A rotating machine centering calibration method and device, electronic equipment and storage medium

Abstract

The application provides a rotating machinery centering calibration method and device, electronic equipment and storage medium, the method comprises the following steps: setting multiple working condition parameters and collecting multiple sensor data, extracting feature vectors after preprocessing the data; constructing a mapping table, fusing DS evidence theory distribution feature weight, constructing a knowledge base, and identifying the feature vector and physical parameter of the optimal centering state; obtaining real-time multi-sensor data of the equipment to be diagnosed, reading the feature weight from the knowledge base according to the current working condition, finding the matching state in the knowledge base through a dynamic weighted similarity matching algorithm, and obtaining the equivalent physical state through reverse reasoning; calculating the deviation of the equivalent physical state and the physical parameter of the optimal centering state, generating a calibration adjustment instruction, re-collecting data and verifying the feature matching degree, and if the preset threshold is not reached, iteratively fine-tuning and updating the knowledge base. The application improves the efficiency and reliability of rotating machinery centering calibration.

Description

Technical Field

[0001] This invention relates to the field of intelligent fault diagnosis technology, specifically to a method, apparatus, electronic device, and storage medium for aligning and calibrating rotating machinery. Background Technology

[0002] Rotating machinery is a critical piece of equipment in industries such as energy, chemical engineering, and manufacturing. The alignment accuracy of its rotor shaft system is a core factor affecting the stability, efficiency, and lifespan of the equipment. Misalignment faults can cause abnormal vibrations, accelerate component wear, and even lead to catastrophic shutdowns. Therefore, accurate diagnosis and rapid calibration of the shaft system alignment are of great engineering significance.

[0003] Currently, with the advancement of sensing technology and intelligent algorithms, condition monitoring and diagnosis methods based on multi-sensor information fusion have become the mainstream trend in industrial operation and maintenance. By simultaneously acquiring multi-source signals such as vibration and displacement, and extracting features with clear physical meaning in the frequency domain, time domain, and trajectory, misalignment fault modes of equipment can be identified more comprehensively and reliably. Machine learning models such as Deep Belief Networks (DBN) and Support Vector Machines (SVM) are also being applied to learn the complex mapping relationship between faults and health states from high-dimensional features, improving the qualitative identification capability of shaft misalignment faults. These methods significantly reduce reliance on human experience, realizing a shift from "experience-driven" to "data-driven" approaches.

[0004] However, existing intelligent diagnostic methods still face a key bottleneck in practical applications: the disconnect between diagnostic results and physical execution. Most methods typically output qualitative or semi-quantitative category labels, which cannot be directly converted into quantifiable physical adjustment parameters to guide on-site maintenance. Furthermore, the model's decision-making process lacks interpretability and is difficult to validate in a closed-loop manner with the physical mechanisms of the mechanical system. This leads to insufficient trust in the diagnostic results among maintenance personnel, ultimately requiring repeated trial-and-error calibration based on expert experience. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, apparatus, electronic device and storage medium for aligning and calibrating rotating machinery, so as to solve the technical problem that the diagnostic results of existing intelligent diagnostic methods are disconnected from physical execution and difficult to form a closed loop verification with the physical mechanism of the mechanical system.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for centering and calibrating rotating machinery, comprising: Multiple operating condition parameters are set on the benchmark platform, and signal data from vibration sensors, eddy current sensors, and speed sensors are collected under each operating condition. After preprocessing the signal data, multidimensional physical feature vectors are extracted. A mapping table between working conditions and feature importance is constructed based on multidimensional physical feature vectors. Feature weights are allocated according to the mapping table and DS evidence theory. A knowledge base of centering state and feature mapping is constructed, and the feature vector and physical parameters of the optimal centering state are identified in the knowledge base. The system acquires real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed. Based on the current operating conditions, it reads feature weights from the knowledge base, searches for matching pairs in the knowledge base using a dynamic weighted similarity matching algorithm, and inversely infers the equivalent physical state of the device to be diagnosed based on the physical parameters in the matching samples. Calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate calibration adjustment instructions for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the knowledge base.

[0007] In one possible implementation, the preprocessing includes adaptive anti-interference preprocessing, which includes performing wavelet threshold denoising on the acquired raw signal and performing adaptive notch filtering to filter out power frequency and harmonic interference.

[0008] In one possible implementation, the multidimensional physical feature vector includes vibration signal frequency domain features, eddy current displacement signal trajectory features, and multi-signal correlation features.

[0009] In one possible implementation, the working condition and feature importance mapping table is based on historical full-condition operation samples, fault labeling datasets, and objective quantitative indicators to determine the importance priority of each physical feature, thus forming the working condition and feature importance mapping table.

