Handheld integrated inspection instrument and inspection method thereof

Abstract

The invention discloses a handheld integrated itinerant detector and an itinerant detection method thereof. The itinerant detector comprises a multi-source sensing unit and a magnetic attraction positioning module, the inspection method comprises the following steps: establishing a multi-dimensional calibration parameter of a target device, adsorbing on the surface of the target device based on a magnetic attraction positioning module, loading a spatial layout, and generating an environment calibration model; synchronously operating the far / close distance temperature measurement module and the vibration sensing module, and outputting multi-source fusion data; performing dynamic correlation analysis, analyzing vibration spectrum energy distribution by using a feature extraction algorithm, constructing a vibration-temperature coupling feature vector in combination with temperature gradient change, inputting a fault diagnosis result of a pre-trained fault diagnosis model, generating a visual early warning map, and feeding back the early warning map in real time through an interactive interface; according to the method, the limitation of single parameter monitoring is overcome, the detection precision is remarkably improved, and a real-time, rapid and accurate fault detection result is provided through self-adaptive switching of a temperature measurement mode and a multi-frequency-band vibration acquisition strategy.

Description

Technical Field

[0001] The present invention relates to the field of patrol inspection technology, and in particular to a handheld integrated patrol inspection instrument and a patrol inspection method thereof. Background Art

[0002] In industrial manufacturing, mechanical equipment (such as wind turbines and reactors) operates under harsh conditions of high temperature, high pressure, and high vibration for extended periods. Fatigue damage and localized overheating of key components (such as bearings, gearboxes, and motors) can easily lead to sudden failures, resulting in production interruptions and even accidents. Traditional manual inspections, which rely on empirical judgment and random checks using distributed instruments, suffer from low efficiency, incomplete coverage, and difficulty tracing data. This makes it difficult to achieve real-time, accurate fault warnings, especially in confined spaces or high-risk environments.

[0003] Existing handheld fault detection devices mostly use a single parameter monitoring mode. For example, infrared thermometers can only obtain the surface temperature distribution of the device, while vibration analyzers are limited to collecting mechanical vibration spectra. Although some integrated devices attempt to combine temperature and vibration sensors, their data processing remains at the independent parameter threshold alarm stage, lacking dynamic correlation analysis between temperature gradient changes and vibration spectrum energy distribution. This data fragmentation makes it impossible to effectively identify complex faults caused by the combined effects of thermal stress and mechanical shock (such as local overheating of bearings caused by lubrication failure and the superposition of high-frequency shock), resulting in missed detections or increased false detection rates, seriously restricting fault diagnosis accuracy and operational efficiency in complex industrial scenarios.

[0004] Therefore, there is an urgent need for a portable, intelligent, integrated detection method to improve the reliability and response speed of industrial equipment status monitoring. Summary of the Invention

[0005] The purpose of the present invention is to provide a handheld integrated inspection instrument and an inspection method thereof to solve the above technical problems.

[0006] To achieve this object, the present invention adopts the following technical solutions: A patrol inspection method for a handheld integrated patrol inspection instrument, the patrol inspection instrument comprising a multi-source sensing unit and a magnetic positioning module, the multi-source sensing unit comprising a vibration sensing module and a long / short distance temperature measurement module; the patrol inspection method comprising the following steps: S1, establishing multi-dimensional calibration parameters of the target device through the multi-source sensing unit of the inspection instrument, based on the magnetic positioning module adsorbed on the surface of the target device and loading the spatial layout map, generating an environmental calibration model including temperature and vibration reference values; S2, synchronously running the long / short distance temperature measurement module and the vibration sensing module, adaptively switching between long-range temperature field scanning and short-range hotspot tracking modes according to the reference value in the environmental calibration model, and collecting vibration signals in a time-sharing manner based on a preset frequency band, and outputting multi-source fusion data with aligned timestamps; S3, performing dynamic correlation analysis on the multi-source fusion data, analyzing the vibration spectrum energy distribution using a feature extraction algorithm, constructing a vibration-temperature coupling feature vector based on the temperature gradient change, and inputting the vector into a pre-trained fault diagnosis model to generate a fault diagnosis result with a fault type and confidence level; S4, based on the spatial mapping relationship between the fault diagnosis result and the environmental calibration model, generate a visual warning map, and provide real-time feedback of the warning map through an interactive interface; the warning map includes a thermal identification of the fault location, a spectrum abnormality interval, and a confidence level.

[0007] Optionally, establishing multi-dimensional calibration parameters of the target device by the multi-source sensing unit of the inspection instrument specifically includes the following steps: S11, the flexible contact surface of the magnetic positioning module is attached to the surface of the target device, triggering the dynamic pressure detection unit to provide real-time feedback on the matching degree between the adsorption force and the curvature of the contact surface. When the adsorption force reaches a preset threshold, a positioning signal is activated and a magnetic state parameter is generated. S12, loading a spatial layout diagram of the target device based on the positioning signal, identifying the real-time position of the inspection instrument in the spatial layout diagram through the gyroscope in the multi-source sensing unit, and establishing a spatial mapping relationship between the surface coordinate system of the inspection instrument and the spatial layout diagram; S13, under the said spatial mapping relationship, controls the long / short distance temperature measurement module to obtain the initial temperature distribution matrix of the target device surface in the reference scanning mode, and synchronously triggers the vibration sensing module to collect the environmental background vibration spectrum in the low-frequency steady-state mode to generate an initial reference parameter set of temperature and vibration.

[0008] Optionally, the synchronously triggered vibration sensing module collects the ambient background vibration spectrum in a low-frequency steady-state mode to generate an initial reference parameter set of temperature and vibration, and then further includes: S14, based on the initial reference parameter set, analyzing the temperature field uniformity and vibration background noise characteristics through the calibration algorithm of the multi-source sensing unit, dynamically adjusting the emissivity compensation coefficient of the corresponding temperature measurement module and the filter frequency band threshold of the vibration sensing module, and constructing multi-dimensional calibration parameters; S15, integrating the multi-dimensional calibration parameters with the spatial mapping relationship, and performing weighted interpolation based on typical operating condition data in the historical fault database to generate an environmental calibration model including a temperature gradient tolerance range, a vibration spectrum baseline, and a fault-sensitive area identifier.

