned, namely, the vibration local time-frequency semantic static energy factor. In this way, the status and importance of each vibration local time-frequency semantic feature in its data distribution and the overall vibration local time-frequency semantics can be quantified, providing a basis for the subsequent selection of the initial clustering center. At the same time, by identifying those vibration local time-frequency semantic feature points that are relatively isolated or have a high local influence, it is helpful to discover potential data cluster cores. In this way, the uniqueness and influence of each vibration local time-frequency semantic in the feature space can be quantified, providing a basis for the subsequent aggregation process.

[0059] In particular, the specific process of calculating the vibration local time-frequency semantic static energy factor of each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector includes: calculating the mean and variance of the vibration local time-frequency semantic coding feature vector to obtain the vibration local time-frequency semantic mean and the vibration local time-frequency semantic variance; calculating the positional subtraction of the vibration local time-frequency semantic coding feature vector and the vibration local time-frequency semantic mean, and calculating the fourth power of each position of the feature vector after the subtraction to obtain the vibration local time-frequency semantic difference modulation vector; calculating the expected value of the vibration local time-frequency semantic difference modulation vector to obtain the vibration local time-frequency semantic expected value; dividing the vibration local time-frequency semantic expected value by the square of the vibration local time-frequency semantic variance, and inputting the obtained value into the sigmoid function to obtain the vibration local time-frequency semantic static energy factor.

[0060] Specifically, the feature significant aggregation unit 352 is used to perform feature significant aggregation based on the imitation field on the set of vibration local time-frequency semantic coding feature vectors based on the set of vibration local time-frequency semantic static energy factors to obtain the vibration full-frequency domain significant clustering prototype coding features.

[0061] In the technical solution of the present invention, first, the vibration local time-frequency semantic coding feature vector corresponding to the largest vibration local time-frequency semantic static energy factor in the set of vibration local time-frequency semantic static energy factors is selected as the initial center vector of vibration local time-frequency semantic clustering. That is, a reliable starting point is determined to perform iterative clustering operations. This point should represent the most representative part of the vibration local time-frequency semantic set, which can serve as the basis for subsequent clustering to ensure the rationality of the clustering results. This enhances the adaptability of the clustering process to real scenes and stability detection tasks, and reduces the adverse effects caused by random initialization. Next, based on the spatial span between each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector and the initial center vector of the vibration local time-frequency semantic clustering, as well as the static energy factor of each vibration local time-frequency semantic coding feature vector and the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering, the vibration local time-frequency semantic dynamic aggregation energy factor of each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector is calculated to obtain a set of vibration local time-frequency semantic dynamic aggregation energy factors; in particular, the dynamic aggregation energy factor not only takes into account the proximity of the vibration local time-frequency semantics in position, but also includes the importance difference of the intrinsic attributes of the vibration local time-frequency semantics presented at each point, aiming to more accurately capture the semantic interaction between vibration local time-frequency, so as to measure the semantic importance and contribution of each vibration local time-frequency semantic in the overall set and subsequent feature aggregation process. Next, the set of the vibration local time-frequency semantic dynamic aggregation energy factors is normalized to obtain a set of normalized vibration local time-frequency semantic dynamic aggregation energy factors, and the set of the normalized vibration local time-frequency semantic dynamic aggregation energy factors is masked to obtain a set of vibration local time-frequency semantic dynamic aggregation weight factors; that is, the set of the vibration local time-frequency semantic dynamic aggregation energy factors is input into the gated mask unit to dynamically adjust the importance weight of each vibration local time-frequency semantic dynamic aggregation energy factor. In this way, by nonlinearly transforming and screening the original energy factors, the model's attention to key vibration local time-frequency semantic information is enhanced, while suppressing insignificant parts, thereby giving higher weight to key answer semantics related to subsequent tasks during the aggregation process. Finally, based on the vibration local time-frequency semantic dynamic aggregation weight factors, the weighted sum of the set of vibration local time-frequency semantic encoding feature vectors is performed to obtain a vibration full-frequency domain significant clustering prototype encoding vector.Here, the set of dynamic aggregation weight factors based on the local time-frequency semantics of the vibration is used as the weighting coefficient to calculate the weighted sum of the set of local time-frequency semantic coding feature vectors of the vibration, so as to combine all the local time-frequency semantic feature information of the vibration, and at the same time, appropriately emphasize or weaken the feature expression according to their respective semantic importance, so as to generate a more concise and representative vibration full-frequency domain significant clustering prototype coding vector.

