A method and system for monitoring the health of a treadmill

Abstract

This invention relates to the field of treadmill technology, and more particularly to a method and system for monitoring the health status of a treadmill. The method includes the following steps: acquiring structural response data from different locations during treadmill operation; performing time alignment, frequency band separation, and interference suppression to form preprocessed data, improving analysis reliability; extracting modal parameters, including frequency, damping, and modal vectors, from the preprocessed data to generate a modal parameter set characterizing vibration characteristics; constructing and reducing the dimensionality of the modal parameter set through multi-source correlation features to obtain low-dimensional features, highlighting key modes; statistically analyzing low-dimensional features grouped by operating status to form a modal baseline and comparing it with current features to obtain offsets, quantifying the structural state; constructing a matrix based on the offsets; analyzing the contribution of components to modal anomalies through matrix decomposition; and combining this with reverse verification to determine the fault source and structural region. This invention improves upon the limitations of single sensors by analyzing multi-location responses through matrix decomposition, achieving more accurate health assessment.

Description

Technical Field

[0001] This invention relates to the field of treadmill technology, and in particular to a method and system for monitoring the health status of a treadmill. Background Technology

[0002] With the increasing awareness of fitness, treadmills, as a highly efficient aerobic exercise device, have been widely used in home fitness, school sports training, and commercial gyms. Treadmills can simulate the outdoor running environment, allowing for controllable adjustments to speed, incline, and exercise intensity, providing users with a convenient, continuous, and safe way to exercise. Meanwhile, with the rise of smart fitness equipment, treadmills are beginning to integrate more sensors and control systems, such as speed sensors, load sensors, heart rate monitoring devices, and vibration sensors, enabling the equipment to collect exercise data and mechanical operating status in real time. However, under frequent use and complex load conditions, key components of treadmills are prone to wear, loosening, or performance degradation.

[0003] Traditional treadmill health monitoring mostly relies on a single sensor or simple indicators, which cannot reflect the differences in structural response at multiple locations, resulting in low monitoring accuracy. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a method and system for monitoring the health status of a treadmill, aiming to improve the problem that traditional treadmill health monitoring mostly relies on a single sensor or simple indicators, which cannot reflect the differences in response at multiple locations, resulting in low monitoring accuracy.

[0005] In a first aspect, the present invention provides the following technical solution: a method for monitoring the health status of a treadmill, comprising the following steps:

[0006] Structural response data from different locations of the treadmill during operation is acquired and time-aligned to obtain aligned data. Frequency band separation and interference suppression are then performed on the aligned data to form preprocessed data.

[0007] Based on the preprocessed data, structural modal parameters, including frequency, damping, and mode vectors, are extracted to obtain a set of modal parameters;

[0008] Perform multi-source correlation feature construction and dimensionality reduction on the modal parameter set to generate a low-dimensional correlation feature set;

[0009] The low-dimensional correlation feature set is grouped and statistically analyzed according to the running state to generate a modal baseline, which is then compared with the current low-dimensional correlation features to obtain a modal offset set.

[0010] Based on the set of modal offsets, a matrix is ​​constructed to represent the relationship between the position of each sensor and the modal offset. The contribution of each component to the modal anomaly is analyzed through matrix decomposition and reverse verification to determine the fault source component.

[0011] Based on the fault source component, determine the structural region related to the offset.

[0012] By adopting the above technical solution, a matrix is ​​constructed based on the modal offset set and matrix decomposition is performed. Then, the contribution of each component to the modal anomaly is analyzed. This improves the problem that traditional treadmill health monitoring mostly relies on a single sensor or simple indicators, which cannot reflect the structural response differences at multiple locations, resulting in low monitoring accuracy.

[0013] Furthermore, acquiring structural response data from different locations of the treadmill during operation includes the following steps:

[0014] Data collection begins simultaneously at multiple collection points on the treadmill;

[0015] The structural response data collected at each location is recorded in real time;

[0016] Add a time stamp to each piece of collected data;

[0017] Save the recorded data from each collection location to the data buffer;

[0018] The buffer data is integrated in chronological order to form a structured response data set.

[0019] Furthermore, the frequency band separation and interference suppression of the aligned data includes the following steps:

[0020] Determine the target frequency band range from the time-aligned data;

[0021] Filter signal components that exceed the target frequency band;

[0022] Smoothing or denoising is performed on high-frequency noise or low-frequency drift in the data.