[0009] Optionally, the long / short distance temperature measurement module and the vibration sensing module are operated synchronously, and the long-distance temperature field scanning mode and the short-distance hotspot tracking mode are adaptively switched according to the reference value in the environmental calibration model. The vibration signal is collected in a time-sharing manner based on a preset frequency band, and multi-source fusion data with aligned timestamps is output. Specifically, the following steps are included: S21, based on the temperature gradient tolerance range and vibration spectrum baseline in the environmental calibration model, calculating the initial resolution of the long-range temperature field scanning and the trigger threshold of the short-range hotspot tracking, and generating multi-mode switching decision parameters; S22, based on the multi-mode switching decision parameters, synchronously start the temperature field scanning thread of the long / short distance temperature measurement module and the frequency band acquisition thread of the vibration sensing module through a dynamic priority scheduling algorithm, and allocate independent clock sources to ensure timing synchronization of the two modules; S23, in the long-range temperature field scanning mode, controls the temperature measurement module to cover the target device surface with a rotational path, compares the scanned temperature with the temperature gradient tolerance range in the environmental calibration model in real time, and switches to the short-range hotspot tracking mode for focused temperature measurement when it is detected that the local temperature deviation exceeds the trigger threshold.

[0010] Optionally, the real-time comparison of the scanning temperature with the temperature gradient tolerance range in the environmental calibration model, when it is detected that the local temperature deviation exceeds the trigger threshold, switches to the close-range hotspot tracking mode for focused temperature measurement, and then further includes: S24, synchronously with the temperature measurement mode switching, controls the vibration sensing module to divide the vibration signal acquisition into a low-frequency steady-state segment and a high-frequency transient segment based on a preset frequency band division strategy, and alternately activates the pressure detection units in the corresponding frequency bands using a time-division multiplexing method; S25, synchronizes and marks the temperature gradient of the temperature measurement module and the vibration time domain signal of the vibration sensing module through the timestamp alignment unit, and based on the spatial mapping relationship of the inspection instrument surface coordinate system, fuses and encodes the temperature-vibration data according to the spatial grid coordinates, and outputs a multi-source fusion data stream with timestamp alignment and spatial coordinate binding.

[0011] Optionally, dynamic correlation analysis is performed on the multi-source fusion data, a feature extraction algorithm is used to analyze the vibration spectrum energy distribution, a vibration-temperature coupling feature vector is constructed in combination with the temperature gradient change, and a pre-trained fault diagnosis model is input to generate a fault diagnosis result with fault type and confidence level, specifically including: S31, performing adaptive noise reduction processing on the vibration time domain signal in the multi-source fusion data stream by the built-in data processing module of the inspection instrument, separating the effective vibration component from the background noise based on the vibration spectrum baseline in the environmental calibration model, and generating a denoised vibration spectrum sequence; S32. Calculate the band energy entropy of the vibration spectrum sequence using the multi-scale wavelet packet decomposition algorithm, extract the characteristics of the high-frequency impact energy ratio, low-frequency resonance main frequency offset, and energy distribution dispersion, and construct a vibration multi-dimensional feature vector; S33. Synchronously analyze the temperature gradient change rate in the multi-source fusion data stream, combine it with the temperature gradient tolerance interval of the environmental calibration model, calculate the local temperature rise rate and the thermal diffusion uniformity index, and generate a temperature dynamic feature vector; S34. Align the vibration multi-dimensional feature vector and the temperature dynamic feature vector according to the time stamp, perform coupling analysis through the dynamic weight allocation layer in the pre-trained fault diagnosis model, and output the fault type, location identifier, and confidence score to form a fault diagnosis result.

[0012] Optionally, generate a visual warning map based on the spatial mapping relationship between the fault diagnosis result and the environmental calibration model, and real-time feedback the warning map through the interaction interface; the warning map includes the thermal identification of the fault location, the abnormal spectrum interval, and the confidence level, specifically including: S41. Based on the spatial mapping relationship of the environmental calibration model and the location identifier in the fault diagnosis result, spatially align the surface coordinate system of the inspection instrument with the location identifier to generate three-dimensional point cloud data of the fault thermal distribution; S42. Perform multi-layer overlay rendering on the three-dimensional point cloud data of the fault thermal distribution and the abnormal vibration spectrum interval, and perform transparency grading coloring on the abnormal area in combination with the confidence level to generate a visual warning map containing temperature-vibration coupling abnormal characteristics; S43. Adaptively adjust the resolution of the visual warning map through the dynamic compression algorithm built in the interaction interface, load the map slices, and synchronously display the frequency domain comparison curve of the abnormal spectrum interval on the terminal screen; S44. Receive the annotation instruction or confidence correction parameter input by the user through the interaction interface, real-time update the fault level identifier in the warning map, and transmit the corrected data back to the fault diagnosis model for online weight fine-tuning.

[0013] Optionally, after generating a visual warning map based on the spatial mapping relationship between the fault diagnosis result and the environmental calibration model, and real-time feedback the warning map through the interaction interface, it further includes: S5. Upload the multi-source fusion data, warning map, and fault prompt to the cloud database, and iteratively optimize the reference value of the environmental calibration model and the correlation threshold of the fault diagnosis model based on historical data to form a closed-loop self-learning monitoring link.

[0014] The present invention also provides a handheld integrated inspection instrument, which realizes detection by adopting the inspection method of the handheld integrated inspection instrument as described above. The handheld integrated inspection instrument specifically includes a rod body, which is a segmented telescopic structure. The front end of the rod body integrates a multi-source sensing unit and a magnetic adsorption positioning module. The multi-source sensing unit includes a far / near distance temperature measurement module and a vibration sensing module; The magnetic adsorption positioning module is arranged at the bottom of the front end of the rod body and includes a flexible contact surface, a permanent magnet and a pressure detection unit, and is used for adsorbing the surface of the target device and feeding back the contact state; A data processing module and an inertial navigation unit are built in the middle section of the rod body. The inertial navigation unit includes a gyroscope and an accelerometer; An interaction unit and a detachable power module are arranged at the rear end of the rod body. The interaction unit is provided with a touch screen and a button part.

