Gear pair detection device and detection method

Abstract

The invention relates to the technical field of detection devices, in particular to a gear pair detection device and method, and the device comprises a base, a first detection assembly, a second detection assembly, a main shaft assembly, a collection assembly, and a controller. The first detection assembly, the second detection assembly and the main shaft assembly are arranged on the base; the acquisition assembly is arranged on the second detection assembly and the main shaft assembly, the controller is arranged on the base, and the acquisition assembly is connected with the controller; the main shaft assembly is used for placing a main gear and driving the main gear to rotate; the first detection assembly is used for positioning the main gear; the second detection assembly is used for driving the auxiliary gear to rotate, axially pressurizing the auxiliary gear and driving the auxiliary gear to be close to and engaged with the main gear; concrete is poured into the base, and a balancing weight mounting groove is reserved in the outer side of the base; the method has the beneficial effects that the accuracy of a detection result is remarkably improved, gear pair detection requirements in different scenes are met, and the reliability and consistency of a judgment result are ensured.

Description

Technical Field

[0001] This invention relates to the field of testing equipment technology, and in particular to a gear pair testing device and testing method. Background Technology

[0002] Currently, gear pairs are the core transmission components of new energy vehicle reducers, and their meshing quality directly determines the vibration and noise level of the reducer, thus affecting the overall vehicle ride comfort. In new energy vehicles, due to the lack of masking effect from traditional engine noise, abnormal noise is generated during gear pair meshing. This type of noise mainly originates from tooth surface errors in the gear machining process (such as tool accuracy deviations, machine tool vibration interference, etc.). These errors are excited by meshing forces during meshing, generating additional vibration and noise, which not only detracts from the driving experience but may also indirectly reflect potential problems in the transmission stability of the gear pair, becoming a key pain point in the quality control of new energy vehicle reducers. Therefore, how to quickly identify and intercept products with abnormal noise during the gear pair manufacturing stage to prevent them from entering the final assembly stage is a core problem that urgently needs to be solved in the industry.

[0003] Currently, the invention patent with application number CN202311522386.8, classified under G01M 13 / 028 and G01M 17 / 007, proposes a comprehensive evaluation method based on the angular fluctuations and meshing vibration acceleration of the active and driven gears. It extracts ghost frequency-related order features through Fourier transform and obtains a weighted comprehensive evaluation index, providing a new approach for online identification of ghost frequency noise. However, this technology still has room for optimization in practical applications: on the one hand, the weight adjustment logic of its comprehensive evaluation index relies on a large number of samples for verification, resulting in insufficient adaptability to gear pairs of different specifications; on the other hand, the matching degree between the simulation of gear pair meshing conditions and actual vehicle loading conditions during the detection process needs improvement, which may lead to some potential ghost frequency noise gear pairs not being effectively intercepted, resulting in insufficient detection accuracy. Summary of the Invention

[0004] (a) Technical problems to be solved In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a gear pair detection device and detection method, which solves the technical problems of insufficient adaptability and accuracy of the prior art for gear pairs of different specifications.

[0005] (II) Technical Solution To achieve the above objectives, the main technical solutions adopted by the present invention include: On one hand, the present invention provides a gear pair detection device, comprising a base, a first detection assembly, a second detection assembly, a spindle assembly, a data acquisition component, and a controller; the first detection assembly, the second detection assembly, and the spindle assembly are disposed on the base; the data acquisition component is disposed on the second detection assembly and the spindle assembly, and the controller is disposed on the base, with the data acquisition component connected to the controller; the spindle assembly is used to place the main gear and drive the main gear to rotate, and the first detection assembly is used to position the main gear; the second detection assembly has a secondary gear detection station, used to drive the secondary gear to rotate and apply axial pressure to the secondary gear, and to drive the secondary gear to approach and mesh with the main gear; the base is filled with concrete, and a counterweight mounting groove is reserved on the outer side of the base; the data acquisition component is used to acquire the torque time-domain signal, angle time-domain signal, and vibration time-domain signal of the main gear and the secondary gear; the controller has a pre-trained multimodal analysis model, which is used to analyze the data acquired by the data acquisition component to determine whether the meshing state of the main gear and the secondary gear is qualified.