[0023] Standardize or adjust the amplitude of each data point;

[0024] The processed data is used to form a preprocessed dataset for subsequent analysis.

[0025] Furthermore, the extraction of structural modal parameters includes the following steps:

[0026] The preprocessed data is divided according to the sampling time period;

[0027] Perform spectral analysis on the data for each time period to identify the main vibration frequencies;

[0028] Damping fitting is performed on the data corresponding to each major vibration frequency, and the damping ratio is calculated.

[0029] The vibration mode of each mode is calculated based on the multi-channel response signals of each time period, forming a mode vector;

[0030] The frequency, damping, and mode vectors for each time period are summarized to generate a set of modal parameters.

[0031] Furthermore, the multi-source correlation feature construction and dimensionality reduction of the modal parameter set includes the following steps:

[0032] The modal parameter set is organized into a feature matrix according to time period and acquisition channel;

[0033] Standardize or normalize the modal parameters in the matrix;

[0034] Calculate the correlation or covariance between modal parameters and construct a multi-source correlation feature matrix;

[0035] Identify and remove redundant or highly correlated features;

[0036] Apply dimensionality reduction methods to the remaining features to map the high-dimensional features to a low-dimensional space;

[0037] Output a set of low-dimensional correlation features.

[0038] Furthermore, the step of grouping and statistically analyzing the low-dimensional correlation feature set according to its operating state includes the following steps:

[0039] Based on preset operational status indicators, the low-dimensional correlation feature set is divided into multiple groups;

[0040] Perform statistical analysis on the feature vectors within each group, including calculating the mean and variance;

[0041] The statistical results of each group are summarized to form a set of modal baselines for the corresponding operating states;

[0042] The currently collected low-dimensional correlation features are grouped into corresponding groups based on the same operational status indicators.

[0043] Furthermore, the construction of the matrix representing the relationship between the position of each sensor and the modal offset includes the following steps:

[0044] Collect modal offset data for each sensor;

[0045] Arrange the data in order of sensor location to form rows of a matrix;

[0046] Use the offset of each modal feature or time period as the column of the matrix;

[0047] The matrix data is standardized or normalized to form a feature matrix.

[0048] Furthermore, the matrix decomposition includes the following steps:

[0049] Input the constructed feature matrix into the matrix factorization algorithm;

[0050] Decompose the matrix to obtain the fundamental matrix and coefficient matrix or singular values ​​and left and right singular vectors;

[0051] Analyze the decomposition results to determine the contribution of each sensor or structural component to the modal offset;

[0052] The output contribution matrix provides a preliminary judgment for fault source identification.

[0053] Furthermore, the reverse verification includes the following steps:

[0054] Based on the matrix factorization results, predict the contribution of each candidate fault source component to the modal offset of each sensor;

[0055] The predicted results are compared with the actual modal offsets, and the error or residual is calculated.

[0056] By combining time delay analysis, vibration propagation analysis, and modal vector contribution analysis, a comprehensive evaluation of candidate fault sources is conducted.

[0057] The rationality of candidate fault sources is determined based on the comprehensive evaluation results, and candidate components that do not meet the conditions are eliminated.

[0058] Output the set of fault source components after reverse verification and comprehensive evaluation.

[0059] Secondly, the present invention provides the following technical solution: a treadmill health status monitoring system, the system comprising:

[0060] The data acquisition module is used to acquire structural response data from different positions of the treadmill during operation, perform time alignment to obtain aligned data, and perform frequency band separation and interference suppression on the aligned data to form preprocessed data.

[0061] The modal extraction module is used to extract structural modal parameters, including frequency, damping and modal vectors, based on the preprocessed data to obtain a set of modal parameters;

[0062] The feature construction module is used to perform multi-source correlation feature construction and dimensionality reduction on the modality parameter set to generate a low-dimensional correlation feature set;

[0063] The modal baseline module is used to group and statistically analyze the low-dimensional correlation feature set according to the running state, generate a modal baseline, and compare it with the current low-dimensional correlation features to obtain a modal offset set.

[0064] The matrix analysis module is used to construct a matrix representation of the relationship between the position of each sensor and the modal offset based on the set of modal offsets, and to analyze the contribution of each component to the modal anomaly through matrix decomposition and reverse verification to determine the fault source component.

[0065] The fault location module is used to determine the structural region related to the offset based on the fault source component.