[0015] Compared with the prior art, the present invention has the following beneficial effects: Based on the multi-source sensing unit and the magnetic adsorption positioning module of the handheld integrated inspection instrument, first, an environmental calibration model including temperature and vibration reference values is established by adsorbing the surface of the target device and loading the space layout diagram; then, the far-distance temperature field scanning and the near-distance hot spot tracking modes are adaptively switched, the vibration signals are collected in different frequency bands, and the multi-source fusion data with time stamps aligned is output; the fusion data is dynamically correlated and analyzed by the data processing module, the vibration spectrum energy distribution characteristics are extracted, and the vibration-temperature coupling feature vector is constructed in combination with the temperature gradient, and the fault type and confidence result are generated by inputting into the pre-trained model; the diagnosis result is mapped to the space model to generate a visual warning map, and the comprehensive information including the fault position, the abnormal spectrum interval and the confidence level is fed back in real time through the interaction interface; this method utilizes the coupling characteristics of the vibration spectrum energy distribution and the temperature gradient change, overcomes the limitations of single-parameter monitoring, significantly improves the detection accuracy, and optimizes the data acquisition efficiency and improves the instantaneity efficiency by adaptively switching the temperature measurement mode and the vibration acquisition strategy in different frequency bands, providing real-time, fast and accurate fault detection results. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] The structures, proportions, sizes, etc. shown in the attached drawings of this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the conditions for the implementation of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0018] Figure 1 It is a schematic flowchart of the inspection method of the handheld integrated inspection instrument in the first embodiment; Figure 2 It is a front view structural schematic diagram of the handheld integrated inspection instrument in the second embodiment; Figure 3 It is an axonometric view structural schematic diagram of the handheld integrated inspection instrument in the second embodiment; Figure 4 It is a side view structural schematic diagram of the handheld integrated inspection instrument in the second embodiment. Detailed implementation manners

[0019] To make the invention objectives, features, and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the attached drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0020] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "upper", "lower", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the attached drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. It should be noted that when a component is considered to be "connected" to another component, it may be directly connected to the other component or there may be intermediate components present simultaneously.

[0021] The technical solutions of the present invention will be further described below in conjunction with the attached drawings and through specific implementation manners.

[0022] Embodiment 1: Combined with Figure 1 As shown, the embodiment of the present invention provides an inspection method for a handheld integrated inspection instrument. The inspection instrument includes a multi-source sensing unit 20 and a magnetic adsorption positioning module 30. The multi-source sensing unit 20 includes a vibration sensing module 22 and a far / near distance temperature measurement module 21; the inspection method includes the following steps: S1, establish multi-dimensional calibration parameters of the target device through the multi-source sensing unit 20 of the inspection instrument, adsorb the magnetic positioning module 30 on the surface of the target device and load the spatial layout map to generate an environmental calibration model including temperature and vibration reference values; The magnetic positioning module 30 achieves spatial calibration and adsorption fixation on the device surface, ensuring that the multi-source sensors maintain consistent physical positioning throughout each inspection. The environmental calibration model integrates the device's normal operating temperature distribution baseline (coordinated by the near and far temperature measurement modules) and vibration spectrum characteristics (based on a pre-set frequency band) to establish a multidimensional parameter mapping network corresponding to the device's physical structure. This step addresses the poor data comparability caused by sensor position offset in traditional mobile detection, providing a stable spatial-parameter correlation benchmark for subsequent dynamic analysis.

[0023] S2: Synchronously operate the long-range / short-range temperature measurement module 21 and the vibration sensing module 22, adaptively switch between the long-range temperature field scanning and short-range hotspot tracking modes according to the reference value in the environmental calibration model, and collect vibration signals in a time-sharing manner based on the preset frequency band, and output multi-source fusion data with aligned timestamps; An adaptive switching strategy between long-range temperature field scanning (wide-area temperature trend identification) and short-range hotspot tracking (localized focus on abnormal temperature rises), combined with a frequency-segmented, time-series acquisition mechanism for vibration signals (such as continuous monitoring in the low-frequency band and triggered sampling in the high-frequency band), reduces data volume while ensuring the capture of key features. Hardware synchronization circuits and timestamp alignment algorithms eliminate data phase differences caused by physical delays in multiple sensors, ensuring a strict temporal correlation between temperature gradient changes and vibration spectrum evolution, laying the data foundation for cross-domain feature fusion.

[0024] S3 performs dynamic correlation analysis on multi-source fusion data, uses feature extraction algorithms to analyze the vibration spectrum energy distribution, and constructs vibration-temperature coupling feature vectors based on temperature gradient changes. This vector is then input into a pre-trained fault diagnosis model to generate fault diagnosis results with fault type and confidence level. Dynamic correlation analysis extracts the frequency-domain energy entropy of the vibration spectrum (indicating mechanical impact intensity) and the spatiotemporal derivatives of the temperature gradient (reflecting thermal conductivity anomalies) to construct a physically interpretable vibration-temperature coupling feature vector. The fault diagnosis model utilizes a graph neural network architecture, embedding device structural relationships (derived from a spatial layout graph) and multi-source features into graph nodes. An attention mechanism is used to capture the weight of the impact of local faults on the global state. This approach overcomes the single-dimensional limitations of traditional threshold alarms and enables feature decoupling and pattern recognition of complex fault modes (such as thermal-vibration coupling failure).

[0025] S4. Based on the spatial mapping relationship between the fault diagnosis result and the environmental calibration model, generate a visual warning map, and feedback the warning map in real time through the interaction interface; the warning map includes the thermal identification of the fault location, the abnormal spectrum interval, and the confidence level.

[0026] The visual warning map uses spatial mapping technology to reverse-project the abstract feature vector onto the device entity model. By overlaying multiple layers of the thermal identification (temperature abnormal area), the spectrum energy cloud map (vibration abnormal frequency band), and the confidence heat zone (fault probability distribution), an interactive augmented reality interface is constructed. This design significantly reduces the cognitive complexity of multi-source data, enabling the operation and maintenance personnel to quickly locate the fault source and evaluate the severity, and improving the on-site decision-making efficiency.

[0027] S5. Upload the multi-source fusion data, the warning map, and the fault prompt to the cloud database, and iteratively optimize the reference value of the environmental calibration model and the correlation threshold of the fault diagnosis model based on historical data to form a closed-loop self-learning monitoring link.

[0028] The closed-loop self-learning monitoring link uses the multi-source fusion data and diagnosis results stored in the cloud, and adopts an incremental learning algorithm to dynamically update the reference parameters of the environmental calibration model (such as temperature drift compensation) and the correlation threshold of the fault diagnosis model (such as feature weight attenuation). At the same time, a transfer learning mechanism is introduced, and the fault feature library of similar devices is used as prior knowledge to accelerate the model convergence speed in new scenarios. This mechanism effectively addresses the impact of time-varying factors such as equipment aging and environmental disturbances on the monitoring system, and realizes the continuous evolution of the diagnostic ability.

