High-speed rail bearing temperature and vibration multi-modal intelligent diagnosis method and system based on working condition spectrum

Abstract

The application provides a high-speed rail bearing temperature vibration multi-modal intelligent diagnosis method and system based on a working condition spectrum, relates to the technical field of intelligent diagnosis, and comprises synchronous collection of vibration, temperature distribution and phase angle signals of a bearing, extraction of an adaptive reference baseline combined with an operating condition. A temperature vibration combined sequence is obtained through phase alignment, a thermal vibration coupling strength spectrum is calculated and compared with the baseline, and a circumferential deviation distribution is obtained. After dynamic threshold correction, the out-of-limit phase interval is mapped to a fault space coordinate on a bearing raceway, and the fault type is determined according to the spectrum characteristics. The application realizes accurate spatial positioning and type identification of high-speed rail bearing faults, and improves the adaptability and accuracy of diagnosis.

Description

Technical Field

[0001] This invention relates to the field of intelligent diagnostic technology, and in particular to a method and system for intelligent diagnosis of temperature and vibration multimodal conditions of high-speed railway bearings based on operating condition spectrum diagrams. Background Technology

[0002] In the field of high-speed railway bearing condition monitoring and fault diagnosis, existing technologies typically rely on single-mode analysis of vibration or temperature signals. The conventional approach involves collecting vibration data using accelerometers mounted on the bearing housing, or acquiring local or overall bearing temperature information using infrared thermometers or a limited number of temperature sensors. These signals are then subjected to spectral analysis, envelope demodulation, or time-frequency transformation, and compared with preset fixed thresholds or benchmarks based on historical statistical data to determine whether the bearing exhibits abnormalities and the possible type of fault. Another common approach is to establish simple operating condition classification models under different train operating speeds or loads, setting independent vibration or temperature alarm thresholds for each condition to reduce the interference of operating condition changes on the diagnostic results.

[0003] However, the aforementioned conventional diagnostic methods have significant drawbacks. Firstly, the lack of a strict synchronous correlation mechanism between vibration and temperature signals during acquisition makes it difficult to accurately reveal their coupling relationship within the bearing's rotation cycle. The asynchronous nature of the impact events reflected by the vibration signals and the temperature field distribution prevents precise spatial correspondence, causing the diagnostic results to remain at the level of overall anomaly assessment, failing to pinpoint the specific spatial location of the fault on the bearing raceway circumference. Secondly, existing methods often employ fixed thresholds or static benchmarks based on coarse operating conditions, making it difficult to adapt to the dynamic effects of continuous changes in speed and load during train operation. This easily leads to false alarms under harsh operating conditions or missed alarms under stable operating conditions with minor faults, thus limiting the accuracy and reliability of the diagnosis. Summary of the Invention

[0004] This invention provides a method and system for intelligent diagnosis of temperature and vibration multimodal conditions of high-speed railway bearings based on operating condition spectrum diagrams, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a multi-modal intelligent diagnostic method for temperature and vibration of high-speed railway bearings based on operating condition spectrum diagrams, comprising:

[0006] The vibration signal, bearing circumferential temperature distribution sequence, and bearing rotation phase angle of the high-speed rail bearing are collected synchronously, and the corresponding train running speed and bearing load are recorded. Based on the train running speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from the pre-stored sub-condition reference library to form an adaptive reference baseline for the current condition.

[0007] Using the bearing rotation phase angle as the timing alignment reference, the vibration signal and the bearing circumferential temperature distribution sequence are synchronously aligned cycle by cycle in the phase domain to obtain a phase-aligned temperature-vibration joint sequence; the phase-aligned temperature-vibration joint sequence is segmented according to the rotation period, the vibration energy density and temperature rise amplitude in each phase interval are extracted, the thermal-vibration coupling strength of each phase interval is calculated, and the thermal-vibration coupling strength spectrum distributed along the bearing circumference is obtained;

[0008] The phase-by-phase interval deviation of the thermal vibration coupling intensity spectrum and the adaptive reference baseline is calculated to obtain the circumferential deviation distribution; the deviation judgment threshold is dynamically corrected according to the train running speed and bearing load, and the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold are mapped to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters.

[0009] The fault type is determined based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and the fault spatial coordinates and fault type identifier are output with adaptive operating conditions.

[0010] Based on the train's operating speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from the pre-stored sub-condition reference library to form an adaptive reference baseline for the current operating condition, including:

[0011] The pre-stored benchmark library for different operating conditions is indexed by several discrete combinations of train speed and bearing load. Based on the historical vibration signals and historical temperature field data of high-speed rail bearings collected under normal operating conditions, the statistical stable vibration spectrum and statistical stable temperature field spatial distribution under each discrete combination value are extracted and pre-stored as benchmark vibration spectrum entries and benchmark temperature field distribution entries corresponding to each discrete combination value.

[0012] The nearest discrete combined value node is located in the pre-stored sub-condition benchmark library index based on the current train speed and bearing load. When the deviation between the current condition parameters and the located node is within the preset tolerance range, the first vibration spectrum and the first temperature field distribution of the node are directly extracted. When the deviation exceeds the preset tolerance range, several nearest discrete combined value nodes are selected, and inverse proportional weighted interpolation is performed based on the distance between the current condition parameters and each node to obtain the second vibration spectrum and the second temperature field distribution.

[0013] The first vibration spectrum, the second vibration spectrum, the first temperature field distribution, and the second temperature field distribution are combined to form an adaptive reference baseline for the current operating condition.

[0014] The phase-aligned temperature-vibration joint sequence is segmented according to the rotation period. The vibration energy density and temperature rise amplitude in each phase interval are extracted, and the thermal-vibration coupling intensity in each phase interval is calculated to obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference, including:

[0015] The phase-aligned temperature and vibration joint sequence is divided into several phase intervals according to the phase range corresponding to one revolution of the bearing. The segmented vibration signal and segmented temperature sequence of each phase interval are extracted based on the start and end phase boundaries of each phase interval.

[0016] Time-frequency decomposition is performed on the segmented vibration signal of each phase interval to extract the integral of the vibration power spectrum within the phase angle range of the bearing characteristic frequency band, which is taken as the vibration energy density; for the segmented temperature sequence of each phase interval, the maximum positive deviation between the temperature measurement value at each circumferential position and the corresponding temperature reference value of the reference temperature field distribution is taken as the temperature rise amplitude.

[0017] The ratio of vibration energy density to the corresponding vibration reference value in the adaptive reference baseline is used as the normalized vibration deviation index, and the ratio of temperature rise amplitude to the corresponding temperature reference value in the reference temperature field distribution is used as the normalized temperature rise deviation index. The normalized vibration deviation index and the normalized temperature rise deviation index of each phase interval are multiplied, and the product result is used as the thermal vibration coupling intensity of that phase interval. The thermal vibration coupling intensities of all phase intervals are arranged in circumferential phase order to obtain the thermal vibration coupling intensity spectrum.

[0018] The deviation judgment threshold is dynamically corrected based on the train running speed and bearing load. The phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold are mapped to fault space coordinates on the bearing raceway surface using bearing raceway geometric parameters, including:

[0019] Using the current train speed and bearing load as a joint index, the corresponding threshold correction coefficient is extracted from the pre-stored threshold correction mapping relationship; the pre-stored static reference threshold is corrected according to the threshold correction coefficient to obtain the dynamically corrected threshold.