[0010] In one possible implementation, the allocation of feature weights based on the DS evidence theory fused from the mapping table includes: Based on the DS evidence theory, an identification framework is constructed for each physical feature in the physical feature vector to identify the importance of the physical feature to state identification. The basic probability allocation function is initialized based on the mapping table between working conditions and feature importance and the real-time quality of features; The Dempster combination rule is used to fuse the basic probability assignment functions of multiple features to obtain a joint basic probability assignment function; The support of each physical feature is calculated based on the joint basic probability assignment function, and then normalized to obtain the dynamic weights.

[0011] In one possible implementation, the construction of the midpoint state and feature mapping knowledge base includes: Construct a three-layer structured storage architecture comprising a working condition layer, a feature layer, and a physical state layer; The operating parameters, physical feature vectors, and optimal physical parameter set of the optimal alignment state are stored in a three-layer structured storage architecture. The misalignment fault conditions are quantitatively set using a laser alignment instrument. Data collection, preprocessing, and feature extraction are performed repeatedly on the misalignment fault conditions to obtain multiple sets of condition parameter-physical feature vector-physical parameter data pairs, which are then stored in batches into a three-layer structured storage architecture. Based on the working condition and feature importance mapping table, the dynamic weight of each sample data in the three-layer structured storage architecture is calculated by DS evidence theory and stored in the working condition layer. At the same time, the quantitative mapping relationship between feature values ​​and physical parameters is calibrated by linear regression and nonlinear fitting, thus completing the construction of the centering state and feature mapping knowledge base.

[0012] In one possible implementation, the dynamic weighted similarity matching algorithm adaptively selects the corresponding similarity algorithm based on the correlation between physical features, specifically as follows: For strong correlations between physical features, the Mahalanobis distance algorithm is used. For weak correlations between physical features, the weighted Euclidean distance algorithm is used; If the physical features are approximately independent, the cosine similarity algorithm is used.

[0013] On the other hand, the present invention also provides a rotating machinery centering calibration device, comprising: The data acquisition and preprocessing module is used to set multiple working condition parameters on the benchmark platform and acquire signal data from vibration sensors, eddy current sensors, and speed sensors under each working condition. After preprocessing the signal data, multi-dimensional physical feature vectors are extracted. The knowledge base construction module is used to construct a mapping table between working conditions and feature importance based on multi-dimensional physical feature vectors, allocate feature weights according to the mapping table and DS evidence theory, construct a knowledge base of centering state and feature mapping, and identify the feature vector and physical parameters of the optimal centering state in the knowledge base. The reverse reasoning module is used to acquire real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed, read feature weights from the knowledge base according to the current working conditions, search for matching pairs in the knowledge base through a dynamic weighted similarity matching algorithm, and obtain the equivalent physical state of the device to be diagnosed by reverse reasoning based on the physical parameters in the matching samples. The calibration module is used to calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate calibration adjustment instructions for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and iteratively fine-tune and update the knowledge base if the feature matching degree does not reach the preset threshold.

[0014] In a second aspect, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the rotating machinery alignment and calibration method described in any of the above implementations.

[0015] Thirdly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps in the rotating machinery alignment and calibration method described in any of the above implementations.

[0016] The beneficial effects of this invention are as follows: The rotating machinery alignment calibration method provided by this invention, by constructing an alignment state and feature mapping knowledge base and integrating DS evidence theory to allocate feature weights, can reverse-engineer the signal data of real-time acquired vibration sensors, eddy current sensors, and speed sensors into the equivalent physical state of the equipment to be diagnosed. It then calculates the physical deviation from the optimal alignment state and automatically generates calibration adjustment commands for adjusting the alignment state, thereby achieving a leap from qualitative identification to quantitative diagnosis of misalignment faults and solving the problem of disconnect between diagnostic results and physical execution. By allocating feature weights through DS evidence theory, the reasoning process becomes interpretable and forms a closed-loop verification with the mechanical physical mechanism. Furthermore, by re-acquiring data and verifying the feature matching degree, if the feature matching degree does not reach a preset threshold, iterative fine-tuning and updating of the alignment state and feature mapping knowledge base are performed, forming an adaptive intelligent calibration process that significantly improves the efficiency and reliability of alignment calibration. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart of an embodiment of the rotational machinery alignment and calibration method provided by the present invention; Figure 2 This is a schematic diagram illustrating the workflow of the rotating machinery alignment and calibration method provided by the present invention; Figure 3 This is a schematic diagram of the closed-loop iterative process of the centering and calibration method for rotating machinery provided by the present invention; Figure 4 A schematic diagram of an embodiment of the centering and calibration device for rotating machinery provided by the present invention; Figure 5 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] This invention provides a method, apparatus, electronic device, and storage medium for aligning and calibrating rotating machinery, which are described below.