[0029] The working principle of the present invention is as follows: Based on the multi-source sensing unit 20 and the magnetic adsorption positioning module 30 of the handheld integrated inspection instrument, first establish an environmental calibration model including temperature and vibration reference values by adsorbing the surface of the target device and loading the spatial layout map; then adaptively switch between the far-distance temperature field scanning and the near-distance hot spot tracking modes, collect vibration signals in different frequency bands and output multi-source fusion data with time stamps aligned; perform dynamic correlation analysis on the fusion data through the data processing module, extract the vibration spectrum energy distribution characteristics and construct a vibration-temperature coupling feature vector in combination with the temperature gradient, and input it into the pre-trained model to generate the fault type and confidence result; map the diagnosis result to the spatial model to generate a visual warning map, and feedback the comprehensive information including the fault location, the abnormal spectrum interval, and the confidence level in real time through the interaction interface; this method utilizes the coupling characteristics of the vibration spectrum energy distribution and the temperature gradient change, overcomes the limitations of single-parameter monitoring, significantly improves the detection accuracy, and optimizes the data acquisition efficiency and the instantaneity efficiency by adaptively switching the temperature measurement mode and the frequency-band vibration acquisition strategy, providing real-time, fast, and accurate fault detection results.

[0030] In this embodiment, specifically, step S1 specifically includes the following steps: S11. The flexible contact surface 31 of the magnetic adsorption positioning module 30 is attached to the surface of the target device, triggering the dynamic pressure detection unit to feedback the matching degree of the adsorption force and the contact surface curvature in real time. When the adsorption force reaches the preset threshold, the positioning signal is activated to generate the magnetic adsorption state parameters. The flexible contact surface 31 of the magnetic adsorption positioning module 30 adaptively adheres to the device surface, and the built-in dynamic pressure detection unit (such as a piezoelectric film sensor array) is used to monitor the adsorption force distribution and the contact surface curvature matching degree in real time. When the average adsorption force exceeds the preset threshold (such as ≥0.5 MPa) and the local pressure dispersion coefficient is lower than the tolerance range, the positioning signal activation instruction is triggered. This design solves the problem of positioning deviation caused by insufficient contact of traditional rigid magnetic adsorption devices on curved surface devices (such as motor end covers, pipe elbows), ensuring the physical coupling stability between the sensor array and the detection surface.

[0031] S12. Based on the positioning signal, the spatial layout diagram of the target device is loaded, and the gyroscope in the multi-source sensing unit 20 is used to identify the real-time pose of the inspection instrument in the spatial layout diagram, establishing the spatial mapping relationship between the surface coordinate system of the inspection instrument and the spatial layout diagram. Based on the magnetic adsorption positioning signal, the spatial layout diagram of the target device (such as a CAD model or point cloud data) is loaded. Through the MEMS gyroscope and accelerometer in the multi-source sensing unit 20, the attitude angles (pitch angle, yaw angle) and displacement vectors (X / Y / Z axis offsets) of the inspection instrument are calculated in real time. Combining the geometric features of the spatial layout diagram (such as bolt hole positions, heat dissipation ribs), the ICP (Iterative Closest Point) algorithm is used to establish the rigid transformation matrix between the surface coordinate system of the inspection instrument (with the magnetic adsorption center as the origin) and the global coordinate system of the device. This process realizes the accurate spatial registration of sensor data and the physical structure of the device, eliminating the confusion of the data spatial reference system caused by the difference in the handheld posture.

[0032] S13. Under the spatial mapping relationship, the far / near temperature measurement module 21 is controlled to obtain the initial temperature distribution matrix of the target device surface in the reference scanning mode, and the vibration sensing module 22 is synchronously triggered to collect the ambient background vibration spectrum in the low-frequency steady state mode, generating the initial reference parameter set of temperature and vibration.

[0033] Under the constraint of the spatial mapping relationship, the long-distance temperature measurement module (such as an infrared thermal imager) obtains the surface temperature distribution matrix in a wide-angle scanning mode. At the same time, the short-distance temperature measurement module (such as a laser temperature measurement probe) performs high-precision point temperature sampling on key areas (such as bearing housings and gearboxes) at a preset step size (such as a 5-mm spacing) to construct an initial temperature reference field. The low-frequency steady-state mode (such as a 0.1-1 kHz bandwidth) of the vibration sensing module 22 (such as a triaxial accelerometer) is synchronously started, and non-steady-state interferences (such as personnel movement and environmental noise) are filtered by the time-domain averaging method to extract the baseline of the inherent vibration spectrum of the equipment. This step forms an initial temperature-vibration characteristic template corresponding to the health state of the equipment.

[0034] S14. Based on the initial reference parameter set, the temperature field uniformity and vibration background noise characteristics are analyzed through the calibration algorithm of the multi-source sensing unit 20, and the emissivity compensation coefficient of the corresponding temperature measurement module and the filtering frequency band threshold of the vibration sensing module 22 are dynamically adjusted to construct multi-dimensional calibration parameters. Calculate the surface temperature field uniformity index (such as the standard deviation σ_T ≤ 2 °C) based on the initial temperature distribution matrix, and dynamically compensate the emissivity error of the long-distance temperature measurement module (such as the 0.1-0.2 emissivity deviation caused by the metal oxide layer). At the same time, analyze the background noise energy distribution of the vibration spectrum (such as the frequency band energy entropy H_v), and adaptively adjust the cut-off frequency of the band-pass filter of the vibration sensing module 22 (such as reducing the high-frequency band threshold from 10 kHz to 8 kHz to suppress electromagnetic interference). The online optimization of sensor parameters is realized through the calibration algorithm to improve the signal-to-noise ratio of abnormal signals in subsequent dynamic monitoring.

[0035] S15. Integrate the multi-dimensional calibration parameters with the spatial mapping relationship, and perform weighted interpolation based on the typical working condition data in the historical fault database to generate an environmental calibration model including the temperature gradient tolerance interval, the vibration spectrum baseline, and the identification of the fault-sensitive area.

[0036] Fuse the optimized multi-dimensional calibration parameters (temperature gradient, vibration band energy) with the equipment spatial layout diagram, and combine the typical working condition data (such as the temperature-vibration correlation curve of bearing oil shortage and winding overheating) of similar equipment in the historical fault database. The Gaussian process regression algorithm is used to perform weighted interpolation on the current calibration parameters to generate an environmental calibration model including a dynamic tolerance interval (such as the temperature gradient ΔT ≤ ±3 °C / min), a vibration spectrum energy baseline (such as the energy in the 1-kHz frequency band ≤ 0.5 g² / Hz), and a fault-sensitive area (such as the risk area of poor coupling alignment). This model enhances the sensitivity to compound faults through prior knowledge injection and reduces the false alarm probability during the cold start phase of new equipment.