[0020] Traverse all phase intervals in the circumferential deviation distribution, extract the phase intervals where the circumferential deviation value exceeds the dynamically corrected threshold, and form a set of fault candidate phase intervals.

[0021] Extract the inner ring pitch circle radius, outer ring pitch circle radius, and bearing contact angle from the pre-stored bearing raceway geometry parameters; use the phase angle where the peak thermal vibration coupling intensity corresponding to each fault candidate phase interval is located as the angle input, and use the product of the phase angle and the inner ring pitch circle radius to determine the circumferential arc length position of the phase angle on the inner ring raceway surface, and use the product of the phase angle and the outer ring pitch circle radius to determine the circumferential arc length position of the phase angle on the outer ring raceway surface;

[0022] Based on the bearing contact angle, the circumferential arc length positions of the inner ring raceway surface and the outer ring raceway surface are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault space coordinates of each fault candidate phase interval on the bearing raceway surface.

[0023] Based on the bearing contact angle, the circumferential arc length positions of the inner and outer ring raceway surfaces are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault spatial coordinates of each candidate fault phase interval on the bearing raceway surface, including:

[0024] For the frequency components of the thermal vibration coupling intensity spectrum corresponding to each fault candidate phase interval, the proportion of vibration power in the inner ring fault characteristic frequency band, the outer ring fault characteristic frequency band, and the rolling element fault characteristic frequency band to the total power of the frequency components is calculated respectively; the fault characteristic frequency type corresponding to the frequency band with the highest proportion is used to determine the fault raceway type of each fault candidate phase interval.

[0025] For the fault candidate phase interval of the fault raceway type outer raceway, the product of the peak phase angle and the outer raceway pitch circle radius is used as the position of the outer raceway surface circumferential arc length.

[0026] For the fault candidate phase interval of the fault raceway type is the inner ring raceway, the peak phase angle is subtracted from the inner ring rotation offset determined by the current bearing rotation phase angle and the inner ring rotation speed ratio, and then multiplied by the inner ring pitch circle radius to obtain the position of the inner ring raceway surface circumferential arc length.

[0027] For the candidate phase interval of the fault raceway type of rolling element raceway, the position of the arc length on the surface of the rolling element is obtained by dividing the peak phase angle by the ratio of the rolling element's revolution to its rotational speed and then multiplying it by the rolling element radius.

[0028] The fault type is determined based on the frequency component characteristics of the thermal coupling intensity spectrum corresponding to the phase interval of the fault spatial coordinates, including:

[0029] The frequency values ​​of the main frequency component, the frequency values ​​of each secondary frequency component, and the amplitude ratio between each frequency component are extracted from the thermal vibration coupling intensity spectrum to form the frequency component feature vector of each phase interval.

[0030] The frequency component feature vector is matched dimension-by-dimensionally with each fault type feature template in the pre-stored fault type feature template library to calculate the feature deviation between the frequency component feature vector and each fault type feature template. Based on the principle of minimum feature deviation, the fault type corresponding to the fault type feature template with the minimum feature deviation is determined as the fault type corresponding to each fault space coordinate.

[0031] A second aspect of the present invention provides a high-speed rail bearing temperature and vibration multimodal intelligent diagnostic system based on operating condition spectrum diagrams, comprising:

[0032] The signal acquisition unit is used to synchronously acquire the vibration signal of the high-speed rail bearing, the bearing circumferential temperature distribution sequence and the bearing rotation phase angle, and record the corresponding train running speed and bearing load.

[0033] The baseline extraction unit is used to extract the corresponding reference vibration spectrum and reference temperature field distribution from the pre-stored sub-condition reference library based on the train running speed and bearing load, and to form an adaptive reference baseline for the current condition.

[0034] The timing alignment unit is used to synchronize the vibration signal and the bearing circumferential temperature distribution sequence in the phase domain with the bearing rotation phase angle as the timing alignment reference, so as to obtain a phase-aligned temperature-vibration joint sequence.

[0035] The coupling analysis unit is used to segment the phase-aligned temperature-vibration joint sequence according to the rotation period, extract the vibration energy density and temperature rise amplitude in each phase interval, calculate the thermal-vibration coupling intensity of each phase interval, and obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference.

[0036] The deviation calculation unit is used to perform phase-by-phase interval deviation calculation between the thermal vibration coupling intensity spectrum and the adaptive reference baseline to obtain the circumferential deviation distribution.

[0037] The fault location unit is used to dynamically correct the deviation judgment threshold based on the train running speed and bearing load, and to map the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters.

[0038] The fault identification unit is used to determine the fault type based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and output the fault spatial coordinates and fault type identifier that are adaptive to the operating conditions.

[0039] A third aspect of the present invention provides an electronic device, comprising:

[0040] processor;

[0041] Memory used to store processor-executable instructions;

[0042] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0043] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0044] This scheme extracts the vibration energy density and temperature rise amplitude within the phase interval and calculates the thermo-vibration coupling strength. It can sensitively capture the synergistic characteristics of periodic impact and frictional heat generation caused by local defects (such as spalling and cracks), enhancing the detection capability of early and subtle faults. A thermo-vibration coupling strength spectrum distributed along the bearing circumference is generated, visually presenting the spatial distribution of fault energy and providing a basis for preliminary fault location.

[0045] By comparing the thermal vibration coupling intensity spectrum with the adaptive reference baseline phase-by-phase interval deviation, abnormal features caused by faults can be effectively separated, suppressing the influence of inherent bearing characteristics and normal operating fluctuations. Dynamically correcting the deviation judgment threshold based on real-time operating conditions avoids false alarms or missed alarms under varying operating conditions with fixed thresholds, improving the adaptability and robustness of the diagnostic system. Mapping the out-of-limit phase interval to specific spatial coordinates on the bearing raceway surface achieves precise fault conversion from the signal domain to the physical spatial domain, providing clear location guidance for maintenance and repair.

[0046] By combining the spectral characteristics of the phase interval corresponding to the fault spatial coordinates, fault type identification is performed. This fully utilizes the differences in characteristic frequencies excited by different fault modes (such as pitting and scratching) to achieve accurate identification of the fault nature. The final output of the adaptive fault spatial coordinates and type identifier forms a complete diagnostic conclusion, significantly improving the intelligence level of high-speed railway bearing condition monitoring and the accuracy of maintenance decisions. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the intelligent diagnostic method for multi-modal temperature and vibration of high-speed railway bearings based on operating condition spectrum diagrams, according to an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 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.

[0049] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0050] Figure 1 This is a flowchart illustrating the intelligent diagnostic method for multi-modal temperature and vibration of high-speed railway bearings based on operating condition spectrum diagrams, as described in this embodiment of the invention. Figure 1 As shown, the method includes:

[0051] The vibration signal, bearing circumferential temperature distribution sequence, and bearing rotation phase angle of the high-speed rail bearing are collected synchronously, and the corresponding train running speed and bearing load are recorded. Based on the train running speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from the pre-stored sub-condition reference library to form an adaptive reference baseline for the current condition.