[0024] Figure 1 This is a schematic flowchart of an embodiment of the rotating machinery alignment and calibration method provided by the present invention. The executing entity of the rotating machinery alignment and calibration method of this application can be an industrial control computer, an embedded controller, an edge computing device, or a cloud server. It is applicable to installation, maintenance, alignment, online status calibration, and periodic operation and maintenance debugging scenarios of rotating machinery in key equipment in industries such as energy, chemical, and manufacturing. It is particularly suitable for high-precision automated alignment and calibration under complex working conditions. Figures 1-3 The methods for centering and calibrating rotating machinery include: S101. Set multiple working condition parameters on the reference platform, and collect signal data from vibration sensor, eddy current sensor and speed sensor under each working condition. After preprocessing the signal data, extract multidimensional physical feature vectors.

[0025] It should be noted that the multi-condition parameters set on the reference platform in this application embodiment cover typical industrial operating conditions such as speed and load. Vibration sensors collect vibration acceleration signal data, eddy current sensors collect rotor radial displacement signal data, and speed sensors collect shaft speed signal data.

[0026] S102. Construct a mapping table between working conditions and feature importance based on multi-dimensional physical feature vectors. Based on the mapping table, integrate DS evidence theory to allocate feature weights, construct a knowledge base of centering state and feature mapping, and identify the feature vector and physical parameters of the optimal centering state in the knowledge base.

[0027] It should be noted that before constructing the mapping table between working conditions and feature importance, ideal alignment of the shaft system is pre-completed on the benchmark platform. The criteria for determining ideal alignment are: the radial runout deviation and end face runout deviation of the shaft system are both lower than the preset industry allowable error threshold, and the vibration amplitude is within the lowest stable range. Then, multi-sensor data of all working conditions under this ideal alignment state are collected. After preprocessing the multi-sensor data, multi-dimensional feature vectors are extracted and the corresponding physical parameters are calibrated to obtain the initial optimal alignment physical features and physical parameters.

[0028] S103. Obtain the real-time signal data of the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed. Read the feature weights from the alignment state and feature mapping knowledge base according to the current working conditions. Search for the matching alignment state in the alignment state and feature mapping knowledge base through the dynamic weighted similarity matching algorithm. Obtain the equivalent physical state of the device to be diagnosed by reverse reasoning based on the physical parameters in the matching sample.

[0029] It should be noted that the device to be diagnosed is a rotating mechanical shaft system device of the same model and structure as the reference platform. The equivalent physical state refers to the current actual physical alignment state of the shaft system of the device to be diagnosed, which is obtained by back-reasoning based on the matched physical feature vector. That is, it is a physical quantity that is equivalent to the state of the reference platform and can be directly used for calibration.

[0030] S104. Calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate a calibration adjustment command for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the alignment state and feature mapping knowledge base.

[0031] In summary, the rotating machinery alignment calibration method provided by this invention, by constructing an alignment state and feature mapping knowledge base and integrating DS evidence theory to allocate feature weights, can reverse-engineer the signal data from real-time acquired vibration sensors, eddy current sensors, and speed sensors into the equivalent physical state of the equipment to be diagnosed. It then calculates the physical deviation from the optimal alignment state and automatically generates calibration adjustment commands to adjust the alignment state, thereby achieving a leap from qualitative identification to quantitative diagnosis of misalignment faults and solving the problem of disconnect between diagnostic results and physical execution. The DS evidence theory-based allocation of feature weights makes the reasoning process interpretable and forms a closed-loop verification with the mechanical physical mechanism. Furthermore, by re-acquiring data and verifying the feature matching degree, if the feature matching degree does not reach a preset threshold, iterative fine-tuning and updating of the alignment state and feature mapping knowledge base are performed, forming an adaptive intelligent calibration process that significantly improves the efficiency and reliability of alignment calibration.

[0032] In some embodiments of the present invention, the preprocessing includes adaptive anti-interference preprocessing, which includes: performing wavelet threshold denoising on the acquired raw signal, and performing adaptive notch filtering to filter out power frequency and harmonic interference.

[0033] Specifically, the wavelet threshold denoising process performed on the acquired raw signal includes: An improved Sure threshold function is used to remove random noise. The improved Sure threshold function is defined as follows: Given wavelet coefficients ,in As a scale, For the location, first calculate the standard Sure threshold. :

[0034] in, For noise standard deviation estimation, The number of coefficients; Improved threshold for:

[0035] in, As an adaptive adjustment factor, The first norm of the coefficients is used; a soft thresholding function is then applied using this threshold.