[0062] Among them, the specific process of calculating the vibration local time-frequency semantic dynamic aggregation energy factor of each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector includes: calculating the product of the first weighting parameter, the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering and the static energy factor of the vibration local time-frequency semantic coding feature vector to obtain the first vibration local time-frequency semantic dynamic aggregation energy factor; calculating the product of the square of the spatial span between the vibration local time-frequency semantic coding feature vector and the initial center vector of the vibration local time-frequency semantic clustering and the second weighting parameter to obtain the second vibration local time-frequency semantic dynamic aggregation energy factor; dividing the first vibration local time-frequency semantic dynamic aggregation energy factor and the second vibration local time-frequency semantic dynamic aggregation energy factor to obtain the vibration local time-frequency semantic dynamic aggregation energy factor.

[0063] More specifically, the specific process of normalizing the set of vibration local time-frequency semantic dynamic aggregation energy factors includes: calculating the exponential function value with the natural base e as the base and the negative number of each vibration local time-frequency semantic dynamic aggregation energy factor in the set of vibration local time-frequency semantic dynamic aggregation energy factors as the power to obtain a set of vibration local time-frequency semantic dynamic aggregation energy exponential factors; calculating the positional addition of the set of vibration local time-frequency semantic dynamic aggregation energy exponential factors and a constant one, and calculating the inverse of the value obtained by the positional addition to obtain the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors.

[0064] More specifically, the specific process of masking the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors includes: in response to each normalized vibration local time-frequency semantic dynamic aggregation energy factor in the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors being greater than a predetermined threshold, setting the normalized vibration local time-frequency semantic dynamic aggregation energy factors greater than the predetermined threshold to the original value, and setting the rest to zero, so as to obtain the set of vibration local time-frequency semantic dynamic aggregation weight factors.

[0065] In summary, in the above embodiment, the set of the vibration local time-frequency semantic coding feature vectors is dynamically aggregated in the feature distribution field to obtain the vibration full-frequency domain significant clustering prototype coding features, including: the set of the vibration local time-frequency semantic coding feature vectors is dynamically aggregated in the feature distribution field to obtain the vibration full-frequency domain significant clustering prototype coding features according to the following feature distribution dynamic aggregation formula; wherein the feature distribution dynamic aggregation formula is:

[0066] X={x1,x2,...,x i ,...,x n}

[0067]

[0068] x c =x m

[0069] in,

[0070]

[0071] w si =mask(w i )

[0072]

[0073] Where X is the set of the vibration local time-frequency semantic coding feature vectors, x1, x2, ..., x i ,...,x n are respectively the 1st, 2nd, ..., ith, ..., nth vibration local time-frequency semantic coding feature vectors in the set of the vibration local time-frequency semantic coding feature vectors, is the eigenvalue of each position of the ith vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector, j is a positive integer, μ i and σ i 2 They are v i The mean and variance of , E(·) is the expected value, sigmoid(·) is the sigmoid function, is the vibration local time-frequency semantic static energy factor corresponding to the i-th vibration local time-frequency semantic encoding feature vector, arg max j (·) is the index corresponding to the maximum value, m is the index of the largest vibration local time-frequency semantic static energy factor, x c is the initial center vector of the vibration local time-frequency semantic clustering, a and b are the first weighting parameter and the second weighting parameter respectively, is xm The corresponding vibration local time-frequency semantic static energy factor, Count(x i →x m ) represents x i With x m The number of eigenvectors between is the vibration local time-frequency semantic dynamic aggregation energy factor corresponding to the i-th vibration local time-frequency semantic encoding feature vector, w i is the normalized vibration local time-frequency semantic dynamic aggregation energy factor corresponding to the i-th vibration local time-frequency semantic encoding feature vector, mask(·) is the masking process, θ is the predetermined threshold, w si is the vibration local time-frequency semantic dynamic aggregation weight factor corresponding to the i-th vibration local time-frequency semantic coding feature vector, n is the number of vectors in the set of vibration local time-frequency semantic coding feature vectors, x f It is the prototype encoding feature of significant clustering of vibration in the full frequency domain.