[0052] Using the bearing rotation phase angle as the timing alignment reference, the vibration signal and the bearing circumferential temperature distribution sequence are synchronously aligned cycle by cycle in the phase domain to obtain a phase-aligned temperature-vibration joint sequence; the phase-aligned temperature-vibration joint sequence is segmented according to the rotation period, the vibration energy density and temperature rise amplitude in each phase interval are extracted, the thermal-vibration coupling strength of each phase interval is calculated, and the thermal-vibration coupling strength spectrum distributed along the bearing circumference is obtained;

[0053] The phase-by-phase interval deviation of the thermal vibration coupling intensity spectrum and the adaptive reference baseline is calculated to obtain the circumferential deviation distribution; the deviation judgment threshold is dynamically corrected according to the train running speed and bearing load, and the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold are mapped to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters.

[0054] The fault type is determined based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and the fault spatial coordinates and fault type identifier are output with adaptive operating conditions.

[0055] For example, based on the train's operating speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from a pre-stored sub-condition reference library to form an adaptive reference baseline for the current operating condition, including:

[0056] The pre-stored benchmark library for different operating conditions is indexed by several discrete combinations of train speed and bearing load. Based on the historical vibration signals and historical temperature field data of high-speed rail bearings collected under normal operating conditions, the statistical stable vibration spectrum and statistical stable temperature field spatial distribution under each discrete combination value are extracted and pre-stored as benchmark vibration spectrum entries and benchmark temperature field distribution entries corresponding to each discrete combination value.

[0057] The nearest discrete combination value node is located in the pre-stored sub-condition benchmark library index based on the current train speed and bearing load.

[0058] When the deviation between the current operating parameters and the located node is within the preset tolerance range, the first vibration spectrum and the first temperature field distribution of the node are directly extracted.

[0059] When the deviation exceeds the preset tolerance range, select the nearest discrete combination value nodes, and perform inverse proportional weighted interpolation based on the distance between the current operating parameters and each node to obtain the second vibration spectrum and the second temperature field distribution respectively.

[0060] The first vibration spectrum, the second vibration spectrum, the first temperature field distribution, and the second temperature field distribution are combined to form an adaptive reference baseline for the current operating condition.

[0061] During the operation of high-speed rail bearings, different combinations of train speed and bearing load lead to significant differences in bearing vibration characteristics and temperature distribution. To achieve accurate condition monitoring, a benchmark library for different operating conditions needs to be established to store normal operating characteristic data under different conditions. The construction process of this benchmark library includes: firstly, determining the discrete value points of train speed, for example, setting a speed node every 20 km / h within the operating range of 60 km / h to 350 km / h, for a total of 15 speed levels; simultaneously determining the discrete value points of bearing load, dividing the bearing load into several levels such as unloaded, half-loaded, and fully loaded according to the actual passenger load of the high-speed train, specifically set as discrete values ​​every 5 kN within the range of 5 kN to 25 kN. Through the combination of speed and load, a multi-dimensional operating condition index grid is formed.

[0062] For each discrete combination value, vibration signals and temperature field data were continuously collected under normal operating conditions of the high-speed rail bearing. Vibration signal acquisition needed to cover a sufficiently long time window, typically no less than 30 days of continuous operating data, with a sampling frequency set to 25.6 kHz to capture the bearing's characteristic frequencies and their higher harmonic components. Temperature field data was acquired using temperature sensors placed at key locations such as the bearing's outer ring, inner ring, cage, and rolling elements, with a sampling interval of 1 second. A Fast Fourier Transform was performed on the collected historical vibration signals to obtain the frequency domain vibration spectrum. Statistical analysis was performed on the vibration spectra at different times; after removing abnormal fluctuation data, the mean and standard deviation of the amplitude at each frequency point were calculated, ultimately forming the statistically stable vibration spectrum under this operating condition. This spectrum reflects the vibration energy distribution law of the bearing under specific speed and load combinations under normal conditions.

[0063] For temperature field data processing, the temperature measurements from multiple sensors are used to construct a spatial distribution of the temperature field. Time-series analysis of historical data is performed to identify steady-state characteristics of temperature changes, eliminating the influence of transient conditions such as startup and braking, and extracting the mean steady-state temperature and its spatial gradient distribution at each measuring point under these conditions. Considering the influence of ambient temperature on bearing temperature, ambient temperature compensation is also required to convert the measured temperature into a temperature rise relative to the ambient temperature. After statistical processing, a statistically stable spatial distribution of the temperature field is obtained, including the temperature values ​​at key locations and their allowable fluctuation ranges.

[0064] The statistically stable vibration spectrum and statistically stable temperature field spatial distribution corresponding to each discrete combination value are stored as entries in a sub-condition benchmark library. The benchmark library adopts a multi-dimensional index structure, using train speed and bearing load as primary keys to achieve fast retrieval. Each entry contains a complete array of spectrum data, a temperature field distribution matrix, and corresponding statistical characteristic parameters.

[0065] After obtaining the real-time operating parameters of a current train speed of 185 km / h and a bearing load of 13.5 kN, the Euclidean distance between this point and all discrete nodes is calculated in the index grid of the benchmark library. Assuming there are four neighboring nodes in the benchmark library with speeds of 180 km / h and 190 km / h, and loads of 10 kN and 15 kN, the calculated distance between the current operating point and the node (180 km / h, 15 kN) is the smallest. The preset tolerance range is determined based on the sensitivity of the operating parameters; for example, the speed tolerance is set to ±3 km / h, and the load tolerance is set to ±1 kN.

[0066] Determine the deviations between the current operating parameters and the nearest nodes: the speed deviation is 185-180=5 km / h, exceeding the speed tolerance; the load deviation is 13.5-15=-1.5 kN, exceeding the load tolerance. Since the deviations exceed the preset tolerance range, interpolation calculations are required. Select the four nearest discrete combined value nodes surrounding the current operating point: node A (180 km / h, 15 kN), node B (190 km / h, 15 kN), node C (180 km / h, 10 kN), and node D (190 km / h, 10 kN).

[0067] Calculate the distance from the current working point to each node, and use the inverse proportional weighted interpolation method to calculate the weight coefficient of each node.

[0068] For the interpolation calculation of the vibration spectrum, the reference vibration spectrum of each node is weighted and summed one by one according to frequency points. This calculation process is repeated for all frequency points in the spectrum to obtain the complete second vibration spectrum. The same weighting coefficient is used for the interpolation calculation of the temperature field distribution.

[0069] For example, the phase-aligned temperature-vibration joint sequence is segmented according to the rotation period, the vibration energy density and temperature rise amplitude in each phase interval are extracted, and the thermal-vibration coupling intensity of each phase interval is calculated to obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference, including:

[0070] The phase-aligned temperature-vibration joint sequence is divided into several phase intervals according to the phase range corresponding to one revolution of the bearing.

[0071] Using the start and end phase boundaries of each phase interval as the interception benchmark, segmented vibration signals and segmented temperature sequences of each phase interval are extracted respectively.

[0072] Time-frequency decomposition is performed on the segmented vibration signal of each phase interval, and the integral of the vibration power spectrum within the characteristic frequency band of the bearing within the phase angle range of that phase interval is extracted as the vibration energy density.

[0073] For the segmented temperature sequence in each phase interval, the maximum positive deviation between the measured temperature value at each circumferential position and the reference temperature value at the corresponding position of the reference temperature field distribution is taken as the temperature rise amplitude.