[0066] The present invention has the following beneficial effects:

[0067] 1. In this invention, a matrix is ​​constructed based on the modal offset set and matrix decomposition is performed to analyze the contribution of each component to modal anomalies. This improves the problem that traditional treadmill health monitoring mostly relies on a single sensor or simple indicators, which cannot reflect the differences in structural responses at multiple locations, resulting in low monitoring accuracy.

[0068] 2. In this invention, by performing multi-source correlation feature construction and dimensionality reduction on the modal parameter set, a low-dimensional correlation feature set is generated, thereby improving the problem that traditional methods mostly directly use the original high-dimensional modal parameters, which cause inaccurate anomaly detection due to severe feature redundancy and difficulty in stable modeling.

[0069] 3. In this invention, by grouping and statistically analyzing low-dimensional correlation features according to their operating states and comparing them with the current features to obtain a set of modal offsets, the problem of traditional monitoring, which mostly only analyzes overall indicators, is improved. Due to the lack of reference baselines for different operating states, the fault location is ambiguous and difficult to identify accurately. Attached Figure Description

[0070] Figure 1 This is a flowchart of a method for monitoring the health status of a treadmill, as proposed in this invention.

[0071] Figure 2 This is a system architecture diagram of a treadmill health status monitoring system proposed in this invention. Detailed Implementation

[0072] The technical solutions in 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example 1

[0074] In a first embodiment of the present invention, the present invention provides a method for monitoring the health status of a treadmill, such as... Figure 1As shown, it includes the following steps:

[0075] Structural response data from different locations of the treadmill during operation is acquired and time-aligned to obtain aligned data. Frequency band separation and interference suppression are then performed on the aligned data to form preprocessed data.

[0076] Furthermore, acquiring structural response data from different locations of the treadmill during operation includes the following steps:

[0077] Data collection begins simultaneously at multiple collection points on the treadmill;

[0078] The structural response data collected at each location is recorded in real time;

[0079] Add a time stamp to each piece of collected data;

[0080] Save the recorded data from each collection location to the data buffer;

[0081] The buffer data is integrated in chronological order to form a structured response data set.

[0082] Specifically, the process of acquiring structural response data from different locations on the treadmill during operation and performing time alignment and preprocessing on the data is achieved through multi-point synchronous acquisition and time stamping. In practice, data acquisition is first initiated simultaneously at multiple acquisition locations on the treadmill, and the structural response signal acquired at each location is denoted as... Indicates the sensor number, The sampling time is indicated by the data being recorded in real time by sensors to form a continuous time series. Each data point is appended with a sampling timestamp to ensure the time correspondence between data from different locations. Subsequently, the data from each sampling location is saved to a buffer and integrated in chronological order to generate a structured response dataset. The number of collection locations, set Each row corresponds to a signal sequence from a different sensor location, and each column corresponds to the data collected by each sensor at the same time point. The integrated dataset undergoes frequency band separation and interference suppression processing, targeting specific frequency bands. The signals within the target frequency band retain their original vibration characteristics. Filtering is performed on components exceeding the target frequency band, while smoothing or denoising is applied to high-frequency noise and low-frequency drift. The processed results form a preprocessed dataset. This set is used for subsequent modality analysis and feature extraction; each element This represents the corresponding position sensor signal after filtering and smoothing, ensuring that it can be used for modal parameter extraction and multi-source feature construction under time alignment and frequency band selection.

[0083] By synchronously acquiring and time-aligning structural response data at various locations, and performing frequency band separation and interference suppression, high-quality, denoised preprocessed data can be obtained, providing a reliable foundation for subsequent modal analysis and feature extraction.

[0084] Furthermore, frequency band separation and interference suppression of the aligned data include the following steps:

[0085] Determine the target frequency band range from the time-aligned data;

[0086] Filter signal components that exceed the target frequency band;

[0087] Smoothing or denoising is performed on high-frequency noise or low-frequency drift in the data.

[0088] Standardize or adjust the amplitude of each data point;

[0089] The processed data is used to form a preprocessed dataset for subsequent analysis.