[0074] In particular, the abnormality detection module 360 ​​is used to determine whether the engine main shaft has abnormal vibration under predetermined load conditions based on the vibration full-frequency domain significant clustering prototype coding features. In the technical solution of the present invention application, the vibration full-frequency domain significant clustering prototype coding vector is input into a detection module based on a classifier to obtain a detection result, and the detection result is used to indicate whether the engine main shaft has abnormal vibration under predetermined load conditions. That is, the classifier model is used to classify and judge the vibration full-frequency domain significant clustering prototype coding vector. Through a large amount of training, the classifier can effectively learn the abnormal vibration information contained in the vibration full-frequency domain significant clustering prototype coding vector, and then accurately distinguish whether the engine main shaft has abnormal vibration under predetermined load conditions, and output the corresponding detection result.

[0075] Preferably, based on the original content differences in the frequency domain of each vibration two-dimensional local time-frequency block in the set of the vibration two-dimensional local time-frequency block, after inputting into the vibration local time-frequency feature extractor based on the void convolutional neural network model, each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vector will also have a difference in coding feature distribution, resulting in that when the feature distribution field based on the imitation force field is dynamically aggregated, the imitation force field feature distribution under the difference in coding feature distribution will have different field dynamic aggregation modes, causing the feature distribution of the vibration full-frequency domain significant clustering prototype coding vector to have feature distribution diversity in the full-frequency domain. Therefore, when the vibration full-frequency domain significant clustering prototype coding vector is input into the detection module based on the classifier for classification probability regression, it will affect the accuracy of the classification result.

[0076] Based on this, when the vibration full-frequency domain significant clustering prototype coding vector is input into the detection module based on the classifier for classification probability regression, the vibration full-frequency domain significant clustering prototype coding vector is first optimized, and the optimization process specifically includes:

[0077] Determine the mean and variance of the feature set composed of the eigenvalues ​​of the vibration full-frequency domain significant clustering prototype coding vector, and divide the mean by the variance to obtain the vibration full-frequency domain significant clustering prototype coding probability weight value:

[0078]

[0079] Among them, μ and σ 2 They respectively represent the mean and variance of the feature set composed of the eigenvalues ​​of the vibration full-frequency domain significant clustering prototype coding vector, and θ represents the probability weight value of the vibration full-frequency domain significant clustering prototype coding;

[0080] The vibration full-frequency domain significant clustering prototype coding vector is point-multiplied by the inverse of the maximum eigenvalue of the vibration full-frequency domain significant clustering prototype coding vector to obtain the vibration full-frequency domain significant clustering prototype coding probability constraint vector:

[0081] V1=V⊙v max -1

[0082] Among them, v max -1 represents the inverse of the maximum eigenvalue of the vibration full-frequency domain significant clustering prototype coding vector, V represents the vibration full-frequency domain significant clustering prototype coding vector, ⊙ represents the position point multiplication, and V1 represents the vibration full-frequency domain significant clustering prototype coding probability constraint vector;

[0083] The vibration full-frequency domain significant clustering prototype coding probability constraint vector is added to the vibration full-frequency domain significant clustering prototype coding probability weight value, and the logarithmic value with base 2 is calculated to obtain the vibration full-frequency domain significant clustering prototype coding information interaction vector:

[0084]