[0074] The ratio of vibration energy density to the corresponding vibration reference value in the adaptive reference baseline is used as the normalized vibration deviation index, and the ratio of temperature rise amplitude to the corresponding temperature reference value in the reference temperature field distribution is used as the normalized temperature rise deviation index.

[0075] The normalized vibration deviation index and the normalized temperature rise deviation index are multiplied for each phase interval, and the product result is used as the thermal vibration coupling strength of that phase interval.

[0076] The thermal coupling intensity spectrum is obtained by arranging the thermal coupling intensity of all phase intervals in circumferential phase order.

[0077] After phase alignment, the temperature and vibration joint sequence needs to be further segmented to achieve precise mapping between the bearing's circumferential position and its health status. The phase range corresponding to one rotation of the bearing is 0° to 360°, and this range is divided into several phase intervals at equal intervals. The choice of interval must balance resolution and computational efficiency, typically using 5°, 10°, or 15° intervals. Taking a 10° interval as an example, the complete circumference is divided into 36 phase intervals, each covering a continuous 10° phase range. The first interval corresponds to 0° to 10°, the second interval to 10° to 20°, and so on until the 36th interval corresponds to 350° to 360°. This division method ensures uniform sampling of the bearing's circumferential position, avoiding the omission of critical abnormal areas.

[0078] For each phase interval, its start and end phase boundaries are used as the truncation benchmark for the time series. By finding all data points in the phase alignment sequence whose phase markers fall within a specific phase interval, the corresponding vibration signal segments and temperature measurement value sequences are extracted. For example, for a phase interval of 20° to 30°, the indexes of all times satisfying the condition 20° ≤ phase < 30° in the phase marker sequence are retrieved, and the corresponding vibration amplitudes are extracted from the vibration signal sequence according to these indices to form the segmented vibration signal for that interval; simultaneously, the temperature measurement values ​​at the corresponding times are extracted from the temperature sequence to form the segmented temperature sequence for that interval. Since the bearing will pass through the same phase interval multiple times during continuous operation, this segmented extraction process actually integrates the measurement data of the same circumferential position within multiple rotation cycles, providing a sufficient sample size for subsequent statistical analysis.

[0079] The segmented vibration signals extracted from each phase interval are subjected to time-frequency decomposition processing. Short-time Fourier transform or wavelet transform is used to convert the time-domain vibration signals into time-frequency energy distribution maps. In the time-frequency plane, the boundary range of the bearing's characteristic frequency band is defined, which covers the bearing's inner ring fault frequency, outer ring fault frequency, rolling element fault frequency, and their harmonic components. The typical frequency band width is 2 to 50 times the bearing's rotational frequency. Within the defined frequency band, the power spectral density values ​​at all times within the phase angle range of the phase interval are numerically integrated. The integration operation is accumulated along the phase axis, compressing the vibration energy distributed in the two-dimensional time-frequency space into a single scalar value. This value reflects the cumulative vibration intensity at a specific circumferential position within the observation period, serving as the vibration energy density of that phase interval. This processing eliminates the influence of transient fluctuations and extracts steady-state vibration characteristics.

[0080] For segmented temperature sequences within the same phase interval, the peak deviation extraction method is used to obtain the temperature rise amplitude. All temperature measurements within the phase interval are iterated, and the deviation between the current measurement and the corresponding circumferential temperature reference value in the reference temperature field distribution is calculated point by point. The deviation calculation uses subtraction, i.e., the measured temperature minus the reference temperature. Since the focus is on temperature rise caused by abnormal heat sources, only positive deviation values ​​are retained; negative deviations or deviations close to zero indicate that the temperature at that location is within the normal range and are not included in the anomaly assessment. The maximum value among all positive deviation values ​​in the phase interval is selected as the temperature rise amplitude for that interval. This maximum positive deviation represents the most significant temperature anomaly occurring at that circumferential location within the observation period, corresponding to the peak thermal effect of fault modes such as local friction, lubrication failure, or material defects.

[0081] To ensure comparability of different physical quantities, vibration energy density and temperature rise amplitude need to be normalized. For vibration energy density, the vibration reference value for the corresponding phase interval is retrieved from the adaptive reference baseline. This reference value represents the normal vibration level of the bearing at that circumferential position under healthy conditions. Dividing the currently observed vibration energy density by the vibration reference value yields the normalized vibration deviation index. This index is dimensionless; a value of 1 indicates that the vibration level is consistent with the reference, while a value greater than 1 indicates that the vibration intensity exceeds the normal range, with the degree of exceedance proportional to the index value. Similarly, normalization is performed on the temperature rise amplitude. The temperature reference value for the corresponding phase interval is extracted from the reference temperature field distribution, and the temperature rise amplitude is divided by the temperature reference value to obtain the normalized temperature rise deviation index. The temperature reference value is usually the average temperature at that position under healthy conditions. The normalized temperature rise deviation index is also dimensionless, intuitively reflecting the relative severity of temperature anomalies.

[0082] After obtaining the normalized vibration deviation index and normalized temperature rise deviation index for each phase interval, a product operation is performed to fuse dual-mode information. The two normalized indices for the same phase interval are multiplied, and the product result is taken as the thermo-oscillatory coupling strength for that phase interval. The physical meaning of the product operation is that the thermo-oscillatory coupling strength will only significantly increase when both vibration enhancement and temperature rise occur simultaneously at a certain circumferential position. If a position only exhibits abnormal vibration but normal temperature, or only abnormal temperature but normal vibration, its coupling strength value will be suppressed, thus effectively eliminating false alarms from a single mode. This nonlinear fusion mechanism enhances the sensitivity to real faults and improves the accuracy of diagnosis.

[0083] The thermal-vibration coupling intensities across all phase intervals are arranged in circumferential phase order to form a one-dimensional sequence, namely the thermal-vibration coupling intensity spectrum. The horizontal axis of this spectrum represents the circumferential phase angle, and the vertical axis represents the thermal-vibration coupling intensity value. The peak positions of the spectral lines correspond to the areas with the most significant anomalies on the bearing circumference, and the peak amplitude reflects the severity of the fault. Morphological analysis of the thermal-vibration coupling intensity spectrum can identify the circumferential location, influence range, and development trend of local faults, providing key characteristic basis for subsequent fault localization and classification. The entire processing flow realizes the transformation from raw multi-source time-series data to spatially distributed health indicators, significantly improving the spatial resolution capability of bearing condition monitoring.

[0084] For example, the deviation judgment threshold is dynamically corrected based on the train running speed and bearing load, and the phase interval in the circumferential deviation distribution that exceeds the dynamically corrected threshold is mapped to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters, including: extracting the corresponding threshold correction coefficient from the pre-stored threshold correction mapping relationship using the current train running speed and bearing load as a joint index.

[0085] The pre-stored static baseline threshold is corrected based on the threshold correction coefficient to obtain the dynamically corrected threshold.

[0086] Traverse all phase intervals in the circumferential deviation distribution, extract the phase intervals where the circumferential deviation value exceeds the dynamically corrected threshold, and form a set of fault candidate phase intervals.