[0006] Optionally, the acquisition component includes a torque sensor, a first circular grating, a second circular grating, a first vibration sensor, and a second vibration sensor; the torque sensor, the first circular grating, and the first vibration sensor are mounted on the spindle assembly, while the second circular grating and the second vibration sensor are mounted on the second detection assembly.

[0007] Optionally, the first detection assembly includes a first column, a first slide plate, a first drive mechanism, and a first bearing seat; the first column is fixed to the base, and the first slide plate is slidably disposed on the first column; the first bearing seat is mounted on the first slide plate and is located above the main shaft assembly; the first drive mechanism is disposed on the first column and is connected to the first slide plate to drive the first slide plate to slide up and down, thereby causing the first bearing seat to move closer to and away from the main gear.

[0008] Optionally, the second detection assembly includes a second column, a column drive mechanism, a second slide plate, a second drive mechanism, a power head, a third slide plate, a third drive mechanism, and a second shaft seat. The second column is movably mounted on the base, the column drive mechanism is mounted on the base, and the column drive mechanism is connected to the second column to drive the second column closer to and further away from the main shaft assembly. The second and third slide plates are slidably mounted on the second column, with the second slide plate located above the third slide plate. The power head is located on the second slide plate, and the second shaft seat is located on the third slide plate. The second and third drive mechanisms are mounted on the second column. The second drive mechanism is connected to the second slide plate to drive the second slide plate to slide up and down, causing the power head to move closer to and further away from the secondary gear. The third drive mechanism is connected to the third slide plate to drive the third slide plate to slide up and down, causing the second shaft seat to adjust the height of the secondary gear. The acquisition component is located on the power head.

[0009] On the other hand, the present invention also provides a detection method for the above-mentioned gear pair detection device, comprising the following steps: S1. The controller controls the first detection assembly to approach and abut against the positioning main gear, the controller controls the main shaft assembly to drive the main gear to rotate, and the controller controls the second detection assembly to drive the auxiliary gear to rotate and adjust its height to approach and mesh with the main gear. S2. During the meshing process, the controller receives the analysis data collected by the acquisition component. The analysis data includes: the main gear torque time domain signal T(t), the main gear angle time domain signal θ1(t), the secondary gear angle time domain signal θ2(t), the main gear vibration time domain signal a1(t), and the secondary gear vibration time domain signal a2(t). S3. The controller performs the first preprocessing of the analysis data to obtain preprocessed data. S4. Then the controller inputs the preprocessed data into the trained multimodal analysis model and outputs the comprehensive judgment value S. Specifically, when S≤0.3, the gear pair meshing state is considered qualified; when S>0.3, it is considered unqualified.

[0010] Optionally, step S1 includes: S11. The controller controls the main shaft assembly to drive the main gear to rotate at a low speed of 20~40r / min. The controller also controls the second detection assembly to drive the auxiliary gear to rotate, and the rotation speed of the auxiliary gear is consistent with that of the main gear. The controller further controls the second detection assembly to drive the auxiliary gear to approach and mesh with the main gear. S12. The controller controls the first detection assembly to apply an axial pressure of 20~30N to the auxiliary gear, and the controller controls the second detection assembly to apply an axial pressure of 70~100N to the main gear. S13. The controller controls the first and second detection assemblies to continue driving the main gear and auxiliary gear to rotate, and increases the speed of the main gear and auxiliary gear to 150~200 r / min.