[0037] In this embodiment, specifically, step S2 specifically includes the following steps: S21. Based on the temperature gradient tolerance interval and vibration spectrum baseline in the environmental calibration model, calculate the initial resolution of the far-field temperature field scan and the trigger threshold for the near-field hot spot tracking, and generate multi-mode switching decision parameters. Based on the temperature gradient tolerance interval (such as ΔT ≤ ±3°C / min) and vibration spectrum baseline (such as the energy threshold in the 1 kHz frequency band) in the environmental calibration model, calculate the optimal resolution of the far-field temperature field scan and the trigger threshold for the near-field hot spot tracking (such as the local temperature rise rate ≥ 5°C / s) through the dynamic programming algorithm. This step generates a multi-mode switching decision matrix (including the scan period, focusing area priority weight, etc.) by quantifying the model parameters, realizing the adaptive balance between the global monitoring efficiency and the local anomaly capture sensitivity, and avoiding resource waste in the traditional fixed parameter mode.

[0038] S22. According to the multi-mode switching decision parameters, synchronously start the temperature field scan thread of the far / near distance temperature measurement module 21 and the frequency band acquisition thread of the vibration sensing module 22 through the dynamic priority scheduling algorithm, and allocate an independent clock source to ensure the timing synchronization of the dual modules. Adopt the dynamic priority scheduling algorithm (such as preemptive thread management) to synchronously start the wide-area scan thread (low sampling rate, high coverage) of the far-field temperature measurement module and the dot matrix focusing thread (high sampling rate, local high precision) of the near-field temperature measurement module, and allocate an independent clock source (such as the FPGA hardware timer) for the frequency band acquisition thread of the vibration sensing module 22. Eliminate the timing jitter between multi-threads through the clock signal synchronization protocol (such as the PTP precise time protocol) to ensure that the acquisition interval error of temperature and vibration data ≤ 1 ms, and solve the time axis drift problem caused by asynchronous sampling of multi-sensors.

[0039] S23. In the far-field temperature field scan mode, control the temperature measurement module to cover the surface of the target device in a spiral path, and compare the scanned temperature with the temperature gradient tolerance interval in the environmental calibration model in real time. When it is detected that the local temperature deviation exceeds the trigger threshold, switch to the near-field hot spot tracking mode for focused temperature measurement.

[0040] In the far-field temperature field scan mode, control the infrared thermal imager to cover the target surface in a spiral path (shrinking from the edge of the device to the center circle by circle), and construct a dynamic monitoring window using the temperature gradient tolerance interval. When it is detected that the temperature deviation in a local area exceeds the trigger threshold for N consecutive frames, immediately switch to the near-field mode and drive the laser temperature measurement probe to perform dense dot matrix sampling on the abnormal area. This strategy reduces the ineffective scan area through dynamic path planning, and at the same time reduces the false switching probability caused by instantaneous interference through the multi-level trigger mechanism (temperature rise rate + duration).

[0041] S24, synchronized with the temperature measurement mode switching, controls the vibration sensing module 22 to divide the vibration signal acquisition into a low-frequency steady-state segment and a high-frequency transient segment based on a preset frequency band division strategy, and alternately activates the pressure detection units corresponding to the frequency bands in a time-division multiplexing manner; The vibration signal acquisition is divided into a low-frequency steady-state segment (0.1 - 1 kHz, representing the inherent vibration of the device) and a high-frequency transient segment (1 - 10 kHz, capturing impact events), and the piezoelectric sensing units corresponding to the frequency bands are alternately activated using a time-division multiplexing mechanism. The low-frequency band uses a continuous sampling mode, and the high-frequency band uses an event-triggered sampling mode (such as starting 10 kSPS high-speed acquisition after threshold triggering). The filter bank (such as low-pass / band-pass) and amplification gain are dynamically switched through a hardware switch matrix to achieve efficient capture of signals in different frequency bands on a single sensor hardware, reducing the power consumption and data redundancy of multi-channel parallel acquisition.

[0042] S25, the timestamp alignment unit synchronously marks the temperature gradient of the temperature measurement module and the vibration time-domain signal of the vibration sensing module 22, and based on the spatial mapping relationship of the surface coordinate system of the patrol instrument, fuses and encodes the temperature-vibration data according to the spatial grid coordinates, and outputs a multi-source fusion data stream with timestamp alignment and spatial coordinate binding.

[0043] The timestamp alignment unit interpolates and synchronizes the temperature gradient data and the vibration time-domain signal based on the synchronization module and the hardware clock counter to generate a fusion data packet with a unified time reference. Combining the spatial mapping relationship of the surface coordinate system of the patrol instrument (the rigid transformation matrix from S12), the temperature-vibration data is coordinate-bound and encoded according to the spatial grid on the device surface to form a multi-source data stream including timestamps, spatial coordinates, temperature values, and vibration energy vectors.

[0044] In this embodiment, specifically, step S3 specifically includes: S31, the built-in data processing module of the patrol instrument performs adaptive noise reduction processing on the vibration time-domain signal in the multi-source fusion data stream, separates the effective vibration components and background noise based on the vibration spectrum baseline in the environmental calibration model, and generates a denoised vibration spectrum sequence; Through the adaptive noise reduction algorithm in the data processing module (such as improved LMS filtering combined with blind source separation technology), a dynamic noise template is constructed using the vibration spectrum baseline in the environmental calibration model (such as the energy distribution of the background noise frequency band). Based on this template, the frequency-domain masking method is used to suppress the noise of the vibration time-domain signal while retaining the effective vibration components (such as the bearing impact characteristic frequency band). By iteratively optimizing the noise reduction parameters (such as the filter cut-off frequency, gain coefficient), the signal-to-noise ratio is improved and a denoised vibration spectrum sequence is generated. This step effectively eliminates the contamination of the vibration signal by environmental interference (such as electromagnetic noise, airflow disturbance), providing a high-fidelity input for subsequent feature extraction.