[0090] Specifically, for time-aligned structural response data The process of frequency band separation and interference suppression is achieved through filtering, smoothing, and normalization. First, based on the preset target frequency band... Determine the vibration signal components that need to be retained, and apply filtering to signals outside this frequency band for each sensor signal. After filtering Filtering can be expressed using bandpass filters or frequency domain selection methods. and These represent the Fourier transform and the inverse Fourier transform, respectively. This is the frequency response function, used to preserve the target frequency band signal; subsequently, for... Smoothing or denoising processes are performed to suppress high-frequency noise and low-frequency drift, resulting in... This involves smoothing or filtering operations to ensure greater signal stability in amplitude and trend; finally, each data point is standardized or its amplitude adjusted to make the signals from different sensors comparable under the same dimensions, resulting in the final preprocessed data. This set can be directly used for subsequent modal parameter extraction, feature construction, and dimensionality reduction analysis; among which For the sampling time point, Number the sensor. To ensure the number of data collection locations, the processed dataset guarantees the comparability of time and frequency characteristics of signals from different sensors, providing reliable input for subsequent modal analysis.

[0091] By separating frequency bands, filtering and denoising, and smoothing the data, and by standardizing the data, stable and consistent preprocessed data can be obtained, which improves the accuracy of subsequent modal parameter extraction and feature analysis.

[0092] Structural modal parameters, including frequency, damping, and mode vectors, are extracted from preprocessed data to obtain a set of modal parameters;

[0093] Furthermore, extracting structural modal parameters includes the following steps:

[0094] The preprocessed data is divided according to the sampling time period;

[0095] Perform spectral analysis on the data for each time period to identify the main vibration frequencies;

[0096] Damping fitting is performed on the data corresponding to each major vibration frequency, and the damping ratio is calculated.

[0097] The vibration mode of each mode is calculated based on the multi-channel response signals of each time period, forming a mode vector;

[0098] The frequency, damping, and mode vectors for each time period are summarized to generate a set of modal parameters.

[0099] Specifically, based on preprocessed data The process of extracting structural modal parameters is achieved through time segmentation, spectrum analysis, damping fitting, and modal vector calculation. First, the preprocessed data is divided into sampling time periods. The data is divided into segments to obtain a data matrix for each time period. Subsequently, Spectral analysis is performed on each channel signal to identify the main vibration frequencies. Spectrum analysis can be performed using Fourier transform. Channel number, At the sampling time point, the main frequency components are determined by the peak value of the Fourier transform; for each main frequency... Damping fitting is performed on the response signal to calculate the damping ratio. Logarithmic decay method can be used. The peak value is the maximum amplitude of the vibration signal at time t. For the corresponding period; according to each time period The multi-channel response signal was analyzed, and the vibration mode of each mode was calculated using modal analysis. , forming mode vectors Indicates the first The first channel is for the first The vibrational contribution of each mode; the modal frequency for each time period. Damping ratio and mode vector The modal parameter set is obtained by summarizing. The total number of time periods. The modality parameter set is used to construct multi-source association features, reduce dimensionality, and calculate modal offsets for each segment, providing basic data input for treadmill health status monitoring.

[0100] By performing spectral analysis, damping fitting, and modal vector calculation on the preprocessed data, a comprehensive set of modal parameters can be obtained, providing reliable basic data for subsequent feature construction and anomaly diagnosis.

[0101] Perform multi-source correlation feature construction and dimensionality reduction on the modal parameter set to generate a low-dimensional correlation feature set;

[0102] Furthermore, the multi-source correlation feature construction and dimensionality reduction of the modal parameter set includes the following steps:

[0103] The modal parameter set is organized into a feature matrix according to time period and acquisition channel;

[0104] Standardize or normalize the modal parameters in the matrix;

[0105] Calculate the correlation or covariance between modal parameters and construct a multi-source correlation feature matrix;

[0106] Identify and remove redundant or highly correlated features;

[0107] Apply dimensionality reduction methods to the remaining features to map the high-dimensional features to a low-dimensional space;

[0108] Output a set of low-dimensional correlation features.

[0109] Specifically, for the modal parameter set The process of constructing and reducing multi-source association features is achieved through matrix organization, standardization, correlation calculation, feature selection, and dimensionality reduction. First, the modal parameter set is divided into time periods. and acquisition channels Organize into a feature matrix ;in Indicates the first Time period The mode in the first The measurement frequency of the channel is used to correct for sensor installation errors; the natural frequency of this mode is... The calculation formula is The measured damping ratio for the corresponding channel is given by the inherent damping ratio of this mode. The calculation formula is Let be the modal vector element, representing the th The first channel is for the first The contribution of each mode to the vibration response; for the matrix The modal parameters were standardized to obtain and Each is a matrix Mean and standard deviation; calculate the standardized matrix. Correlation or covariance matrix of each modal parameter This is used to construct a multi-source correlation feature matrix, through calculation. Determine the correlation between features, identify and remove redundant or highly correlated features, and obtain the filtered feature matrix. ; for residual features Applying dimensionality reduction methods The dimension-reduced mapping matrix can be obtained using principal component analysis. For a low-dimensional set of related features, a matrix , The dimension after dimensionality reduction; a set of low-dimensional related features. This data is used to generate modal baselines and modal offsets based on subsequent grouping and statistical analysis according to operating status, providing input data for fault location and health status assessment.