[0011] Optionally, step S3 includes: S31. The controller uses Kalman filtering to denoise the main gear torque time-domain signal T(t) to obtain the denoised torque signal T'(t). S32, The controller determines the transmission ratio between the main gear and the auxiliary gear. ; Z1 is the number of teeth on the main gear, Z2 is the number of teeth on the secondary gear, and the rotation angle difference signal Δθ(t) is calculated using a pre-set formula (Formula 1); the formula (Formula 1) is: ; S33. The controller removes low-frequency interference from the main gear vibration time-domain signal a1(t) and the auxiliary gear vibration time-domain signal a2(t) by using a 50Hz high-pass filter, and obtains the filtered signals a1'(t) and a2'(t). Perform a fast Fourier transform on a1'(t) and a2'(t) to obtain the frequency domain signal A1(f) of the main gear and the frequency domain signal A2(f) of the auxiliary gear. S34. Perform feature normalization to obtain the eigenvalue set: Torque ripple characteristics ; Corner difference characteristics ; Peak factor of main gear in time domain ; Frequency domain amplitude characteristics of 1 times meshing frequency A1(f) m ) is A1(f) in f m The amplitude at (main gear meshing frequency), A ref =100g / Hz, which is the reference amplitude; Peak factor in time domain of auxiliary gear ; Frequency domain twice the meshing frequency amplitude characteristics A2 (2f) m ) is A2(f) in 2f m The amplitude at that point.

[0012] Among them, peak means peak value, which refers to the maximum absolute value of the vibration time-domain signal within the set analysis period; rms is the root mean square value, which is the quantification of the average energy level of the vibration time-domain signal within the analysis period; the time-domain peak factor is the ratio of instantaneous impact to average energy, and its core function is to amplify the instantaneous anomalies caused by the fault.

[0013] Optionally, step S4 includes: S41. The controller inputs the preprocessed data obtained in step S3 into the trained multimodal analysis model; S42. Calculate the comprehensive judgment value S using a linear weighting formula; the linear weighting formula is: ; w1, w2, w3, w4, w5, and w6 are pre-set weighting coefficients; S43. Result Determination and Output: When S≤0.3, the controller determines that the gear pair meshing state is qualified and outputs a qualified signal and various characteristic values; When S>0.3, it is judged as unqualified, and an unqualified signal, S value, and the two features that contribute the most in the feature value group are output.

[0014] Optionally, the multimodal analysis model is trained through the following steps: S04-1. Preset at least 1500 sets of sample data, of which 700 sets are qualified gear pairs and 800 sets are faulty gear pairs.

[0015] S04-2. Preprocess each set of sample data.

[0016] S04-3. Input the preprocessed sample data into the multimodal analysis model to obtain the trained multimodal analysis model.

[0017] Optionally, in step S04-1, the types of faulty gear pairs include main gear installation eccentricity, secondary gear tooth surface wear, tooth pitch error, and tooth surface scratches; the number of each type is not less than 150 groups.

[0018] In step S04-2, the preprocessed sample data is labeled with the true state label and the fault type label.

[0019] (III) Beneficial Effects The beneficial effects of this invention are: This invention provides a gear pair detection device. Its base is made of poured concrete with a pre-reserved counterweight mounting groove, which significantly improves its rigidity to suppress resonance and allows for flexible adjustment of inertia according to gear pair specifications, ensuring stable detection and adaptability to various specifications. The integrated main shaft assembly and second detection assembly achieve coordinated control of the main and auxiliary gear drive, meshing position adjustment, and axial pressure, simulating the dynamic meshing and load states of the gear pair in actual operation, avoiding the scenario disconnect caused by traditional step-by-step operations. Multi-dimensional acquisition components simultaneously collect torque, angle, and vibration signals, comprehensively capturing the dynamic characteristics during meshing, avoiding the one-sidedness of single-signal detection. Through a multi-modal analysis model trained in the controller, based on the comprehensive analysis of complex features using the three types of acquired signals, misjudgments caused by signal distortion or missing information are significantly reduced, significantly improving the accuracy of the detection results. This device adapts to the gear pair detection needs in different scenarios and ensures reliable consistency of judgment results.