[0087] Extract the inner ring pitch circle radius, outer ring pitch circle radius, and bearing contact angle from the pre-stored bearing raceway geometry parameters;

[0088] The phase angle of the peak thermal coupling intensity corresponding to each fault candidate phase interval is used as the angle input. The circumferential arc length position of the phase angle on the inner raceway surface is determined by the product of the phase angle and the inner raceway radius. The circumferential arc length position of the phase angle on the outer raceway surface is determined by the product of the phase angle and the outer raceway radius.

[0089] Based on the bearing contact angle, the circumferential arc length positions of the inner ring raceway surface and the outer ring raceway surface are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault space coordinates of each fault candidate phase interval on the bearing raceway surface.

[0090] When dynamically correcting the deviation judgment threshold, a multi-dimensional parameter joint indexing mechanism is used to achieve adaptive adjustment of the threshold. Specifically, a threshold correction mapping table is established with train speed and bearing load as dual index keys. This mapping table is pre-generated through extensive experimental data training. When threshold correction is required, the current train speed and bearing load values ​​are read and combined to form a joint index key. This joint index key is then used to query and retrieve the threshold correction coefficient corresponding to the current operating condition from the pre-stored threshold correction mapping table. This threshold correction coefficient reflects the influence of different operating conditions on fault judgment sensitivity and typically fluctuates between 0.7 and 1.3.

[0091] After obtaining the threshold correction coefficient, a pre-set static baseline threshold is retrieved from the database. This static baseline threshold is a judgment benchmark value determined through statistical analysis under standard operating conditions, representing the boundary between normal and abnormal states. During the correction calculation, the static baseline threshold and the threshold correction coefficient are multiplied to obtain a dynamically corrected threshold adapted to the current operating conditions. This correction mechanism ensures that the judgment threshold is appropriately raised under high-speed or heavy-load conditions to avoid misjudging normal operating condition fluctuations as fault characteristics; while under low-speed and light-load conditions, the judgment threshold is appropriately lowered to improve the ability to identify early and subtle faults.

[0092] After completing the dynamic threshold correction, the circumferential deviation distribution data is scanned and analyzed point by point. The circumferential deviation distribution is plotted with phase angle as the abscissa and thermal vibration coupling strength deviation value as the ordinate, covering a complete circumferential range from 0 degrees to 360 degrees. During the traversal, the circumferential deviation value of each phase interval is checked sequentially according to a preset angular resolution. When the circumferential deviation value of a certain phase interval exceeds the dynamically corrected threshold, the start and end angle information of that phase interval is recorded, marking it as a candidate fault phase interval. After traversing the entire circumferential range, all marked candidate fault phase intervals are summarized to form a set of candidate fault phase intervals. This set may contain multiple discrete phase intervals, each interval representing a potential fault location on the raceway surface.

[0093] For each phase interval in the set of candidate fault phase intervals, it needs to be mapped from the phase angle domain to the actual physical space coordinate system of the bearing raceway surface. This mapping process depends on the geometric parameters of the bearing. Three key parameters are extracted from the pre-stored bearing raceway geometric parameter library: inner ring pitch circle radius, outer ring pitch circle radius, and bearing contact angle. The inner ring pitch circle radius refers to the radius of the circle formed by the contact point between the inner ring raceway and the rolling element; the outer ring pitch circle radius refers to the radius of the circle formed by the contact point between the outer ring raceway and the rolling element; and the bearing contact angle refers to the angle between the normal direction of the contact line between the rolling element and the raceway and the radial plane of the bearing.

[0094] For each candidate fault phase interval, the phase angle at which the peak thermal coupling strength occurs is located; this phase angle represents the position where the fault characteristic is most significant. This phase angle is used as an angle input parameter to perform spatial coordinate mapping calculations for both the inner and outer raceways. For the inner raceway, the phase angle value is multiplied by the inner raceway pitch circle radius value; the result represents the circumferential arc length distance of the corresponding position on the inner raceway pitch circle. Similarly, the phase angle value is multiplied by the outer raceway pitch circle radius value to obtain the circumferential arc length distance of the corresponding position on the outer raceway pitch circle. This arc length calculation is based on the geometric relationship that the arc length equals the radius multiplied by the central angle, converting the angular domain information into the actual arc length position on the raceway circumference.

[0095] Because of the contact angle on the bearing raceway, the contact between the rolling elements and the raceway is not purely radial, but exhibits a certain degree of oblique contact. Therefore, it is necessary to decompose the circumferential arc length position based on the bearing contact angle. For the circumferential arc length position on the inner ring raceway surface, it is decomposed into two orthogonal components: a circumferential component and an axial component. The circumferential component is obtained by multiplying the circumferential arc length position by the cosine of the contact angle, representing the projection position of the fault point in the tangential direction of the raceway circumference. The axial component is obtained by multiplying the circumferential arc length position by the sine of the contact angle, representing the offset of the fault point in the bearing axial direction. The same decomposition process is applied to the circumferential arc length position on the outer ring raceway surface to obtain the corresponding circumferential and axial components of the outer ring raceway.

[0096] Through the coordinate decomposition described above, each candidate fault phase interval is precisely located as a two-dimensional coordinate pair on the bearing raceway surface. This coordinate pair consists of circumferential and axial components, forming fault spatial coordinates. For the inner raceway, this coordinate indicates the circumferential and axial offset of the fault point relative to the inner raceway reference position; for the outer raceway, this coordinate indicates the circumferential and axial offset of the fault point relative to the outer raceway reference position. This spatial coordinate representation transforms abstract phase angle information into intuitively understandable physical location information, providing a spatial reference for subsequent precise fault location and maintenance guidance. After mapping, all candidate fault phase intervals form a complete fault spatial coordinate set, which accurately describes the spatial distribution characteristics of each potential fault point on the bearing raceway surface, realizing full-chain traceability from thermal vibration coupling monitoring data to the physical location of the fault.

[0097] For example, based on the bearing contact angle, the circumferential arc length positions of the inner ring raceway surface and the outer ring raceway surface are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault spatial coordinates of each fault candidate phase interval on the bearing raceway surface, including:

[0098] For the frequency components of the thermal vibration coupling intensity spectrum corresponding to each fault candidate phase interval, the proportion of vibration power in the frequency band of the inner ring fault characteristic frequency band, the frequency band of the outer ring fault characteristic frequency band and the frequency band of the rolling element fault characteristic frequency band is statistically analyzed to account for the total power of the frequency components.

[0099] The fault roll path type for each fault candidate phase interval is determined by the fault characteristic frequency type corresponding to the frequency band with the highest proportion.

[0100] For the fault candidate phase interval of the fault raceway type outer raceway, the product of the peak phase angle and the outer raceway pitch circle radius is used as the position of the outer raceway surface circumferential arc length.

[0101] For the fault candidate phase interval of the fault raceway type is the inner ring raceway, the peak phase angle is subtracted from the inner ring rotation offset determined by the current bearing rotation phase angle and the inner ring rotation speed ratio, and then multiplied by the inner ring pitch circle radius to obtain the position of the inner ring raceway surface circumferential arc length.

[0102] For the candidate phase interval of the fault raceway type of rolling element raceway, the position of the arc length on the surface of the rolling element is obtained by dividing the peak phase angle by the ratio of the rolling element's revolution to its rotational speed and then multiplying it by the rolling element radius.