[0045] S32. Using the multi-scale wavelet packet decomposition algorithm, calculate the band energy entropy of the vibration frequency spectrum sequence, extract the characteristics of the high-frequency impact energy ratio, the low-frequency resonance main frequency offset, and the energy distribution dispersion degree, and construct a vibration multi-dimensional feature vector. Using the multi-scale wavelet packet decomposition algorithm (such as the db4 wavelet basis function), finely divide the vibration frequency spectrum sequence (such as 16-layer decomposition), and calculate the energy entropy values of each sub-band (characterizing the complexity of energy distribution). Extract the following characteristics: High-frequency impact energy ratio (when the energy ratio in the 5 - 10 kHz frequency band is ≥ 30% of the total energy, it indicates micro-cracks or spalling). Low-frequency resonance main frequency offset (such as when the fundamental frequency offset is ±5%, it corresponds to rotor imbalance or misalignment of the shafting). Energy distribution dispersion degree (quantify the energy distribution dispersion degree between frequency bands through the coefficient of variation, reflecting the randomness of abnormal impacts).

[0046] The constructed vibration multi-dimensional feature vector integrates time-frequency domain characteristics and enhances the characterization ability for compound faults.

[0047] S33. Synchronously analyze the temperature gradient change rate in the multi-source fusion data stream, combine it with the temperature gradient tolerance interval of the environmental calibration model, calculate the local temperature rise rate and the thermal diffusion uniformity index, and generate a temperature dynamic feature vector. Analyze the temperature gradient change rate in the multi-source fusion data (calculated by the temperature field difference based on spatial grid coordinates), combine it with the tolerance interval of the environmental calibration model (such as ΔT_max = 3℃ / min), quantify the local temperature rise rate (such as when dT / dt ≥ 5℃ / s, it triggers an overheat warning) and the thermal diffusion uniformity index (such as when the standard deviation of the temperature field σ_T ≥ 2.5℃, it determines abnormal heat conduction). Invert the local heat source intensity through the thermodynamic transfer equation, and generate a temperature dynamic feature vector including the temperature rise rate, thermal diffusion uniformity, and heat source intensity. This vector can effectively distinguish the normal heat dissipation of the equipment from the abnormal heat generation caused by faults.

[0048] S34. Align the vibration multi-dimensional feature vector and the temperature dynamic feature vector according to the time stamp, perform coupled analysis through the dynamic weight allocation layer in the pre-trained fault diagnosis model, and output the fault type, location identifier, and confidence score to form the fault diagnosis result.

[0049] After aligning the vibration and temperature feature vectors according to the unified timestamp, they are input into a pre-trained graph neural network (GNN) fault diagnosis model. The dynamic weight allocation layer in the model automatically assigns the contribution weights of the vibration and temperature features through an attention mechanism (such as multi-head self-attention) (for example, the weight coefficient of the high-frequency vibration feature for bearing faults is 0.7, and the weight coefficient of the temperature rise rate for winding overheating is 0.8). Combining the relationship of the equipment spatial layout diagram (such as the adjacent node connection between the bearing and the winding), the fault type (such as "bearing lubrication failure"), the location identifier (spatial grid coordinates G12-15), and the confidence score are output. This design realizes the physical association modeling of cross-domain features and significantly improves the identification accuracy of compound faults (such as thermal-vibration coupling failure).

[0050] In this embodiment, specifically, step S4 specifically includes: S41, based on the spatial mapping relationship of the environment calibration model and the location identifier in the fault diagnosis result, align the surface coordinate system of the patrol instrument with the location identifier in space to generate three-dimensional point cloud data of the fault thermal distribution; Based on the spatial mapping relationship in the environment calibration model (such as the rigid transformation matrix established in S12), convert the location identifier (such as spatial grid coordinates G12-15) in the fault diagnosis result to the global coordinate system of the equipment. Through the registration algorithm of the surface coordinate system of the patrol instrument and the three-dimensional model of the equipment (such as ICP iterative optimization), combined with the attitude parameters (pitch angle, yaw angle) of the magnetic adsorption positioning module 30, three-dimensional point cloud data of the fault thermal distribution is generated. This step realizes the accurate mapping of the fault location from the abstract coordinate to the physical entity, eliminates the influence of the attitude change of the handheld device on the positioning accuracy, and ensures the spatial consistency between the visualization result and the actual structure of the equipment.

[0051] S42, perform multi-layer superposition rendering on the three-dimensional point cloud data of the fault thermal distribution and the abnormal interval of the vibration spectrum, and perform transparency grading coloring on the abnormal area combined with the confidence level to generate a visualization warning map containing temperature-vibration coupling abnormal features; Adopt a multi-layer rendering engine (such as the WebGL architecture) to spatially superimpose the temperature abnormal point cloud data (expressed as a pseudo-color thermal map) and the abnormal interval of the vibration spectrum (expressed as a frequency domain energy cloud map). Control the transparency grading (α channel value 0.5-1.0 gradient) through the confidence level (such as the 80%-90% interval), and the high-confidence area is marked with high saturation, and the low-confidence area is displayed semi-transparently. At the same time, mark the frequency domain comparison curve of the abnormal frequency band (such as the normal baseline spectrum and the current spectrum are superimposed and displayed) to construct a visualization map of temperature-vibration coupling abnormality. This design presents through multi-modal data fusion, intuitively reveals the spatial association characteristics of compound faults, and reduces the difficulty of multi-source information integration for maintenance personnel.

[0052] S43. The visualization warning spectrum is adaptively adjusted in resolution through the dynamic compression algorithm built into the interaction interface, the spectrum slices are loaded, and the frequency-domain comparison curve of the spectrum anomaly interval is synchronously displayed on the terminal screen. Through the dynamic compression algorithm (such as the block compression technology based on wavelet transform) built into the interaction interface, according to the terminal screen resolution and the user's zoom operation level, the rendering resolution of the warning spectrum is adaptively adjusted (for example, downsampled to 1 / 4 resolution when displayed in full screen, and the original accuracy is restored when locally magnified). When loading the spectrum slices, the LOD (Level of Detail) technology is used to dynamically schedule data blocks with different precisions to ensure a smooth interaction experience. The synchronously displayed frequency-domain comparison curve supports touch dragging and frequency band focusing, enhancing the user's fine-grained analysis ability of abnormal features.

[0053] S44. Receive the annotation instructions or confidence correction parameters input by the user through the interaction interface, update the fault level identification in the warning spectrum in real time, and transmit the corrected data back to the fault diagnosis model for online weight fine-tuning.