[0020] This invention provides a detection method for a gear pair detection device. Through streamlined design, it transforms hardware functionality into practical detection capabilities. First, it optimizes the base inertia and meshing parameters based on the characteristics of the gear pair, ensuring the detection scenario closely matches actual working conditions and avoiding signal distortion caused by inertia mismatch or static meshing. Then, it purifies multi-source signals through targeted preprocessing and normalizes the signals, eliminating dimensional differences between gear pairs of different specifications. Furthermore, the model training covers various typical gear samples and supports dynamic weight optimization, adapting to gear pairs with different modules and numbers of teeth. This eliminates the need for separate process design for each specification, solving the problem of limited versatility in traditional methods. Simultaneously, it enhances signal authenticity by simulating actual working conditions, extracting multi-dimensional features such as torque fluctuation, rotation angle difference, and vibration in the time and frequency domains. Based on statistical laws, it sets reasonable judgment thresholds, avoiding judgment bias caused by single features or subjective experience. The collaborative analysis of multi-source information significantly improves the ability to identify hidden faults, making qualification judgments more accurate and effectively reducing missed and false judgments, thus balancing the wide applicability of the detection with the accuracy of the results. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the gear pair detection device according to Embodiment 1 of the present invention; Figure 2 This is an exploded view of the gear pair detection device according to Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the structure of the first detection assembly in Embodiment 1 of the present invention; Figure 4 This is an exploded view of the structure of the first detection assembly in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the structure of the second detection assembly in Embodiment 1 of the present invention; Figure 6 This is an exploded view of the structure of the second detection assembly in Embodiment 1 of the present invention; Figure 7 This is a partial structural schematic diagram of the power head of Embodiment 1 of the present invention; Figure 8 This is a schematic diagram of the spindle assembly structure in Embodiment 1 of the present invention; Figure 9 This is a schematic flowchart of the detection method according to Embodiment 2 of the present invention; Figure 10 This is a schematic diagram of the overall structure of the gear pair detection device according to Embodiment 4 of the present invention; Figure 11 This is a side view schematic diagram of the gear pair detection device of Embodiment 4 of the present invention; Figure 12 yes Figure 11 A schematic diagram of section AA.

[0022] [Explanation of Labels in the Attached Image] 1: Base; 11: Counterweight mounting slot; 2: First detection assembly; 21: First column; 22: First slide plate; 23: First drive mechanism; 24: First bearing seat; 3: Second inspection assembly; 31: Second column; 32: Column drive mechanism; 33: Second slide plate; 35: Second drive mechanism; 36: Power head; 361: Power driver; 362: Power mounting base; 363: Power coupling; 364: Power shaft seat; 37: Third slide plate; 38: Third drive mechanism; 39: Second shaft seat; 4: Spindle assembly; 41: Spindle mounting base; 42: Spindle driver; 43: Spindle housing; 5: Outer casing; 51: Automatic sliding door; 52: Door opening and closing actuator; 53: Door opening and closing control panel; 54: First guide rail; 55: Second guide rail. Detailed Implementation

[0023] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.

[0024] Example 1: like Figure 1 and Figure 2 As shown, a specific embodiment of the present invention provides a gear pair detection device, including a base 1, a first detection assembly 2, a second detection assembly 3, a spindle assembly 4, a data acquisition component, and a controller; the first detection assembly 2, the second detection assembly 3, and the spindle assembly 4 are disposed on the base 1; the data acquisition component is disposed on the second detection assembly 3 and the spindle assembly 4, and the controller is disposed on the base 1, with the data acquisition component connected to the controller; the spindle assembly 4 is used to place the main gear and drive the main gear to rotate, and the first detection assembly 2 is used to position the main gear; the second detection assembly 3 has a secondary gear detection station, used to drive the secondary gear to rotate and apply axial pressure to the secondary gear, and to drive the secondary gear to approach and mesh with the main gear; the base 1 is filled with concrete, and a counterweight mounting groove 11 is reserved on the outer side of the base 1; the data acquisition component is used to acquire the torque time-domain signal, angle time-domain signal, and vibration time-domain signal of the main gear and the secondary gear; the controller has a pre-trained multimodal analysis model, which is used to analyze the data acquired by the data acquisition component to determine whether the meshing state of the main gear and the secondary gear is qualified.