[0110] By constructing a multi-source correlation feature matrix and performing dimensionality reduction, a low-dimensional correlation feature set can be obtained, which improves the effectiveness of feature expression and reduces redundancy, providing a simplified and reliable data foundation for modal anomaly analysis.

[0111] The low-dimensional correlation feature set is grouped and statistically analyzed according to the running status to generate a modal baseline, which is then compared with the current low-dimensional correlation feature set to obtain the modal offset set.

[0112] Furthermore, grouping and statistically analyzing the low-dimensional correlation feature set according to its operational status includes the following steps:

[0113] Based on preset operational status indicators, the low-dimensional correlation feature set is divided into multiple groups;

[0114] Perform statistical analysis on the feature vectors within each group, including calculating the mean and variance;

[0115] The statistical results of each group are summarized to form a set of modal baselines for the corresponding operating states;

[0116] The currently collected low-dimensional correlation features are grouped into corresponding groups based on the same operational status indicators.

[0117] Specifically, the operating status index S is set as the load-speed combination condition of the treadmill, including m=3 typical conditions. S1 is the no-load + low-speed condition, corresponding to a load of 0kg and a speed of 0-6km / h; S2 is the light-load + medium-speed condition, corresponding to a load of 50kg and a speed of 6-12km / h; S3 is the heavy-load and high-speed condition, corresponding to a load of 100kg and a speed of 12-18km / h. The low-dimensional associated feature set is then used. The process of grouping and statistically analyzing data according to operational status is achieved through the division of operational status indicators, intra-group statistical analysis, and baseline generation. First, based on preset operational status indicators... For low-dimensional feature vectors Grouping, forming The corresponding running status is , For low-dimensional features, For each group, the sample size is [number]; for each group... Calculate the mean of the eigenvectors within. and variance The statistical results of each group are then summarized to form a modal baseline set. Then, the currently collected low-dimensional correlation features will be... Based on operational status indicators, they are assigned to the corresponding groups and compared with the modal baseline of that group. Perform comparison and calculate modal offset. Modal offset set This is used for subsequent fault source localization and health status assessment based on matrix factorization and reverse verification. Each element represents the degree of offset of the current feature relative to the baseline in each dimension of the low-dimensional space. The input data is a set of low-dimensional correlated features. and operating status indicators The output is a set of modal baselines. and modal offset set The next step will be This is used to construct a matrix representing the relationship between the position of each sensor and the modal offset, providing input for subsequent matrix decomposition and reverse verification.

[0118] By grouping by operating status and statistically analyzing low-dimensional correlation features, modal baselines can be established and modal offset sets can be calculated, thereby quantifying the deviation from the current state and providing a reliable basis for subsequent fault diagnosis.

[0119] Based on the modal offset set, a matrix is ​​constructed to represent the relationship between the position of each sensor and the modal offset. The contribution of each component to the modal anomaly is analyzed through matrix decomposition and reverse verification to determine the fault source component.

[0120] Based on the fault source component, determine the structural region related to the offset;

[0121] Furthermore, constructing a matrix representation of the relationship between the positions of each sensor and the modal offset includes the following steps:

[0122] Collect modal offset data for each sensor;

[0123] Arrange the data in order of sensor location to form rows of a matrix;

[0124] Use the offset of each modal feature or time period as the column of the matrix;

[0125] The matrix data is standardized or normalized to form a feature matrix.