[0103] For the frequency components of the thermal vibration coupling intensity spectrum extracted from each candidate fault phase interval, the fault location on the raceway surface is converted from polar coordinates to a spatial rectangular coordinate system using the bearing contact angle. In this process, it is first necessary to understand the role of the contact angle in coordinate decomposition: the circumferential arc length can be decomposed into circumferential tangential components and axial components, which together form a joint coordinate pair as a spatial coordinate representation of the fault. This decomposition method establishes a mapping relationship from the phase domain to the physical space domain, providing a coordinate reference for subsequent precise positioning of the fault area on the raceway surface.

[0104] Before determining the spatial coordinates of the fault, it is necessary to identify the raceway type corresponding to each candidate fault phase interval. Frequency band power statistical analysis is performed on the frequency components of the acquired thermal-vibration coupling intensity spectrum, dividing the spectrum into three regions: the inner ring fault characteristic frequency band, the outer ring fault characteristic frequency band, and the rolling element fault characteristic frequency band. The inner ring fault characteristic frequency is typically a specific combination of rotational frequency, the number of rolling elements, and geometric parameters; the outer ring fault characteristic frequency is related to the frequency at which the rolling element passes through the outer ring raceway; and the rolling element fault characteristic frequency is determined by the rolling element's rotational frequency. The vibration power within each frequency band is accumulated and summed to calculate the proportion of that frequency band's power to the total power of the frequency components. If the power proportion of the inner ring frequency band is 45%, the outer ring frequency band is 30%, and the rolling element frequency band is 25%, then the fault raceway type of the candidate fault phase interval is determined to be the inner ring raceway. By comparing the power proportions of the three frequency bands, the frequency band corresponding to the highest proportion is selected, and its fault characteristic frequency type is used as the raceway type label for that phase interval.

[0105] For candidate phase intervals identified as outer raceway faults, spatial coordinate calculation is relatively straightforward. The outer raceway is typically stationary during bearing operation, and its fault location is independent of bearing rotation. The peak phase angle of this phase interval is extracted; this angle represents the instantaneous rotational position where the fault response is strongest within the vibration signal acquisition period. The outer raceway pitch circle radius is defined as the radius of the circle containing the outer raceway centerline, and its circumferential arc length on the raceway surface can be calculated. This arc length is further decomposed into circumferential and axial components, forming a spatial coordinate pair for the outer raceway fault. This coordinate pair directly reflects the position of the outer raceway surface defect in the bearing's fixed coordinate system.

[0106] For the candidate phase interval of the inner raceway fault, the calculation process needs to consider the rotational motion of the inner raceway. The inner raceway rotates with the shaft, and its fault location exhibits periodic changes in a fixed coordinate system. First, the peak phase angle is obtained, representing the peak position of the fault signal at the current measurement moment. However, due to the continuous rotation of the inner raceway, an inner raceway rotation offset needs to be introduced for compensation. The inner raceway rotation offset is determined by the current bearing rotation phase angle and the inner raceway rotation speed ratio, defined as the ratio of the inner raceway speed to the relative speed of the bearing housing; in most applications, this ratio is close to 1. Subtracting the rotation offset from the peak phase angle yields the corrected phase angle, which represents the relative position of the fault on the inner raceway. Multiplying the corrected phase angle by the inner raceway pitch circle radius yields the circumferential arc length of the inner raceway surface. This arc length is also decomposed into circumferential and axial components to establish a spatial coordinate system for the inner raceway fault. This compensation method eliminates the influence of inner raceway rotation on fault location, ensuring the accuracy of the fault location in the inner raceway body coordinate system.

[0107] For candidate phase intervals of rolling element raceway faults, the localization process involves the complex motion characteristics of the rolling elements. Within the bearing, the rolling element both revolves around the bearing center and rotates on its own axis; the phase angle at the fault location reflects the coupling effect of these two motions. After extracting the peak phase angle, decoupling is required using the speed ratio of the rolling element's revolution and rotation. This speed ratio is defined as the ratio of the rolling element's rotational angular velocity to its revolution angular velocity, typically derived from the geometric relationship between the rolling element diameter and its pitch circle diameter. Dividing the peak phase angle by the speed ratio yields a correction angle, which reflects the relative position of the fault on the rolling element surface. Multiplying the correction angle by the rolling element radius gives the arc length position on the rolling element surface. Since the rolling element is spherical or cylindrical, its surface coordinates can be described using a local spherical or cylindrical coordinate system. The arc length position is decomposed into tangential and radial components by the contact angle. For spherical rolling elements, a three-dimensional coordinate transformation is also required, combining the rolling element's spatial attitude angle within the cage, to ultimately determine the precise spatial coordinates of the fault point on the rolling element surface.

[0108] Through the above classification and processing strategy, a complete mapping link from the phase domain peak to the physical space coordinates is established for different raceway types.

[0109] For example, determining the fault type based on the frequency component characteristics of the thermal coupling intensity spectrum of the phase interval corresponding to the fault spatial coordinates includes:

[0110] The frequency values ​​of the main frequency component, the frequency values ​​of each secondary frequency component, and the amplitude ratio between each frequency component are extracted from the thermal vibration coupling intensity spectrum to form the frequency component feature vector of each phase interval.

[0111] The frequency component feature vector is matched dimension by dimension with each fault type feature template in the pre-stored fault type feature template library to calculate the feature deviation between the frequency component feature vector and each fault type feature template.

[0112] Based on the principle of minimum feature deviation, the fault type corresponding to the fault type feature template with the smallest feature deviation is determined as the fault type corresponding to each fault space coordinate.

[0113] After obtaining the thermal-oscillatory coupling intensity spectrum in different phase intervals, in-depth analysis of the spectral data is performed to extract frequency component features that characterize the fault characteristics. For the thermal-oscillatory coupling intensity spectrum in each phase interval, its dominant frequency component is first identified. The dominant frequency component refers to the frequency component with the largest amplitude in the thermal-oscillatory coupling intensity spectrum, which usually corresponds to the most significant thermal-oscillatory response mode caused by the fault. By traversing the amplitude distribution of the thermal-oscillatory coupling intensity spectrum using a peak detection algorithm, the frequency value corresponding to the global maximum amplitude point is located, and this frequency value is recorded as the dominant frequency component frequency.

[0114] After determining the dominant frequency component, the secondary frequency components are further extracted. Secondary frequency components refer to frequency elements with significant amplitudes in the thermal coupling intensity spectrum, excluding the dominant frequency. These frequency components may correspond to the harmonic response of the fault, coupled vibration modes, or secondary characteristics of heat wave propagation. An amplitude threshold is set between 15% and 30% of the dominant frequency amplitude, and all frequency points with amplitudes exceeding this threshold are identified as candidate secondary frequency components. The candidate secondary frequency components are then sorted from largest to smallest amplitude, and the top 5 to 8 components are extracted. This multi-frequency component extraction method can comprehensively capture the fault's performance characteristics across different frequency bands.

[0115] Besides the frequency values ​​themselves, the amplitude ratios between frequency components also contain important fault characteristic information. Calculating the amplitude ratio of each secondary frequency component relative to the primary frequency component involves dividing the amplitude A_i of the i-th secondary frequency component by the amplitude A_1 of the primary frequency component, yielding the amplitude ratio R_i. These amplitude ratios reflect the energy distribution relationship between different frequency components and have significant discriminative power for differentiating different types of faults. For example, mechanical wear faults typically exhibit a relatively balanced distribution of amplitudes among multiple secondary frequency components, while thermal imbalance faults often show the primary frequency component dominating and the secondary frequency component amplitudes rapidly decaying.