[0025] Specifically, it effectively solves the core pain points of existing testing devices: its base 1 is made of concrete with a pre-reserved counterweight installation groove, which not only significantly improves its rigidity to suppress resonance, but also allows for flexible adjustment of inertia according to the gear pair specifications, ensuring a stable testing process and adaptability to various specifications; the integrated main shaft assembly 4 and the second testing assembly 3 achieve coordinated control of the main and auxiliary gear drive, meshing position adjustment and axial pressure, which can simulate the dynamic meshing and load state of the gear pair in actual operation, avoiding the scenario disconnect caused by traditional step-by-step operation; the multi-dimensional acquisition component simultaneously acquires three types of signals: torque, angle and vibration signals, comprehensively capturing the dynamic characteristics in the meshing process, avoiding the one-sidedness of single signal detection, and through the multi-modal analysis model trained in the controller, comprehensively analyzes the complex features based on the three types of signals acquired, greatly reducing misjudgments caused by signal distortion or missing information, and significantly improving the accuracy of the test results, adapting to the gear pair testing needs in different scenarios and ensuring the reliability and consistency of the judgment results.

[0026] Furthermore, in this embodiment, the acquisition component includes a torque sensor, a first circular grating, a second circular grating, a first vibration sensor, and a second vibration sensor. The torque sensor, the first circular grating, and the first vibration sensor are mounted on the main shaft assembly 4, while the second circular grating and the second vibration sensor are mounted on the second detection assembly 3. By specifically configuring the torque sensor, dual circular gratings, and dual vibration sensors, and corresponding to the main shaft assembly 4 and the second detection assembly 3 where the main and auxiliary gears are located, the acquisition component achieves synchronous acquisition of full-dimensional signals of the torque of the main gear, the angle between the main and auxiliary gears, and the vibration of the main and auxiliary gears during gear meshing. The torque sensor focuses on the force state of the main gear, the dual circular gratings capture the rotation angle information of the main and auxiliary gears to reflect synchronicity, and the dual vibration sensors accurately distinguish the vibration characteristics of the main and auxiliary gears, avoiding signal confusion caused by a single sensor or mixed acquisition. This component-based, targeted acquisition design ensures a one-to-one correspondence between signals and the main / secondary gears. Through the coordinated capture of multiple physical quantities, it provides accurate data support for distinguishing between main gear faults (such as installation eccentricity) and secondary gear faults (such as tooth surface wear). This solves the problems of ambiguous signal sources and feature cross-interference in traditional detection, and significantly improves the accuracy of fault location.

[0027] Furthermore, such as Figure 3 and Figure 4As shown, the first detection assembly 2 includes a first column 21, a first slide plate 22, a first drive mechanism 23, and a first bearing 24. The first column 21 is fixed to the base 1, and the first slide plate 22 is slidably mounted on the first column 21. The first bearing 24 is mounted on the first slide plate 22 and is located above the main shaft assembly 4. The first drive mechanism 23 is mounted on the first column 21 and connected to the first slide plate 22 to drive the first slide plate 22 to slide up and down, causing the first bearing 24 to move closer to and away from the main gear. The acquisition component is mounted on the first bearing 24. This design facilitates quick mounting and dismounting of the main gear (closer positioning during detection, and further away for removal after detection), improving operational convenience. Furthermore, it ensures the coaxiality and perpendicularity of the main gear during installation, avoiding errors caused by manual alignment. Simultaneously, it can adapt to main gears of different thicknesses and shaft diameters—installation requirements can be met simply by adjusting the height of the first bearing 24 via the first drive mechanism 23, enhancing the device's adaptability to diverse main gears and improving overall versatility.