[0036] in, These are the wavelet coefficients after processing with the soft thresholding function; This is the sign function, used to preserve the positive and negative signs of wavelet coefficients; represents the original wavelet coefficients, where As a scale, For location; To improve the threshold; This is a function that takes the maximum value and sets coefficients less than a threshold to 0, thus eliminating noise.

[0037] The threshold function described above can effectively remove random noise.

[0038] Furthermore, the adaptive notch filtering process is implemented using an adaptive notch filter, the transfer function of which is expressed as:

[0039] in, Let be the transfer function of the adaptive notch filter. For the power frequency that needs to be filtered out, Sampling frequency, The Z-transform operator is used for frequency domain modeling of discrete systems. The polar radius is used to control the notch bandwidth, and its value is adaptively adjusted according to the real-time estimated interference intensity.

[0040] This invention implements improved Sure threshold wavelet denoising on the original signal, adaptively determining the denoising threshold based on signal sparsity and noise intensity, effectively eliminating random noise and sudden interference while retaining fault-sensitive components; then, adaptive notch filtering is performed on the denoised signal, tracking the power frequency in real time and dynamically filtering out power frequency and its harmonic interference to ensure the accuracy of feature extraction.

[0041] In some embodiments of the present invention, the multidimensional physical feature vector includes vibration signal frequency domain features, eddy current displacement signal trajectory features, and multi-signal correlation features.

[0042] It should be noted that the multidimensional physical feature vector can be represented as: Specifically, it includes the following physical characteristics and their calculation formulas: Frequency domain characteristics of vibration signals: amplitude ratio of 2nd harmonic to 1st harmonic :

[0043] in, These represent the amplitudes of the vibration spectrum at the first and second harmonics, respectively. Axial to radial vibration energy ratio :

[0044] in, This represents the effective value of the axial (Z-direction) vibration signal. The effective value of the synthesized vibration signal in the radial (X, Y directions); Eddy current displacement signal trajectory characteristics: ellipticity of the axis trajectory The calculation formula is:

[0045] in, These are the lengths of the major and minor axes of the elliptical trajectory synthesized from orthogonal displacement signals, respectively.

[0046] The covariance matrix was obtained through principal component analysis (PCA). :

[0047] in, This represents the total number of trajectory points involved in the calculation. It is a data matrix composed of displacement signals in orthogonal directions (X, Y). For matrix The transpose of .

[0048] For covariance matrix Perform eigenvalue decomposition to obtain two eigenvalues ​​arranged in descending order of their numerical values. and ,and That is, the two eigenvalues ​​that contribute the largest and second largest to the variance of the discrete points of the trajectory, then the major semi-axis short half-shaft .

[0049] Standard deviation of trajectory point dispersion The calculation formula is:

[0050] in, The coordinates of the trajectory points, The center point of the trajectory.

[0051] Multi-signal correlation characteristics: 2-harmonic phase difference :

[0052] in, and These are the phases of the vibration signal and the displacement signal at their second harmonic frequencies, respectively.

[0053] This invention, through a combination of frequency domain amplitude ratio, energy ratio, axis center trajectory morphology, and multi-signal phase difference, characterizes the axis system state from three dimensions: vibration energy distribution, trajectory morphology, and signal correlation. This avoids the limitations of single features, resulting in a more complete and three-dimensional representation of the centering state, providing comprehensive feature data for subsequent state identification. Furthermore, the three types of features—frequency domain, trajectory, and multi-signal correlation—complement each other physically and possess different correlation characteristics in their data distribution, providing rich and reliable feature inputs for subsequent dynamic weighted similarity matching.

[0054] In some embodiments of the present invention, the working condition and feature importance mapping table is based on historical full-condition operation samples, fault labeling datasets and objective quantitative indicators to determine the importance priority of each physical feature and form the working condition and feature importance mapping table.

[0055] It should be noted that the determination of importance priority requires the comprehensive evaluation of three types of quantitative indicators: long-term operational stability of features, fault sensitivity values, and signal-to-noise ratio benchmark performance, to complete the objective hierarchical classification.

[0056] In some embodiments of the present invention, the step S102 of allocating feature weights according to the mapping table and fusing DS evidence theory includes: S201. Based on the DS evidence theory, construct an identification framework for each physical feature in the physical feature vector to identify the importance of the physical feature to state recognition; identification framework It can be represented as:

[0057] in, This indicates that the physical feature is important for state recognition. This indicates that the physical feature is not important for state recognition; its importance is determined using a quantitative threshold method. The specific determination method is as follows: Assume physical characteristics The signal-to-noise ratio under the current operating conditions is The fault sensitivity index is The stability index under operating conditions is Among them, the fault sensitivity index Defined as:

[0058] Operating stability index Defined as:

[0059] in, This represents the mean of the physical characteristic in the fault samples. This represents the mean of the physical characteristic in the normal sample. This represents the standard deviation of the physical characteristic obtained from multiple samplings under the same operating conditions. This is the average value obtained from multiple samplings of this physical characteristic under the same operating conditions.