[0085] Among them, V1 represents the vibration full-frequency domain significant clustering prototype coding probability constraint vector, θ represents the vibration full-frequency domain significant clustering prototype coding probability weight value, represents the addition by position point, V2 represents the interaction vector of the prototype encoding information of the vibration full frequency domain significant clustering;

[0086] After subtracting the vibration full-frequency domain significant clustering prototype coding probability constraint vector from the vibration full-frequency domain significant clustering prototype coding probability weight value, the inverse of each eigenvalue is calculated to obtain the vibration full-frequency domain significant clustering prototype coding sequence constraint vector:

[0087]

[0088] Among them, V1 represents the vibration full-frequency domain significant clustering prototype coding probability constraint vector, θ represents the vibration full-frequency domain significant clustering prototype coding probability weight value, represents point-by-point subtraction, V3 represents the vibration full-frequency domain significant clustering prototype coding sequence constraint vector;

[0089] Finally, the vibration full-frequency domain significant clustering prototype coding information interaction vector is added to the vibration full-frequency domain significant clustering prototype coding sequence constraint vector to obtain an optimized vibration full-frequency domain significant clustering prototype coding vector.

[0090] That is, considering that when the weight matrix during classification probability regression acts on the vibration full-frequency domain significant clustering prototype coding vector, the distribution diversity of the vibration full-frequency domain significant clustering prototype coding vector causes the uncertainty of the weight stimulus parameters of the weight matrix, which makes the posterior probability density inference of the vibration full-frequency domain significant clustering prototype coding vector through the weight matrix missing, which will affect the accuracy of the probability regression result. The above optimization process simulates the mesoscale interaction structure under probability constraints between the eigenvalue scale and the eigenvector scale of the vibration full-frequency domain significant clustering prototype coding vector through the probability statistical characteristics of the vibration full-frequency domain significant clustering prototype coding vector, so as to construct a bidirectional hidden variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and perform posterior recovery based on the sequence constraint on the interactive information, so as to improve the convergence effect of the vibration full-frequency domain significant clustering prototype coding vector in the probability density domain, and improve the accuracy of the detection result obtained by its input based on the detection module of the classifier.

[0091] As described above, the engine main shaft stability detection system 300 according to the present invention can be implemented in various wireless terminals, such as a server with an engine main shaft stability detection algorithm. In one possible implementation, the engine main shaft stability detection system 300 according to the embodiment of the present invention can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the engine main shaft stability detection system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the engine main shaft stability detection system 300 can also be one of the many hardware modules of the wireless terminal.

[0092] Alternatively, in another example, the engine main shaft stability detection system 300 and the wireless terminal may also be separate devices, and the engine main shaft stability detection system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0093] On the other hand, Figure 4 As shown, the present invention provides an engine main shaft stability detection method, which adopts any one of the engine main shaft stability detection systems 300 in the first aspect, including:

[0094] S1, obtaining a vibration signal of a main shaft of a detected engine under a predetermined load condition collected by a vibration sensor;

[0095] S2, performing discrete wavelet transform on the vibration signal to obtain a vibration two-dimensional time-frequency diagram;

[0096] S3, performing image segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks;

[0097] S4, extracting the vibration local time-frequency features of each vibration two-dimensional local time-frequency image block in the set of vibration two-dimensional local time-frequency image blocks to obtain a set of vibration local time-frequency semantic coding feature vectors;

[0098] S5, performing dynamic aggregation of feature distribution field on the set of vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features;

[0099] S6, determining whether abnormal vibration occurs on the main shaft of the engine under a predetermined load condition based on the vibration full-frequency domain significant clustering prototype coding features.

[0100] In summary, the engine main shaft stability detection method according to the embodiment of the present invention is explained. In the process of detecting the stability of the engine main shaft, the vibration signal of the detected engine main shaft under predetermined load conditions is subjected to discrete wavelet transform and local time-frequency analysis to dig out the tiny vibration characteristics of the detected engine main shaft. Furthermore, the local time-frequency characteristics of the vibration signal are dynamically aggregated based on the characteristic energy distribution to capture the key vibration mode of the engine main shaft, realize the full-frequency domain feature perception of the vibration signal, and based on this, judge whether the engine main shaft has abnormal vibration under predetermined load conditions. In this way, slight changes in the stability of the detected engine main shaft during operation can be captured, which improves the accuracy of abnormality detection.