[0028] Furthermore, such as Figure 8 As shown, in this embodiment, the spindle assembly 4 includes a spindle mounting base 41, a spindle driver 42, a spindle coupling, and a spindle seat 43. The spindle mounting base 41 is fixedly mounted on the base 1, the spindle driver 42 is disposed on the spindle mounting base 41, the spindle seat 43 is mounted on the top of the spindle mounting base 41, and the drive shaft of the spindle driver 42 is connected to the spindle seat 43 through the spindle coupling. A torque sensor is disposed between the drive shaft of the spindle driver 42 and the spindle coupling, a first circular grating is disposed on the spindle mounting base 41, the drive shaft of the spindle driver 42 passes through the first circular grating, and a first vibration sensor is disposed on the spindle mounting base 41. The torque sensor is located between the drive shaft and the coupling, directly measuring the meshing torque; the first circular grating is close to the main shaft mounting base 41 to capture the rotation angle of the main gear; the first vibration sensor is fixed to the main shaft mounting base 41, collecting the vibration of the main gear at close range, reducing the attenuation and distortion of the signal during transmission, ensuring that the collected main gear torque, angle, and vibration signals can truly reflect its meshing state, providing high-quality raw data for subsequent feature extraction, and improving the accuracy of detection from the signal source.

[0029] Furthermore, such as Figure 5 and Figure 6As shown, the second detection assembly 3 includes a second column 31, a column drive mechanism 32, a second slide plate 33, a second drive mechanism 35, a power head 36, a third slide plate 37, a third drive mechanism 38, and a second shaft seat 39. The second column 31 is movably mounted on the base 1, and the column drive mechanism 32 is mounted on the base 1 and connected to the second column 31 to drive the second column 31 to move closer to and away from the main shaft assembly 4. The second slide plate 33 and the third slide plate 37 are slidably mounted on the second column 31. Located above the third slide plate 37, the power head 36 is mounted on the second slide plate 33, and the second bearing 39 is mounted on the third slide plate 37. The second drive mechanism 35 and the third drive mechanism 38 are mounted on the second column 31. The second drive mechanism 35 is connected to the second slide plate 33 to drive the second slide plate 33 to slide up and down, causing the power head 36 to move closer to and away from the auxiliary gear. The third drive mechanism 38 is connected to the third slide plate 37 to drive the third slide plate 37 to slide up and down, causing the second bearing 39 to adjust the height of the auxiliary gear. The acquisition component is located on the power head 36. Through the composite adjustment structure of the movable second column 31, multiple slide plates, and multiple drive mechanisms, multi-dimensional precise adjustment of the auxiliary gear is achieved in the horizontal direction (closer / away from the main shaft assembly 4) and the vertical direction (up and down sliding). Horizontal adjustment can adapt to main and auxiliary gear combinations with different center distances (such as differences in meshing distance between gears of different modules), while vertical adjustment ensures that the axis of the auxiliary gear and the main gear are at the same height, avoiding meshing tilt caused by height deviation.

[0030] Furthermore, such as Figure 1 and Figure 7 As shown, the power head 36 includes a power driver 361, a power mounting base 362, a power coupling 363, and a power shaft seat 364. The power driver 361 is mounted on the second slide plate 33 via the power mounting base 362, and the power shaft seat 364 is mounted on the second slide plate 33. The drive shaft of the power driver 361 is connected to the power shaft seat 364 via the power coupling 363. A second circular grating and a second vibration sensor are disposed on the power shaft seat 364. This drive structure drives the auxiliary gear to rotate, and the second circular grating and the second vibration sensor are integrated into the power shaft seat 364, allowing the sensor to be directly close to the meshing area of ​​the auxiliary gear. On the one hand, this ensures the stability of the auxiliary gear rotation (the coupling buffers the transmission impact) and reduces signal noise caused by drive fluctuations; on the other hand, the sensor collects the rotation angle and vibration signals of the auxiliary gear at close range, which can accurately capture the subtle characteristics of the auxiliary gear during meshing (such as abnormal vibration caused by tooth surface wear and rotation angle fluctuations caused by tooth pitch error), forming a main-auxiliary corresponding acquisition system with the sensor of the main shaft assembly 4. The two work together to extract the independent features of the main and auxiliary gears, avoiding misjudgments of faults caused by signal mixing and improving the ability to distinguish between faults of the main and auxiliary gears.