[0054] Receive the annotation instructions (such as box-selecting the false alarm area) or confidence correction parameters (such as manually adjusting the fault level weight) input by the user through the interaction interface, trigger the real-time update of the warning spectrum (such as reducing the transparency of the false alarm area or modifying the color of the confidence identification). The corrected data is transmitted back to the fault diagnosis model through the incremental learning interface, and the online gradient descent algorithm is used to fine-tune the parameters of the dynamic weight allocation layer (such as the vibration feature weight coefficient is adjusted from 0.7 to 0.65), and the fault sensitive area identification of the environmental calibration model is updated. This mechanism realizes the closed-loop of human-machine collaborative diagnosis and continuously optimizes the model generalization ability through the injection of expert experience.

[0055] Embodiment 2: Combined with Figures 2 to 4 As shown, the present invention also provides a handheld integrated inspection instrument, which realizes detection by using the inspection method of the handheld integrated inspection instrument in Embodiment 1. The handheld integrated inspection instrument specifically includes a rod body 10, the rod body 10 is a segmented telescopic structure, and the front end of the rod body 10 is integrated with a multi-source sensing unit 20 and a magnetic adsorption positioning module 30. The multi-source sensing unit 20 includes a far / near temperature measurement module 21 and a vibration sensing module 22; The magnetic adsorption positioning module 30 is arranged at the bottom of the front end of the rod body 10, and includes a flexible contact surface 31, a permanent magnet and a pressure detection unit, and is used for adsorbing the surface of the target device and feeding back the contact state; The middle section of the rod body 10 is built-in with a data processing module and an inertial navigation unit, and the inertial navigation unit includes a gyroscope and an accelerometer; The rear end of the rod body 10 is provided with an interaction 40 and a detachable power module 50, and the interaction 40 is provided with a touch screen 41 and a button part 42.

[0056] The operation process is as follows: After the inspection instrument is started, the operator adjusts the front multi-source sensing unit 20 to the detection area of the target device through the retractable rod body 10. The flexible contact surface 31 of the magnetic adsorption positioning module 30 fits the surface of the device, and the pressure detection unit monitors the adsorption force state in real time. When the adsorption is stable, a positioning signal is triggered. The inertial navigation unit calculates the pose of the rod body 10 through the gyroscope and accelerometer, and establishes a spatial mapping relationship between the multi-source sensing unit 20 and the device surface in combination with the magnetic adsorption positioning parameters. The far / near temperature measurement module 21 adaptively switches between the wide-area scanning and focusing temperature measurement modes according to the environmental calibration model. The vibration sensing module 22 collects vibration signals in different frequency bands. The data processing module performs spatio-temporal alignment, noise reduction, and feature extraction on the temperature-vibration data, and generates a fault diagnosis result through a pre-trained model. The touch screen 41 of the interaction 40 displays the visual warning spectrum and the spectrum comparison curve in real time, supports the user to manually correct the diagnosis parameters, and the corrected data is transmitted back to the cloud to optimize the model. At the same time, the detachable power module ensures long-term continuous operation. Through the coordination of multi-source sensing integration, magnetic adsorption positioning, and dynamic data processing technologies, this inspection instrument realizes the accurate positioning and visual diagnosis of complex faults, and greatly improves the reliability and response efficiency of on-site inspection of industrial equipment.

[0057] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A patrol inspection method for a handheld integrated patrol inspection instrument, characterized in that, The inspection instrument includes a multi-source sensing unit and a magnetic positioning module, wherein the multi-source sensing unit includes a vibration sensing module and a long / short distance temperature measurement module; the inspection method includes the following steps: S1, establishing multi-dimensional calibration parameters of the target device through the multi-source sensing unit of the inspection instrument, based on the magnetic positioning module adsorbed on the surface of the target device and loading the spatial layout map, generating an environmental calibration model including temperature and vibration reference values; S2, synchronously running the long / short distance temperature measurement module and the vibration sensing module, adaptively switching between long-range temperature field scanning and short-range hotspot tracking modes according to the reference value in the environmental calibration model, and collecting vibration signals in a time-sharing manner based on a preset frequency band, and outputting multi-source fusion data with aligned timestamps; S3, performing dynamic correlation analysis on the multi-source fusion data, analyzing the vibration spectrum energy distribution using a feature extraction algorithm, constructing a vibration-temperature coupling feature vector based on the temperature gradient change, and inputting the vector into a pre-trained fault diagnosis model to generate a fault diagnosis result with a fault type and confidence level; S4, based on the spatial mapping relationship between the fault diagnosis result and the environmental calibration model, generate a visual warning map, and provide real-time feedback of the warning map through an interactive interface; the warning map includes a thermal identification of the fault location, a spectrum abnormality interval, and a confidence level.

2. The inspection method of the handheld integrated inspection instrument according to claim 1, characterized in that, The method of establishing multi-dimensional calibration parameters of the target device by the multi-source sensing unit of the inspection instrument specifically includes the following steps: S11, the flexible contact surface of the magnetic positioning module is attached to the surface of the target device, triggering the dynamic pressure detection unit to provide real-time feedback on the matching degree between the adsorption force and the curvature of the contact surface. When the adsorption force reaches a preset threshold, a positioning signal is activated and a magnetic state parameter is generated. S12, loading a spatial layout diagram of the target device based on the positioning signal, identifying the real-time position of the inspection instrument in the spatial layout diagram through the gyroscope in the multi-source sensing unit, and establishing a spatial mapping relationship between the surface coordinate system of the inspection instrument and the spatial layout diagram; S13, under the said spatial mapping relationship, controls the long / short distance temperature measurement module to obtain the initial temperature distribution matrix of the target device surface in the reference scanning mode, and synchronously triggers the vibration sensing module to collect the environmental background vibration spectrum in the low-frequency steady-state mode to generate an initial reference parameter set of temperature and vibration.

3. The inspection method of the handheld integrated inspection instrument according to claim 2, wherein The synchronously triggered vibration sensing module collects the ambient background vibration spectrum in a low-frequency steady-state mode to generate an initial reference parameter set of temperature and vibration, and then further includes: S14, based on the initial reference parameter set, analyzing the temperature field uniformity and vibration background noise characteristics through the calibration algorithm of the multi-source sensing unit, dynamically adjusting the emissivity compensation coefficient of the corresponding temperature measurement module and the filter frequency band threshold of the vibration sensing module, and constructing multi-dimensional calibration parameters; S15, integrating the multi-dimensional calibration parameters with the spatial mapping relationship, and performing weighted interpolation based on typical operating condition data in the historical fault database to generate an environmental calibration model including a temperature gradient tolerance range, a vibration spectrum baseline, and a fault-sensitive area identifier.