[0060] If and only if the signal-to-noise ratio Fault sensitivity index And operating condition stability index When this physical feature is deemed "important" to state recognition, it is denoted as... ;when , or When the feature is determined to be in the critical importance interval, a portion of its basic probability allocation is assigned to the uncertain set. ;when or or When this feature is deemed "not important" for state recognition, it is denoted as... .

[0061] S202. Initialize the basic probability allocation function based on the mapping table of working conditions and feature importance and the real-time quality of features.

[0062] The real-time quality of the features refers to the feature signal-to-noise ratio and effectiveness quantification value, which are calculated in real time based on the currently acquired signal and dynamically updated according to the operating conditions. The signal-to-noise ratio... The calculation formula is:

[0063] Validity Quantification The calculation formula is:

[0064] in, For signal power, Noise power; The standard deviation of the characteristic amplitude, The mean of the characteristic amplitude, For the current eigenvalue, The optimal baseline eigenvalue, , Let be the normalized weighted coefficients, and satisfy . .

[0065] Based on the mapping table between operating conditions and feature importance, and the basic probability allocation function (BPA) for initializing real-time feature quality: And satisfy The unallocated portion is uncertain. .

[0066] S203. The Dempster combination rule is used to fuse the basic probability assignment functions of multiple features to obtain the joint basic probability assignment function.

[0067] Among them, the joint basic probability assignment function The formula is as follows:

[0068] Where K= As a conflict factor, The number of features.

[0069] S204. Calculate the support of each physical feature based on the joint fundamental probability assignment function, and normalize it to obtain the dynamic weights. The support is expressed as... Dynamic weights Represented as:

[0070] This invention achieves objective grading of feature importance through objective quantitative indicators (signal-to-noise ratio, fault sensitivity, and operating condition stability) and quantitative thresholds, avoiding subjective bias. It effectively handles evidence conflicts and uncertain information by utilizing the uncertainty allocation and Dempster combination rules of DS evidence theory, enhancing fusion robustness. Feature weights are dynamically updated based on real-time signal-to-noise ratio and validity quantification values, adapting to changes in operating conditions and signal quality fluctuations. The resulting dynamic weights can adaptively update with operating conditions, highlighting high-value features, suppressing noise interference, and significantly improving the accuracy of state recognition and reverse reasoning. Furthermore, the entire process is transparent and interpretable, solving the problems of black box and low reliability in traditional diagnostics.

[0071] In some embodiments of the present invention, the construction of the midpoint state and feature mapping knowledge base in step S102 includes: S301. Construct a three-layer structured storage architecture that includes a working condition layer, a feature layer, and a physical state layer; The operating condition layer stores operating condition parameters such as rotational speed and load, as well as a mapping table of operating conditions and feature importance under the corresponding operating conditions, and feature weight parameters of DS evidence theory.

[0072] Feature layer: Stores standardized multidimensional feature vectors for corresponding working conditions and physical alignment states, as well as quality parameters such as effective value, threshold range, and signal-to-noise ratio of the features; among which, the effective value, threshold range, and signal-to-noise ratio parameters are obtained by statistical fitting of massive steady-state samples from the benchmark platform and statistical calculation of the feature distribution interval across the entire domain.

[0073] Physical state layer: Stores the shaft alignment physical parameters that correspond one-to-one with the feature vectors, including physical quantities that can be directly used for calibration, such as radial horizontal / vertical offset, angular offset, and bearing seat adjustment, as well as the reference physical parameters for the optimal alignment state.

[0074] S302. Store the working parameters, physical feature vector, and optimal physical parameter set of the optimal alignment state into a three-layer structured storage architecture; use it as the benchmark reference system of the knowledge base and mark it as the root node of the knowledge base.

[0075] S303. The misalignment fault conditions are quantitatively set using a laser alignment instrument. Data acquisition, preprocessing, and feature extraction are performed repeatedly on the misalignment fault conditions to obtain multiple sets of condition parameter-physical feature vector-physical parameter data pairs, which are then stored in batches into a three-layer structured storage architecture.

[0076] It should be noted that, in this embodiment of the invention, a gradient-based misalignment fault condition is quantitatively set using a laser alignment instrument, including: radial offset gradient (0mm, ±0.02mm, ±0.05mm, ±0.10mm, ±0.20mm, ±0.50mm). For each quantitatively set misalignment state, the full-condition data acquisition, preprocessing, and feature extraction process in step 101 is repeated to obtain corresponding data pairs of "multiple sets of condition parameters - physical feature vectors - physical quantitative physical offset parameters", and batch storage is completed to build a sample library covering all fault degrees and all operating conditions.