[0031] Furthermore, such as Figure 1and Figure 2 As shown, in this embodiment, the second column 31 is designed with enhanced strength compared to the first column 21, and the cross-sectional dimension of the second column 31 is larger than that of the first column 21. The second detection assembly 3 needs to bear the multi-directional adjustment, driving, and signal acquisition of the sub-gear, and the force is more complex (including horizontal thrust, vertical adjustment force, and the reaction force of the sub-gear rotation). The larger cross-sectional dimension can improve its rigidity and anti-deformation ability, and avoid the position deviation of the sub-gear caused by the deformation of the column during the adjustment or detection process (such as the bending of the column during horizontal movement and the vibration during up and down sliding). This enhanced design synergizes with the function of the first column 21 (the first column 21 mainly bears the positioning of the main gear and the force is relatively simple), jointly ensuring the position accuracy and stability when the main and sub-gears are meshed, providing a structural basis for the multi-sensor synchronous acquisition of high-quality signals, reducing the detection error caused by the deformation of the equipment, and improving the long-term use reliability of the device and the consistency of the detection results.

[0032] Embodiment 2: As Figure 9 shown, this embodiment provides a detection method for the gear pair detection device described in Embodiment 1, including the steps of: S1. Control the first detection assembly 2 to approach and abut against the main gear for positioning through the controller, control the main shaft assembly 4 to drive the main gear to rotate through the controller, and control the second detection assembly 3 to drive the sub-gear to rotate and adjust the height to approach and mesh with the main gear through the controller.

[0033] S2. During the meshing process, the controller receives the analysis data collected by the acquisition component, and the analysis data includes: the main gear torque time-domain signal T(t), the main gear angle time-domain signal θ1(t), the sub-gear angle time-domain signal θ2(t), the main gear vibration time-domain signal a1(t), and the sub-gear vibration time-domain signal a2(t).

[0034] S3. The controller performs the first preprocessing on the analysis data to obtain the preprocessed data.

[0035] S4. Then the controller inputs the preprocessed data into the trained multi-modal analysis model and outputs a comprehensive judgment value S. Among them, when S≤0.3, it is determined that the meshing state of the gear pair is qualified; when S>0.3, it is determined to be unqualified.

[0036] Specifically, the hardware functionality is transformed into actual detection capabilities through a streamlined design: First, the inertia and meshing parameters of the base 1 are optimized based on the characteristics of the gear pair to ensure that the detection scenario closely matches the actual working conditions, avoiding signal distortion caused by inertia mismatch or static meshing. Then, targeted preprocessing is used to purify multi-source signals and normalize the signals, eliminating dimensional differences between gear pairs of different specifications. The model training covers a variety of typical gear samples and supports dynamic weight optimization, making it adaptable to the detection of gear pairs with different modules and numbers of teeth. There is no need to design separate processes for each specification, solving the problem of limited versatility of traditional methods. At the same time, the realism of the signals is improved by simulating actual working conditions in stages, extracting multi-dimensional features such as torque fluctuation, rotation angle difference, vibration time domain and frequency domain, and setting reasonable judgment thresholds based on statistical laws, avoiding judgment bias caused by single features or subjective experience. The collaborative analysis of multi-source information greatly improves the ability to identify hidden faults, making the qualification judgment more accurate, effectively reducing missed judgments and false judgments, and taking into account both the wide applicability of the detection and the high accuracy of the results.

[0037] Further, step S1 includes: S11. The controller controls the main shaft assembly 4 to drive the main gear to rotate at a low speed of 20~40r / min. The controller controls the second detection assembly 3 to drive the auxiliary gear to rotate. The rotation speed of the auxiliary gear is consistent with that of the main gear. The controller controls the second detection assembly 3 to drive the auxiliary gear to approach and mesh with the main gear. S12. The controller controls the first detection assembly 2 to apply an axial pressure of 20~30N to the auxiliary gear, and the controller controls the second detection assembly 3 to apply an axial pressure of 70~100N to the main gear. S13. The controller controls the first detection assembly 2 and the second detection assembly 3 to continue driving the main gear and the auxiliary gear to rotate, and increases the speed of the main gear and the auxiliary gear to 150~200 r / min.