4. The inspection method of the handheld integrated inspection instrument according to claim 1, characterized in that, Run the far / near - distance temperature measurement module and the vibration sensing module synchronously, adaptively switch between the far - distance temperature field scanning and near - distance hot - spot tracking modes according to the reference values in the environmental calibration model, and collect vibration signals based on a preset frequency band in a time - sharing manner to output multi - source fusion data with aligned timestamps. The specific steps are as follows: S21. Based on the temperature gradient tolerance interval and vibration spectrum baseline in the environmental calibration model, calculate the initial resolution of far - distance temperature field scanning and the trigger threshold of near - distance hot - spot tracking, and generate multi - mode switching decision parameters; S22. According to the multi - mode switching decision parameters, synchronously start the temperature field scanning thread of the far / near - distance temperature measurement module and the frequency - band acquisition thread of the vibration sensing module through a dynamic priority scheduling algorithm, and allocate independent clock sources to ensure the timing synchronization of the two modules; S23. In the far - distance temperature field scanning mode, control the temperature measurement module to cover the surface of the target device in a spiral path, and compare the scanned temperature with the temperature gradient tolerance interval in the environmental calibration model in real time. When it is detected that the local temperature deviation exceeds the trigger threshold, switch to the near - distance hot - spot tracking mode for focused temperature measurement.

5. The inspection method of the handheld integrated inspection instrument according to claim 4, characterized in that, The step of comparing the scanned temperature with the temperature gradient tolerance interval in the environmental calibration model in real time. When it is detected that the local temperature deviation exceeds the trigger threshold, switch to the near - distance hot - spot tracking mode for focused temperature measurement, and then further includes: S24. Synchronous with the temperature measurement mode switching, control the vibration sensing module to divide the vibration signal acquisition into a low - frequency steady - state segment and a high - frequency transient segment based on a preset frequency - band division strategy, and alternately activate the pressure detection units corresponding to the frequency bands in a time - division multiplexing manner; S25. Use a timestamp alignment unit to synchronously mark the temperature gradient of the temperature measurement module and the vibration time - domain signal of the vibration sensing module, and based on the spatial mapping relationship of the surface coordinate system of the inspection instrument, fuse and encode the temperature - vibration data according to the spatial grid coordinates to output a multi - source fusion data stream with aligned timestamps and bound spatial coordinates.

6. The inspection method of the handheld integrated inspection instrument according to claim 1, characterized in that Conduct dynamic correlation analysis on the multi - source fusion data, use a feature extraction algorithm to analyze the energy distribution of the vibration spectrum, combine the temperature gradient change to construct a vibration - temperature coupling feature vector, and input it into a pre - trained fault diagnosis model to generate a fault diagnosis result of the fault type and confidence level, specifically including: S31. Through the data - processing module built in the inspection instrument, perform adaptive noise reduction processing on the vibration time - domain signal in the multi - source fusion data stream, separate the effective vibration components and background noise based on the vibration spectrum baseline in the environmental calibration model, and generate a denoised vibration spectrum sequence; S32. Use a multi - scale wavelet packet decomposition algorithm to calculate the frequency - band energy entropy of the vibration spectrum sequence, extract features such as the proportion of high - frequency impact energy, the offset of the low - frequency resonance main frequency, and the dispersion degree of energy distribution, and construct a multi - dimensional vibration feature vector; S33. Synchronously analyze the temperature gradient change rate in the multi - source fusion data stream, and combine the temperature gradient tolerance interval of the environmental calibration model to calculate the local temperature rise rate and the thermal diffusion uniformity index, and generate a temperature dynamic feature vector; S34. Align the vibration multi-dimensional feature vector and the temperature dynamic feature vector according to the time stamp, perform coupled analysis through the dynamic weight allocation layer in the pre-trained fault diagnosis model, and output the fault type, location identifier, and confidence score to form a fault diagnosis result.

7. The inspection method of the handheld integrated inspection instrument according to claim 6, characterized in that, Based on the spatial mapping relationship between the fault diagnosis result and the environment calibration model, generate a visual warning map and real-time feedback the warning map through the interaction interface; the warning map includes a thermal identifier of the fault location, a spectrum abnormal interval, and a confidence level, specifically including: S41. Based on the spatial mapping relationship of the environment calibration model and the location identifier in the fault diagnosis result, spatially align the surface coordinate system of the inspection instrument with the location identifier to generate three-dimensional point cloud data of the fault thermal distribution; S42. Perform multi-layer overlay rendering on the three-dimensional point cloud data of the fault thermal distribution and the vibration spectrum abnormal interval, and perform transparency grading coloring on the abnormal area in combination with the confidence level to generate a visual warning map containing temperature-vibration coupling abnormal features; S43. Adaptively adjust the resolution of the visual warning map through the dynamic compression algorithm built in the interaction interface, load the map slices, and synchronously display the frequency domain comparison curve of the spectrum abnormal interval on the terminal screen; S44. Receive the annotation instruction or confidence correction parameter input by the user through the interaction interface, update the fault level identifier in the warning map in real time, and transmit the corrected data back to the fault diagnosis model for online weight fine-tuning.

8. The inspection method of the handheld integrated inspection instrument according to claim 1, characterized in that, Based on the spatial mapping relationship between the fault diagnosis result and the environment calibration model, generate a visual warning map and real-time feedback the warning map through the interaction interface. After that, it further includes: S5. Upload the multi-source fusion data, warning map, and fault prompt to the cloud database, and iteratively optimize the reference value of the environment calibration model and the correlation threshold of the fault diagnosis model based on historical data to form a closed-loop self-learning monitoring link.

9. A handheld integrated inspection instrument, characterized in that, The inspection method using the handheld integrated inspection instrument described in any one of claims 1 to 8 is used for detection. The handheld integrated inspection instrument specifically includes a rod body, the rod body is a segmented telescopic structure, and the front end of the rod body is integrated with a multi-source sensing unit and a magnetic attraction positioning module. The multi-source sensing unit includes a far / near distance temperature measurement module and a vibration sensing module; The magnetic attraction positioning module is arranged at the bottom of the front end of the rod body and includes a flexible contact surface, a permanent magnet, and a pressure detection unit for adsorbing the surface of the target device and feedbacking the contact state; The middle section of the rod body is built with a data processing module and an inertial navigation unit. The inertial navigation unit includes a gyroscope and an accelerometer; The rear end of the rod body is provided with an interaction unit and a detachable power module. The interaction unit is provided with a touch screen and a button part.

Images