[0126] Specifically, based on the modal offset set The process of constructing a matrix representing the relationship between the position of each sensor and the modal offset involves collecting offset data corresponding to each sensor. Represents the low-dimensional feature dimension. The number of sensors is represented by the matrix. The data from each sensor are arranged in positional order to form the rows of a matrix, and the columns of the matrix correspond to the offsets of each modal feature or time period, thus obtaining the feature matrix. Then the matrix z-score normalization is used to eliminate dimensional differences and amplitude fluctuations, resulting in a normalized feature matrix. The calculation formula is For matrix The global mean is calculated as follows: For matrix The global standard deviation is calculated as follows: The input data is a set of modal offsets. The offset vector between the current sensor and the baseline mode is calculated as follows: The output is a standardized feature matrix. Used for subsequent matrix factorization algorithms Analyze the contribution of each structural component to the modal anomaly; among which The left singular vector matrix represents the sensor feature contribution. The singular value matrix represents the importance of each feature. The right singular vector matrix represents the combination of modal features. The decomposition results, combined with reverse verification, are used to screen reasonable fault source components and determine the structural regions that cause modal shifts. This provides a basis for subsequent health status assessment and fault location. The entire process ensures that the mapping relationship between modal shifts and sensor positions is clear, providing a quantifiable data foundation for fault source analysis.

[0127] By constructing a matrix of sensor positions and modal offsets and performing standardization, the contribution of each sensor to structural anomalies can be quantified, providing a data foundation for subsequent matrix decomposition and fault source localization.

[0128] Furthermore, matrix factorization includes the following steps:

[0129] Input the constructed feature matrix into the matrix factorization algorithm;

[0130] Decompose the matrix to obtain the fundamental matrix and coefficient matrix or singular values ​​and left and right singular vectors;

[0131] Analyze the decomposition results to determine the contribution of each sensor or structural component to the modal offset;

[0132] The output contribution matrix provides a preliminary judgment for fault source identification.

[0133] Specifically, matrix factorization is used to analyze the contribution of each sensor's position to the modal offset by constructing a standardized feature matrix. Input matrix factorization algorithms, such as Singular Value Decomposition (SVD) Indicates the number of sensors. Represents the modal feature dimension. The left singular vector matrix represents the sensor feature contribution. The singular value matrix represents the importance of each feature. The right singular vector matrix represents the modal feature combination. The rank of the matrix or the chosen dimension reduction dimension; the input data is a matrix. ; each of the lines Indicates the first The modal offset vector of each sensor is calculated as follows: , This is the sensor modal offset. and These are the mean and standard deviation, used for standardization; the output is a contribution matrix. Each row represents the contribution of the corresponding sensor to each modal offset, which is used for subsequent reverse verification to screen reasonable fault source components, determine the structural region that causes the modal offset, and provide a quantitative basis for health status assessment and fault location.

[0134] By performing matrix decomposition on the feature matrix, the contribution of each sensor and structural component to the modal offset can be quantified, providing a preliminary basis for identifying potential fault sources and supporting subsequent verification analysis.

[0135] Furthermore, reverse verification includes the following steps:

[0136] Based on the matrix factorization results, predict the contribution of each candidate fault source component to the modal offset of each sensor;

[0137] The predicted results are compared with the actual modal offsets, and the error or residual is calculated.

[0138] By combining time delay analysis, vibration propagation analysis, and modal vector contribution analysis, a comprehensive evaluation of candidate fault sources is conducted.

[0139] The rationality of candidate fault sources is determined based on the comprehensive evaluation results, and candidate components that do not meet the conditions are eliminated.

[0140] Output the set of fault source components after reverse verification and comprehensive evaluation.