[0077] S304. Based on the working condition and feature importance mapping table, the dynamic weight of each sample data point in the three-layer structured storage architecture is calculated using DS evidence theory and stored in the working condition layer. Simultaneously, using sample feature values ​​as input and measured physical offsets as output, a basic linear correlation between features and offsets is fitted through multiple linear regression. Then, multinomial nonlinear fitting is used to correct higher-order coupling errors, calibrating the quantitative mapping relationship between each feature value and the physical offset, thus completing the construction of the centering state and feature mapping knowledge base. This provides physical constraint boundaries for subsequent reverse reasoning.

[0078] This invention constructs a three-layer structured knowledge base—operating condition, feature, and physical state—to achieve a precise one-to-one mapping between feature vectors and axis physical offsets, providing reliable data support and physical constraints for reverse reasoning. Gradient-based quantitative calibration and linear and nonlinear fitting significantly improve the mapping accuracy between features and physical parameters, ensuring that diagnostic results can be directly converted into calibration instructions. Simultaneously, an adaptive similarity selection algorithm based on the Pearson correlation coefficient, combined with dynamic weights, achieves more robust state matching, effectively eliminating feature correlation interference and improving matching speed and accuracy.

[0079] In some embodiments of the present invention, the dynamic weighted similarity matching algorithm adaptively selects the corresponding similarity algorithm based on the correlation between physical features, specifically as follows: For strong correlations between physical features, the Mahalanobis distance algorithm is used. For weak correlations between physical features, the weighted Euclidean distance algorithm is used; If the physical features are approximately independent, the cosine similarity algorithm is used.

[0080] It should be noted that the correlation between physical characteristics is measured using the Pearson correlation coefficient. The calculation is expressed by the following formula:

[0081] in, Features With features The Pearson correlation coefficient between them; Features With features The covariance between features characterizes the degree of overall coordinated fluctuation. , Features , The standard deviation represents the degree of dispersion of a single set of features.

[0082] Correlation is determined by the absolute value of the correlation coefficient. This grading standard refers to the general statistical conventions for classifying correlations and is finally determined by combining the distribution statistics of massive measured feature samples on this benchmark platform: Strong correlation, It is a weak correlation. They are approximately independent.

[0083] When there is a strong correlation between physical features, the Mahalanobis distance algorithm is used. The formula is expressed as:

[0084] in, This is the real-time multidimensional physical feature vector to be matched; For the first in the knowledge base The multidimensional physical feature vector corresponding to each sample data; for The transpose operation. Real-time multidimensional physical feature vector The covariance matrix is ​​used to characterize the correlation between the various feature dimensions.

[0085] For weak correlations between physical features, the weighted Euclidean distance algorithm is used. The formula is expressed as:

[0086] in, No. The dynamic weights of each feature are calculated by fusing the aforementioned DS evidence theory. The th in the current real-time feature vector The values ​​of each feature component; For the first in the knowledge base The first sample feature vector The values ​​of each feature component.

[0087] If the physical features are approximately independent, then the cosine similarity algorithm is used. The formula is expressed as:

[0088] in, This is the real-time multidimensional physical feature vector to be matched; For the first in the knowledge base The multidimensional physical feature vector corresponding to each sample data.

[0089] Furthermore, when performing a matching search in the alignment state and feature mapping knowledge base, the real-time multidimensional physical feature vector of the current device to be diagnosed is first... With the multidimensional physical feature vectors of each sample in the knowledge base Similarity is calculated for each item, and the corresponding distance metric or similarity algorithm is selected based on relevance level. This is achieved through dynamic weighting. Each feature component is weighted to highlight high-value features and suppress noise interference features, thus obtaining a weighted similarity or weighted distance. Then, the N best matching samples with the highest similarity and smallest distance are selected as candidate matching alignment states, sorted from largest to smallest by similarity or distance. Based on the pre-stored shaft alignment physical parameters in the candidate matching samples, a multivariate linear regression combined with a multinomial nonlinear fitting model is used to quantitatively inversely infer the real-time feature vector and the equivalent physical alignment state, thereby obtaining the current equivalent physical alignment state of the device to be diagnosed, including radial horizontal / vertical offset, angular offset, and bearing seat adjustment, which are physical quantities that can be directly used for calibration.