[0101] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An engine main shaft stability detection system, characterized in that: include: A signal acquisition module, used to acquire a vibration signal of a main shaft of a detected engine under a predetermined load condition collected by a vibration sensor; A signal processing module, used for performing discrete wavelet transform on the vibration signal to obtain a vibration two-dimensional time-frequency diagram; An image segmentation module, used for performing image segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks; A vibration local time-frequency feature extraction module, used to extract the vibration local time-frequency features of each vibration two-dimensional local time-frequency image block in the set of vibration two-dimensional local time-frequency image blocks to obtain a set of vibration local time-frequency semantic coding feature vectors; A feature aggregation module, used for performing dynamic aggregation of the set of the vibration local time-frequency semantic coding feature vectors in the feature distribution field to obtain vibration full-frequency domain significant clustering prototype coding features, including: a static energy factor calculation unit, used for calculating the vibration local time-frequency semantic static energy factor of each vibration local time-frequency semantic coding feature vector in the set of the vibration local time-frequency semantic coding feature vectors to obtain a set of vibration local time-frequency semantic static energy factors; a feature significant aggregation unit, used for performing feature significant aggregation based on the imitation force field on the set of the vibration local time-frequency semantic coding feature vectors based on the set of the vibration local time-frequency semantic static energy factors to obtain the vibration full-frequency domain significant clustering prototype coding features; The abnormality detection module is used to determine whether the engine main shaft has abnormal vibration under predetermined load conditions based on the vibration full-frequency domain significant clustering prototype coding characteristics.

2. The engine main shaft stability detection system according to claim 1, characterized in that: The image segmentation module is used to perform grid segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks.

3. The engine main shaft stability detection system according to claim 2, characterized in that: The vibration local time-frequency feature extraction module is used to: input each vibration two-dimensional local time-frequency block in the set of vibration two-dimensional local time-frequency blocks into a vibration local time-frequency feature extractor based on a void convolutional neural network model to obtain a set of vibration local time-frequency semantic coding feature vectors.

4. The engine main shaft stability detection system according to claim 3, characterized in that: The static energy factor calculation unit is used to: Calculating the mean and variance of the vibration local time-frequency semantic coding feature vector to obtain the vibration local time-frequency semantic mean and the vibration local time-frequency semantic variance; Calculating the positional subtraction between the vibration local time-frequency semantic coding feature vector and the vibration local time-frequency semantic mean, and calculating the fourth power of each position of the feature vector after the subtraction to obtain the vibration local time-frequency semantic difference modulation vector; Calculating the expected value of the vibration local time-frequency semantic difference modulation vector to obtain the vibration local time-frequency semantic expected value; The vibration local time-frequency semantics expected value is divided by the square of the vibration local time-frequency semantics variance, and the obtained value is input into a sigmoid function to obtain the vibration local time-frequency semantics static energy factor.

5. The engine main shaft stability detection system according to claim 4, characterized in that: The characteristic significant polymerization unit includes: A vibration local time-frequency semantic clustering initial center extraction subunit is used to extract the vibration local time-frequency semantic encoding feature vector corresponding to the largest vibration local time-frequency semantic static energy factor in the set of vibration local time-frequency semantic static energy factors as the vibration local time-frequency semantic clustering initial center vector; A vibration local time-frequency semantic dynamic aggregation energy factor calculation subunit is used to calculate the vibration local time-frequency semantic dynamic aggregation energy factor of each vibration local time-frequency semantic encoding feature vector in the set of vibration local time-frequency semantic encoding feature vectors based on the spatial span between each vibration local time-frequency semantic encoding feature vector in the set of vibration local time-frequency semantic encoding feature vectors and the initial center vector of the vibration local time-frequency semantic clustering and the static energy factor of each vibration local time-frequency semantic encoding feature vector and the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering to obtain a set of vibration local time-frequency semantic dynamic aggregation energy factors; A normalization subunit, used for normalizing the set of vibration local time-frequency semantic dynamic aggregation energy factors to obtain a set of normalized vibration local time-frequency semantic dynamic aggregation energy factors; A masking subunit, configured to perform masking processing on the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors to obtain a set of vibration local time-frequency semantic dynamic aggregation weight factors; The vibration full-frequency domain significant clustering representation subunit is used to obtain the vibration full-frequency domain significant clustering prototype coding vector by weighting the set of the vibration local time-frequency semantic coding feature vectors based on the vibration local time-frequency semantic dynamic aggregation weight factor.