[0116] The extracted dominant frequency, secondary frequency values, and amplitude ratios are combined in a predetermined order to form the frequency component feature vector for that phase interval. This feature vector can be represented as a numerical sequence containing 12 to 18 dimensions, with the first half storing frequency value information and the second half storing amplitude ratio information. Through this structured feature representation, the originally complex thermal vibration coupling intensity spectrum is transformed into a feature vector form that is easy to calculate and compare.

[0117] To accurately identify fault types, a fault type feature template library needs to be established in advance. This library is formed by analyzing and summarizing historical data from a large number of known fault samples, and it stores frequency component feature templates for various typical faults. Each fault type corresponds to a feature template, and the structure of the template is consistent with the feature vector of the frequency component to be identified, also including feature dimensions such as the dominant frequency, secondary frequency, and amplitude ratio. Common fault types include bearing wear, rotor imbalance, uneven thermal expansion, poor lubrication, installation deviation, and structural fatigue. Each fault type has a standard feature template and an allowable feature fluctuation range stored in the template library.

[0118] During fault identification, the extracted frequency component feature vector is matched dimension-by-dimensionally with the fault type feature templates in the template library. For each corresponding dimension of the feature vector and template, the numerical difference between them is calculated. For the frequency dimension, the relative deviation between the frequency to be identified and the template frequency is calculated; for the amplitude ratio dimension, the absolute difference between the amplitude ratio to be identified and the amplitude ratio of the template is calculated. Since the physical meaning and numerical range of different dimensions differ, the deviations of each dimension need to be normalized to eliminate the influence of dimensions. The normalization method can be to divide the deviation of each dimension by the standard deviation value of that dimension in the template library, making the deviations of different dimensions comparable.

[0119] After calculating and normalizing the deviations of each dimension, the overall feature deviation is calculated through a weighted summation. Weight coefficients are assigned to different dimensions; the dominant frequency dimension typically receives a higher weight because dominant frequency features have a strong ability to distinguish fault types. The secondary frequency dimension and amplitude ratio dimension are assigned corresponding weights based on their sensitivity to specific fault types. The normalized deviations of each dimension are multiplied by their corresponding weights and then summed to obtain the feature deviation value of that frequency component feature vector relative to a fault type feature template. The smaller the feature deviation value, the higher the similarity between the feature to be identified and the fault type feature template.

[0120] The feature deviation between the feature vector of the frequency component to be identified and the feature templates of all fault types in the template library is calculated sequentially, forming a sequence of feature deviation values. The feature template with the smallest feature deviation value is then found in this sequence; the fault type feature template corresponding to this minimum value is the best matching template. Based on the principle of minimum feature deviation, the fault type represented by this best matching template is determined as the actual fault type corresponding to the current fault space coordinates. This fault identification method based on template matching and the minimum deviation criterion can fully utilize historical fault knowledge to achieve rapid and accurate identification of newly occurring faults.

[0121] For multiple fault spatial coordinate points, the aforementioned feature extraction and template matching processes are performed separately to determine the corresponding fault type for each spatial coordinate. When different spatial coordinate points identify the same fault type, it indicates that the fault has a certain spatial distribution range within the equipment; when adjacent spatial coordinate points identify different fault types, it suggests the possible occurrence of multiple concurrent faults or a spatial transformation of fault types. By comprehensively analyzing the fault type identification results of all spatial coordinate points, a comprehensive understanding of the equipment's fault state distribution can be obtained, providing detailed evidence for subsequent fault handling and maintenance decisions.

[0122] A second aspect of the present invention provides a high-speed rail bearing temperature and vibration multimodal intelligent diagnostic system based on operating condition spectrum diagrams, comprising:

[0123] The signal acquisition unit is used to synchronously acquire the vibration signal of the high-speed rail bearing, the bearing circumferential temperature distribution sequence and the bearing rotation phase angle, and record the corresponding train running speed and bearing load.

[0124] The baseline extraction unit is used to extract the corresponding reference vibration spectrum and reference temperature field distribution from the pre-stored sub-condition reference library based on the train running speed and bearing load, and to form an adaptive reference baseline for the current condition.

[0125] The timing alignment unit is used to synchronize the vibration signal and the bearing circumferential temperature distribution sequence in the phase domain with the bearing rotation phase angle as the timing alignment reference, so as to obtain a phase-aligned temperature-vibration joint sequence.

[0126] The coupling analysis unit is used to segment the phase-aligned temperature-vibration joint sequence according to the rotation period, extract the vibration energy density and temperature rise amplitude in each phase interval, calculate the thermal-vibration coupling intensity of each phase interval, and obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference.

[0127] The deviation calculation unit is used to perform phase-by-phase interval deviation calculation between the thermal vibration coupling intensity spectrum and the adaptive reference baseline to obtain the circumferential deviation distribution.

[0128] The fault location unit is used to dynamically correct the deviation judgment threshold based on the train running speed and bearing load, and to map the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters.

[0129] The fault identification unit is used to determine the fault type based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and output the fault spatial coordinates and fault type identifier that are adaptive to the operating conditions.

[0130] A third aspect of the present invention provides an electronic device, comprising:

[0131] processor;

[0132] Memory used to store processor-executable instructions;

[0133] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0134] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0135] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-modal intelligent diagnostic method for temperature and vibration of high-speed railway bearings based on operating condition spectrum diagrams, characterized in that, include: The vibration signal, bearing circumferential temperature distribution sequence, and bearing rotation phase angle of the high-speed rail bearing are collected synchronously, and the corresponding train running speed and bearing load are recorded. Based on the train running speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from the pre-stored sub-condition reference library to form an adaptive reference baseline for the current condition. Using the bearing rotation phase angle as the timing alignment reference, the vibration signal and the bearing circumferential temperature distribution sequence are synchronously aligned cycle by cycle in the phase domain to obtain a phase-aligned temperature-vibration joint sequence. The phase-aligned temperature-vibration joint sequence is segmented according to the rotation period. The vibration energy density and temperature rise amplitude in each phase interval are extracted, and the thermal-vibration coupling intensity in each phase interval is calculated to obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference, including: The phase-aligned temperature and vibration joint sequence is divided into several phase intervals according to the phase range corresponding to one revolution of the bearing. The segmented vibration signal and segmented temperature sequence of each phase interval are extracted based on the start and end phase boundaries of each phase interval. Time-frequency decomposition is performed on the segmented vibration signal of each phase interval to extract the integral of the vibration power spectrum within the phase angle range of the bearing characteristic frequency band, which is taken as the vibration energy density; for the segmented temperature sequence of each phase interval, the maximum positive deviation between the temperature measurement value at each circumferential position and the corresponding temperature reference value of the reference temperature field distribution is taken as the temperature rise amplitude. The ratio of vibration energy density to the corresponding vibration reference value in the adaptive reference baseline is used as the normalized vibration deviation index, and the ratio of temperature rise amplitude to the corresponding temperature reference value in the reference temperature field distribution is used as the normalized temperature rise deviation index. The normalized vibration deviation index and the normalized temperature rise deviation index of each phase interval are multiplied, and the product result is used as the thermal vibration coupling intensity of that phase interval. The thermal vibration coupling intensities of all phase intervals are arranged in circumferential phase order to obtain the thermal vibration coupling intensity spectrum. The phase-by-phase interval deviation of the thermal vibration coupling intensity spectrum and the adaptive reference baseline is calculated to obtain the circumferential deviation distribution; the deviation judgment threshold is dynamically corrected according to the train running speed and bearing load, and the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold are mapped to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters. The fault type is determined based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and the fault spatial coordinates and fault type identifier are output with adaptive operating conditions.