[0141] Specifically, reverse verification involves obtaining the contribution matrix through matrix decomposition. This is used to predict the contribution of each candidate fault source component to the modal offset of each sensor. The prediction result is denoted as... Indicates the number of sensors. Representing feature dimension, Predict the offset matrix. For modal feature dimensions; This is a linear combination operation based on SVD decomposition. For the contribution matrix, The predicted offset is reconstructed by matrix multiplication using the transpose of the right singular vector matrix; the prediction result is then... With actual modal offset Compare and calculate the error or residual. Indicates the first The first sensor The deviation of each modal characteristic is analyzed; combined with time delay analysis, vibration propagation analysis, and modal vector contribution analysis, a comprehensive evaluation of candidate fault sources is conducted, generating a rationality index matrix. This represents the time delay effect matrix. Represents the vibration propagation path matrix. Represents the modal vector contribution matrix; For comprehensive evaluation functions, The residual matrix The 2-norm is used for normalization. This means taking the diagonal elements of the matrix to obtain the rationality index of each candidate fault source; Transpose the matrix that contributes to the modal vectors to ensure dimension matching in matrix multiplication; The larger the value, the more likely it is to be a candidate fault source. The stronger the interpretability of mode shifts, the greater the impact of time delay on the matrix. , For the number of sensors, This represents the number of candidate fault sources. Indicates the first The vibration signal of the first candidate fault source propagates to the second... The time delay of each sensor is calculated through cross-correlation analysis, specifically as follows: It is a cross-correlation function. Time search range; vibration propagation path matrix , Indicates the first The candidate fault source to the first The vibration attenuation coefficient of each sensor is determined based on the treadmill structural drawings, specifically as follows: , Candidate fault sources With sensors Straight-line distance; Modal vector contribution matrix That is, the modal vector matrix extracted earlier. Indicates the first The first channel is for the first The contribution of each modality; based on Determine the validity of candidate fault sources, eliminate components that do not meet the criteria, and output the set of filtered fault source components. The reasonableness threshold. Based on the statistical significance level, the calculation formula is: The average of the rationality indices for all candidate fault sources. The standard deviation of the rationality index is used to identify the specific structural components that cause modal shifts, providing a quantitative basis for treadmill health status analysis and fault location. The next step can be used to guide structural maintenance and risk warning.

[0142] By combining reverse verification with time delay, vibration propagation and modal contribution analysis, reasonable fault source components can be accurately screened and their corresponding structural regions can be determined, thereby improving the reliability and accuracy of fault location.

[0143] Example 2:

[0144] In commercial gyms, the simultaneous operation of numerous treadmills subjects the equipment to repetitive vibrations and loads over extended periods, leading to wear, loosening, and potential malfunctions in structural components that are difficult to detect promptly, posing challenges to safety management and maintenance. To address these issues, this invention provides a treadmill health status monitoring system, the structure of which is as follows: Figure 2 As shown.

[0145] The specific implementation process of the system is as follows: First, the data acquisition module acquires structural response data from different positions of the treadmill during operation, and performs time alignment, frequency band separation and interference suppression on the data to form high-quality preprocessed data, ensuring the accuracy of subsequent modal analysis and eliminating noise interference;

[0146] Subsequently, the modal extraction module extracts structural modal parameters, including frequency, damping, and modal vectors, based on the preprocessed data, and generates a set of modal parameters. This step can accurately characterize the vibration characteristics of each component of the treadmill and provide basic information for fault diagnosis.

[0147] Next, the feature construction module performs multi-source correlation feature construction and dimensionality reduction on the modal parameter set to form a low-dimensional correlation feature set, thereby effectively integrating multi-channel information and reducing data dimensionality, and improving the efficiency of subsequent statistics and comparison.

[0148] The modal baseline module groups and statistically analyzes the low-dimensional correlation feature set according to the running state to generate a modal baseline, and compares it with the current low-dimensional correlation features to obtain a modal offset set. This process can reflect the changes in structural performance in real time and provide a benchmark for anomaly detection.

[0149] The matrix analysis module constructs a matrix representation based on the modal offset set to represent the relationship between the position of each sensor and the modal offset. Through matrix decomposition and reverse verification, it analyzes the contribution of each component to the modal anomaly, identifies the fault source component, and thus accurately locates potential fault points, improving the reliability of fault diagnosis.

[0150] Finally, the fault location module determines the structural area related to the offset based on the fault source component, providing maintenance personnel with targeted maintenance solutions, which can effectively reduce operational risks and extend equipment life.

[0151] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the health status of a treadmill, characterized in that, Includes the following steps: Structural response data from different locations of the treadmill during operation is acquired and time-aligned to obtain aligned data. Frequency band separation and interference suppression are then performed on the aligned data to form preprocessed data. Based on the preprocessed data, structural modal parameters, including frequency, damping, and mode vectors, are extracted to obtain a set of modal parameters; Perform multi-source correlation feature construction and dimensionality reduction on the modal parameter set to generate a low-dimensional correlation feature set; The low-dimensional correlation feature set is grouped and statistically analyzed according to the running state to generate a modal baseline, which is then compared with the current low-dimensional correlation features to obtain a modal offset set. Based on the set of modal offsets, a matrix is ​​constructed to represent the relationship between the position of each sensor and the modal offset. The contribution of each component to the modal anomaly is analyzed through matrix decomposition and reverse verification to determine the fault source component. Based on the fault source component, determine the structural region related to the offset.

2. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The process of acquiring structural response data from different locations of the treadmill during operation includes the following steps: Data collection begins simultaneously at multiple collection points on the treadmill; The structural response data collected at each location is recorded in real time; Add a time stamp to each piece of collected data; Save the recorded data from each collection location to the data buffer; The buffer data is integrated in chronological order to form a structured response data set.

3. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The frequency band separation and interference suppression of the aligned data includes the following steps: Determine the target frequency band range from the time-aligned data; Filter signal components that exceed the target frequency band; Smoothing or denoising is performed on high-frequency noise or low-frequency drift in the data. Standardize or adjust the amplitude of each data point; The processed data is used to form a preprocessed dataset for subsequent analysis.

4. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The extraction of structural modal parameters includes the following steps: The preprocessed data is divided according to the sampling time period; Perform spectral analysis on the data for each time period to identify the main vibration frequencies; Damping fitting is performed on the data corresponding to each major vibration frequency, and the damping ratio is calculated. The vibration mode of each mode is calculated based on the multi-channel response signals of each time period, forming a mode vector; The frequency, damping, and mode vectors for each time period are summarized to generate a set of modal parameters.

5. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The process of constructing and reducing the multi-source correlation features of the modal parameter set includes the following steps: The modal parameter set is organized into a feature matrix according to time period and acquisition channel; Standardize or normalize the modal parameters in the matrix; Calculate the correlation or covariance between modal parameters and construct a multi-source correlation feature matrix; Identify and remove redundant or highly correlated features; Apply dimensionality reduction methods to the remaining features to map the high-dimensional features to a low-dimensional space; Output a set of low-dimensional correlation features.

6. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The step of grouping and statistically analyzing the low-dimensional correlation feature set according to its operating status includes the following steps: Based on preset operational status indicators, the low-dimensional correlation feature set is divided into multiple groups; Perform statistical analysis on the feature vectors within each group, including calculating the mean and variance; The statistical results of each group are summarized to form a set of modal baselines for the corresponding operating states; The currently collected low-dimensional correlation features are grouped into corresponding groups based on the same operational status indicators.

7. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The construction matrix representing the relationship between the position of each sensor and the modal offset includes the following steps: Collect modal offset data for each sensor; Arrange the data in order of sensor location to form rows of a matrix; Use the offset of each modal feature or time period as the column of the matrix; The matrix data is standardized or normalized to form a feature matrix.

8. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The matrix decomposition includes the following steps: Input the constructed feature matrix into the matrix factorization algorithm; Decompose the matrix to obtain the fundamental matrix and coefficient matrix or singular values ​​and left and right singular vectors; Analyze the decomposition results to determine the contribution of each sensor or structural component to the modal offset; The output contribution matrix provides a preliminary judgment for fault source identification.

9. The method for monitoring the health status of a treadmill according to claim 1, characterized in that, The reverse verification includes the following steps: Based on the matrix factorization results, predict the contribution of each candidate fault source component to the modal offset of each sensor; The predicted results are compared with the actual modal offsets, and the error or residual is calculated. By combining time delay analysis, vibration propagation analysis, and modal vector contribution analysis, a comprehensive evaluation of candidate fault sources is conducted. The rationality of candidate fault sources is determined based on the comprehensive evaluation results, and candidate components that do not meet the conditions are eliminated. Output the set of fault source components after reverse verification and comprehensive evaluation.

10. A treadmill health status monitoring system, characterized in that, A treadmill health status monitoring method according to any one of claims 1-7, the system comprising: The data acquisition module is used to acquire structural response data from different positions of the treadmill during operation, perform time alignment to obtain aligned data, and perform frequency band separation and interference suppression on the aligned data to form preprocessed data. The modal extraction module is used to extract structural modal parameters, including frequency, damping and modal vectors, based on the preprocessed data to obtain a set of modal parameters; The feature construction module is used to perform multi-source correlation feature construction and dimensionality reduction on the modality parameter set to generate a low-dimensional correlation feature set; The modal baseline module is used to group and statistically analyze the low-dimensional correlation feature set according to the running state, generate a modal baseline, and compare it with the current low-dimensional correlation features to obtain a modal offset set. The matrix analysis module is used to construct a matrix representation of the relationship between the position of each sensor and the modal offset based on the set of modal offsets, and to analyze the contribution of each component to the modal anomaly through matrix decomposition and reverse verification to determine the fault source component. The fault location module is used to determine the structural region related to the offset based on the fault source component.

Images