[0090] It should be noted that after executing the calibration and adjustment command, the multi-sensor data of the device under diagnosis is re-acquired and new feature vectors are extracted. ,calculate Similarity to the features of the optimal alignment state; if the similarity does not reach the preset threshold. The calibration step size is adaptively adjusted based on the mapping relationship between physical deviation and feature deviation. Large step size is used for fast convergence for large deviations and small step size is used for fine adjustment for small deviations. The calibration adjustment and feature similarity verification are performed iteratively until the similarity reaches the standard. At the same time, the new "aligned state-feature" data pairs and the corresponding updated feature weights are synchronously added to the knowledge base to complete the dynamic optimization and continuous iteration of the knowledge base, further improving the accuracy and robustness of subsequent state matching and reverse reasoning.

[0091] To better implement the rotating machinery alignment and calibration method in the embodiments of the present invention, based on the rotating machinery alignment and calibration method, correspondingly, as follows: Figure 4 As shown, this embodiment of the invention also provides a rotating machinery alignment calibration device, the rotating machinery alignment calibration device 400 comprising: The data acquisition and preprocessing module 401 is used to set multi-condition parameters on the reference platform, and to acquire signal data from vibration sensors, eddy current sensors, and speed sensors under each condition. After preprocessing the signal data, multi-dimensional physical feature vectors are extracted. The knowledge base construction module 402 is used to construct a mapping table between working conditions and feature importance based on multi-dimensional physical feature vectors, allocate feature weights according to the mapping table and fused with DS evidence theory, construct a knowledge base of centering state and feature mapping, and identify the feature vector and physical parameters of the optimal centering state in the knowledge base; The reverse reasoning module 403 is used to acquire real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed, read feature weights from the knowledge base according to the current working conditions, search for matching pairs in the knowledge base through a dynamic weighted similarity matching algorithm, and obtain the equivalent physical state of the device to be diagnosed by reverse reasoning based on the physical parameters in the matching samples. The calibration module 404 is used to calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate a calibration adjustment command for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the alignment state and feature mapping knowledge base.

[0092] The rotating machinery alignment calibration device 400 provided in the above embodiments can realize the technical solutions described in the above rotating machinery alignment calibration method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above rotating machinery alignment calibration method embodiments, and will not be repeated here.

[0093] like Figure 5 As shown, the present invention also provides an electronic device 500. The electronic device 500 includes a processor 501, a memory 502, and a display 503. Figure 5Only some components of the electronic device 500 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0094] In some embodiments, processor 501 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 502 or process data, such as the rotating machinery alignment calibration method of the present invention.

[0095] In some embodiments, processor 501 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 501 may be local or remote. In some embodiments, processor 501 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-cloud, multi-cloud, or any combination thereof.

[0096] In some embodiments, memory 502 may be an internal storage unit of electronic device 500, such as a hard disk or memory of electronic device 500. In other embodiments, memory 502 may also be an external storage device of electronic device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 500.

[0097] Furthermore, the memory 502 may include both internal storage units of the electronic device 500 and external storage devices. The memory 502 is used to store application software and various types of data installed on the electronic device 500.

[0098] In some embodiments, display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 503 is used to display information from electronic device 500 and to display a visual user interface. Components 501-503 of electronic device 500 communicate with each other via a system bus.

[0099] In one embodiment, when processor 501 executes the rotating machinery alignment calibration program in memory 502, the following steps can be implemented: Multiple operating condition parameters are set on the benchmark platform, and signal data from vibration sensors, eddy current sensors, and speed sensors are collected under each operating condition. After preprocessing the signal data, multidimensional physical feature vectors are extracted. A mapping table between working conditions and feature importance is constructed based on multidimensional physical feature vectors. Feature weights are allocated according to the mapping table and DS evidence theory. A knowledge base of centering state and feature mapping is constructed, and the feature vector and physical parameters of the optimal centering state are identified in the knowledge base. The system acquires real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed. Based on the current operating conditions, it reads feature weights from the knowledge base, searches for matching pairs in the knowledge base using a dynamic weighted similarity matching algorithm, and inversely infers the equivalent physical state of the device to be diagnosed based on the physical parameters in the matching samples. Calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate calibration adjustment instructions for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the knowledge base.

[0100] It should be understood that when the processor 501 executes the rotary mechanical alignment calibration program in the memory 502, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0101] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions of the rotating machinery alignment and calibration methods provided in the above-described method embodiments.

[0102] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0103] The rotating machinery alignment and calibration method, apparatus, electronic device, and storage medium provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for centering and calibrating rotating machinery, characterized in that, include: Multiple operating condition parameters are set on the benchmark platform, and signal data from vibration sensors, eddy current sensors, and speed sensors are collected under each operating condition. After preprocessing the signal data, multidimensional physical feature vectors are extracted. A mapping table between working conditions and feature importance is constructed based on multidimensional physical feature vectors. Feature weights are allocated according to the mapping table and DS evidence theory. A knowledge base of centering state and feature mapping is constructed, and the feature vector and physical parameters of the optimal centering state are identified in the knowledge base. The system acquires real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed. Based on the current operating conditions, it reads feature weights from the knowledge base, searches for matching pairs in the knowledge base using a dynamic weighted similarity matching algorithm, and inversely infers the equivalent physical state of the device to be diagnosed based on the physical parameters in the matching samples. Calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate calibration adjustment instructions for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the knowledge base.