6. The engine main shaft stability detection system according to claim 5, characterized in that: The vibration local time-frequency semantic dynamic aggregation energy factor calculation subunit is used to: Calculate the product of the first weighting parameter, the static energy factor of the initial center vector of the vibration local time-frequency semantic clustering, and the static energy factor of the vibration local time-frequency semantic coding feature vector to obtain a first vibration local time-frequency semantic dynamic aggregation energy factor; Calculating the product of the square of the spatial span between the vibration local time-frequency semantic coding feature vector and the vibration local time-frequency semantic clustering initial center vector and a second weighting parameter to obtain a second vibration local time-frequency semantic dynamic aggregation energy factor; The first vibration local time-frequency semantic dynamic aggregation energy factor and the second vibration local time-frequency semantic dynamic aggregation energy factor are divided to obtain the vibration local time-frequency semantic dynamic aggregation energy factor.

7. The engine main shaft stability detection system according to claim 6, characterized in that: The normalization subunit is used for: Calculate an exponential function value with a natural base e as the base and the negative number of each vibration local time-frequency semantic dynamic aggregation energy factor in the set of vibration local time-frequency semantic dynamic aggregation energy factors as the power to obtain a set of vibration local time-frequency semantic dynamic aggregation energy exponential factors; The set of vibration local time-frequency semantic dynamic aggregation energy index factors and the constant one are added in position, and the inverse of the value obtained by the positional addition is calculated to obtain the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors.

8. The engine main shaft stability detection system according to claim 7, characterized in that: The mask sub-unit is used to: in response to each normalized vibration local time-frequency semantic dynamic aggregation energy factor in the set of normalized vibration local time-frequency semantic dynamic aggregation energy factors being greater than a predetermined threshold, set the normalized vibration local time-frequency semantic dynamic aggregation energy factors greater than the predetermined threshold to the original value, and set the rest to zero, so as to obtain the set of vibration local time-frequency semantic dynamic aggregation weight factors.

9. The engine main shaft stability detection system according to claim 8, characterized in that: The anomaly detection module is used to: The vibration full-frequency domain significant clustering prototype coding vector is input into a classifier-based detection module to obtain a detection result, and the detection result is used to indicate whether abnormal vibration occurs in the engine main shaft under a predetermined load condition.

10. A method for detecting the stability of an engine main shaft, using an engine main shaft stability detection system according to any one of claims 1 to 9, characterized in that: include: Acquiring a vibration signal of a main shaft of a detected engine under a predetermined load condition collected by a vibration sensor; Performing discrete wavelet transform on the vibration signal to obtain a two-dimensional vibration time-frequency diagram; Performing image segmentation on the vibration two-dimensional time-frequency image to obtain a set of vibration two-dimensional local time-frequency image blocks; Extracting vibration local time-frequency features of each vibration two-dimensional local time-frequency image block in the set of vibration two-dimensional local time-frequency image blocks to obtain a set of vibration local time-frequency semantic coding feature vectors; Performing dynamic aggregation in the feature distribution field on the set of the vibration local time-frequency semantic coding feature vectors to obtain vibration full-frequency domain significant clustering prototype coding features; Based on the vibration full-frequency domain significant clustering prototype coding features, it is determined whether the engine main shaft has abnormal vibration under predetermined load conditions.

Images