2. The method according to claim 1, characterized in that, Based on the train's operating speed and bearing load, the corresponding reference vibration spectrum and reference temperature field distribution are extracted from the pre-stored sub-condition reference library to form an adaptive reference baseline for the current operating condition, including: The pre-stored benchmark library for different operating conditions is indexed by several discrete combinations of train speed and bearing load. Based on the historical vibration signals and historical temperature field data of high-speed rail bearings collected under normal operating conditions, the statistical stable vibration spectrum and statistical stable temperature field spatial distribution under each discrete combination value are extracted and pre-stored as benchmark vibration spectrum entries and benchmark temperature field distribution entries corresponding to each discrete combination value. The nearest discrete combined value node is located in the pre-stored sub-condition benchmark library index based on the current train speed and bearing load. When the deviation between the current condition parameters and the located node is within the preset tolerance range, the first vibration spectrum and the first temperature field distribution of the node are directly extracted. When the deviation exceeds the preset tolerance range, several nearest discrete combined value nodes are selected, and inverse proportional weighted interpolation is performed based on the distance between the current condition parameters and each node to obtain the second vibration spectrum and the second temperature field distribution. The first vibration spectrum, the second vibration spectrum, the first temperature field distribution, and the second temperature field distribution are combined to form an adaptive reference baseline for the current operating condition.

3. The method according to claim 1, characterized in that, The deviation judgment threshold is dynamically corrected based on the train running speed and bearing load. The phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold are mapped to fault space coordinates on the bearing raceway surface using bearing raceway geometric parameters, including: Using the current train speed and bearing load as a joint index, the corresponding threshold correction coefficient is extracted from the pre-stored threshold correction mapping relationship; the pre-stored static reference threshold is corrected according to the threshold correction coefficient to obtain the dynamically corrected threshold. Traverse all phase intervals in the circumferential deviation distribution, extract the phase intervals where the circumferential deviation value exceeds the dynamically corrected threshold, and form a set of fault candidate phase intervals. Extract the inner ring pitch circle radius, outer ring pitch circle radius, and bearing contact angle from the pre-stored bearing raceway geometry parameters; use the phase angle where the peak thermal vibration coupling intensity corresponding to each fault candidate phase interval is located as the angle input, and use the product of the phase angle and the inner ring pitch circle radius to determine the circumferential arc length position of the phase angle on the inner ring raceway surface, and use the product of the phase angle and the outer ring pitch circle radius to determine the circumferential arc length position of the phase angle on the outer ring raceway surface; Based on the bearing contact angle, the circumferential arc length positions of the inner ring raceway surface and the outer ring raceway surface are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault space coordinates of each fault candidate phase interval on the bearing raceway surface.

4. The method according to claim 3, characterized in that, Based on the bearing contact angle, the circumferential arc length positions of the inner and outer ring raceway surfaces are decomposed into circumferential and axial components, respectively. The joint coordinate pair of the circumferential and axial components is used as the fault spatial coordinates of each candidate fault phase interval on the bearing raceway surface, including: For the frequency components of the thermal vibration coupling intensity spectrum corresponding to each fault candidate phase interval, the proportion of vibration power in the inner ring fault characteristic frequency band, the outer ring fault characteristic frequency band, and the rolling element fault characteristic frequency band to the total power of the frequency components is calculated respectively; the fault characteristic frequency type corresponding to the frequency band with the highest proportion is used to determine the fault raceway type of each fault candidate phase interval. For the fault candidate phase interval of the fault raceway type outer raceway, the product of the peak phase angle and the outer raceway pitch circle radius is used as the position of the outer raceway surface circumferential arc length. For the fault candidate phase interval of the fault raceway type is the inner ring raceway, the peak phase angle is subtracted from the inner ring rotation offset determined by the current bearing rotation phase angle and the inner ring rotation speed ratio, and then multiplied by the inner ring pitch circle radius to obtain the position of the inner ring raceway surface circumferential arc length. For the candidate phase interval of the fault raceway type of rolling element raceway, the position of the arc length on the surface of the rolling element is obtained by dividing the peak phase angle by the ratio of the rolling element's revolution to its rotational speed and then multiplying it by the rolling element radius.

5. The method according to claim 1, characterized in that, The fault type is determined based on the frequency component characteristics of the thermal coupling intensity spectrum corresponding to the phase interval of the fault spatial coordinates, including: The frequency values ​​of the main frequency component, the frequency values ​​of each secondary frequency component, and the amplitude ratio between each frequency component are extracted from the thermal vibration coupling intensity spectrum to form the frequency component feature vector of each phase interval. The frequency component feature vector is matched dimension-by-dimensionally with each fault type feature template in the pre-stored fault type feature template library to calculate the feature deviation between the frequency component feature vector and each fault type feature template. Based on the principle of minimum feature deviation, the fault type corresponding to the fault type feature template with the minimum feature deviation is determined as the fault type corresponding to each fault space coordinate.

6. A high-speed rail bearing temperature and vibration multimodal intelligent diagnostic system based on operating condition spectrum diagrams, used to implement the method as described in any one of claims 1-5, characterized in that, include: The signal acquisition unit is used to synchronously acquire the vibration signal of the high-speed rail bearing, the bearing circumferential temperature distribution sequence and the bearing rotation phase angle, and record the corresponding train running speed and bearing load. The baseline extraction unit is used to extract the corresponding reference vibration spectrum and reference temperature field distribution from the pre-stored sub-condition reference library based on the train running speed and bearing load, and to form an adaptive reference baseline for the current condition. The timing alignment unit is used to synchronize the vibration signal and the bearing circumferential temperature distribution sequence in the phase domain with the bearing rotation phase angle as the timing alignment reference, so as to obtain a phase-aligned temperature-vibration joint sequence. The coupling analysis unit is used to segment the phase-aligned temperature-vibration joint sequence according to the rotation period, extract the vibration energy density and temperature rise amplitude in each phase interval, calculate the thermal-vibration coupling intensity of each phase interval, and obtain the thermal-vibration coupling intensity spectrum distributed along the bearing circumference. The deviation calculation unit is used to perform phase-by-phase interval deviation calculation between the thermal vibration coupling intensity spectrum and the adaptive reference baseline to obtain the circumferential deviation distribution. The fault location unit is used to dynamically correct the deviation judgment threshold based on the train running speed and bearing load, and to map the phase intervals in the circumferential deviation distribution that exceed the dynamically corrected threshold to the fault space coordinates on the bearing raceway surface using the bearing raceway geometric parameters. The fault identification unit is used to determine the fault type based on the frequency component characteristics of the thermal vibration coupling intensity spectrum in the phase interval corresponding to the fault spatial coordinates, and output the fault spatial coordinates and fault type identifier that are adaptive to the operating conditions.

7. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 5.

Images