2. The rotational machinery alignment and calibration method according to claim 1, characterized in that, The preprocessing includes adaptive anti-interference preprocessing, which includes: performing wavelet threshold denoising on the acquired raw signal, and performing adaptive notch filtering to filter out power frequency and harmonic interference.

3. The rotational machinery alignment and calibration method according to claim 1, characterized in that, The multidimensional physical feature vector includes the frequency domain features of vibration signals, the trajectory features of eddy current displacement signals, and the correlation features of multiple signals.

4. The method for centering and calibrating rotating machinery according to claim 1, characterized in that, The working condition and feature importance mapping table is based on historical full-condition operation samples, fault labeling datasets and objective quantitative indicators to determine the importance priority of each physical feature, thus forming the working condition and feature importance mapping table.

5. The rotational machinery alignment and calibration method according to claim 1, characterized in that, The process of assigning feature weights based on the DS evidence theory fusion according to the mapping table includes: Based on the DS evidence theory, an identification framework is constructed for each physical feature in the physical feature vector to identify the importance of the physical feature to state identification. The basic probability allocation function is initialized based on the mapping table between working conditions and feature importance and the real-time quality of features; The Dempster combination rule is used to fuse the basic probability assignment functions of multiple features to obtain a joint basic probability assignment function; The support of each physical feature is calculated based on the joint basic probability assignment function, and then normalized to obtain the dynamic weights.

6. The rotational machinery alignment and calibration method according to claim 1, characterized in that, The construction of the midpoint state and feature mapping knowledge base includes: Construct a three-layer structured storage architecture comprising a working condition layer, a feature layer, and a physical state layer; The operating parameters, physical feature vectors, and optimal physical parameter set of the optimal alignment state are stored in a three-layer structured storage architecture. The misalignment fault conditions are quantitatively set using a laser alignment instrument. Data collection, preprocessing, and feature extraction are performed repeatedly on the misalignment fault conditions to obtain multiple sets of condition parameter-physical feature vector-physical parameter data pairs, which are then stored in batches into a three-layer structured storage architecture. Based on the working condition and feature importance mapping table, the dynamic weight of each sample data in the three-layer structured storage architecture is calculated by DS evidence theory and stored in the working condition layer. At the same time, the quantitative mapping relationship between feature values ​​and physical parameters is calibrated by linear regression and nonlinear fitting, thus completing the construction of the centering state and feature mapping knowledge base.

7. The method for centering and calibrating rotating machinery according to claim 1, characterized in that, The dynamic weighted similarity matching algorithm adaptively selects the corresponding similarity algorithm based on the correlation between physical features, specifically as follows: For strong correlations between physical features, the Mahalanobis distance algorithm is used. For weak correlations between physical features, the weighted Euclidean distance algorithm is used; If the physical features are approximately independent, the cosine similarity algorithm is used.

8. A centering and calibration device for rotating machinery, characterized in that, include: The data acquisition and preprocessing module is used to set multiple working condition parameters on the benchmark platform and acquire signal data from vibration sensors, eddy current sensors, and speed sensors under each working condition. After preprocessing the signal data, multi-dimensional physical feature vectors are extracted. The knowledge base construction module is used to construct a mapping table between working conditions and feature importance based on multi-dimensional physical feature vectors, allocate feature weights according to the mapping table and DS evidence theory, construct a knowledge base of centering state and feature mapping, and identify the feature vector and physical parameters of the optimal centering state in the knowledge base. The reverse reasoning module is used to acquire real-time signal data from the vibration sensor, eddy current sensor, and speed sensor of the device to be diagnosed, read feature weights from the knowledge base according to the current working conditions, search for matching pairs in the knowledge base through a dynamic weighted similarity matching algorithm, and obtain the equivalent physical state of the device to be diagnosed by reverse reasoning based on the physical parameters in the matching samples. The calibration module is used to calculate the physical parameter deviation between the equivalent physical state and the optimal alignment state, generate calibration adjustment instructions for adjusting the alignment state based on the physical parameter deviation, re-acquire data and verify the feature matching degree, and if the feature matching degree does not reach the preset threshold, iteratively fine-tune and update the alignment state and feature mapping knowledge base.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the rotating machinery alignment and calibration method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can perform the steps in the rotating machinery alignment and calibration method according to any one of claims 1 to 7.

Images