[0038] Specifically, by employing a phased drive control system of low-speed meshing alignment, axial compression, and stepped speed increase, the problems of meshing impact and state distortion caused by static docking or sudden speed increases in traditional gear meshing detection are effectively solved. Low-speed alignment utilizes the natural meshing of gears to guide precise contact, avoiding tooth surface collisions and initial meshing deviations; axial compression simulates actual working load conditions, ensuring that the detection state is consistent with the real working environment of the gear pair; and the stepped speed increase avoids sudden changes in meshing load by gradually increasing the speed, reducing vibration interference and signal distortion. The synergistic effect of these three aspects ensures both the stability and safety of the master and slave gear meshing (avoiding tooth surface damage) and enables the acquired signals to accurately reflect the meshing characteristics at different speeds, providing high-quality raw data for subsequent accurate judgment and improving the adaptability of the detection scenario.

[0039] Further, step S3 includes: S31. The controller uses Kalman filtering to denoise the main gear torque time-domain signal T(t) to obtain the denoised torque signal T'(t). Specifically, the torque sensor on the spindle assembly 4 collects the main gear torque time-domain signal T(t) at a sampling frequency of 2kHz for more than 10 seconds. This signal reflects the load fluctuation during main gear meshing. The torque fluctuation is smooth during normal meshing. If the main gear has installation eccentricity or tooth surface scratches, the torque signal will show periodic spikes.

[0040] The controller performs Kalman filtering on T(t). Based on the state equation and observation equation of the filtering process, the state equation is: .

[0041] The observation equation is: .

[0042] Where A=1, H=1, because the torque signal is short-term constant during stable operation. Process noise w k-1 With observation noise v k The variances are all set to 10. -6 (N·m) 2 Finally, the denoised torque signal T'(t) is obtained, which filters out electromagnetic interference (such as high-frequency noise generated by motor drive) and mechanical vibration noise (such as slight vibration of base 1) in the signal.

[0043] S32, The controller determines the transmission ratio between the main gear and the auxiliary gear. ; Z1 is the number of teeth on the main gear, Z2 is the number of teeth on the auxiliary gear, and the angle difference signal Δθ(t) is calculated using a pre-set formula. In this embodiment, the first and second circular gratings acquire the time-domain signals θ1(t) of the main gear angle and θ2(t) of the auxiliary gear angle at a sampling frequency of 5kHz, respectively. The controller uses linear interpolation to align the sampling times of the two signals. Based on the transmission ratio i of the main gear and auxiliary gear, formula one is: ; Calculate the rotation angle difference signal. During normal meshing, the rotation angles of the main gear and the auxiliary gear satisfy θ1(t) = iθ2(t), and Δθ(t) is close to 0. If there is a pitch error or installation misalignment, Δθ(t) will fluctuate significantly. For example, when the pitch error of the auxiliary gear exceeds the limit, the peak value of Δθ(t) will rise from 0.1° to more than 0.5°.

[0044] S33. The vibration time-domain signal a1(t) of the main gear and the vibration time-domain signal a2(t) of the auxiliary gear are filtered by a 50Hz high-pass filter to remove low-frequency interference, and the filtered signals a1'(t) and a2'(t) are obtained. The fast Fourier transform formula for a1'(t) and a2'(t) is as follows: ; Where N is the number of sampling points, specifically 4096; k is the index frequency; and j is the imaginary unit.

[0045] The frequency domain signal A1(f) of the main gear and the frequency domain signal A2(f) of the auxiliary gear are obtained.

[0046] S34. Perform feature normalization to obtain a set of feature values, wherein the set of feature values ​​includes: Torque ripple characteristics .

[0047] Corner difference characteristics .

[0048] Peak factor of main gear in time domain .

[0049] Frequency domain amplitude characteristics of 1 times meshing frequency A1(f) m ) is A1(f) in f m The amplitude at (main gear meshing frequency), A ref =100g / Hz, which is the reference amplitude; Peak factor in time domain of auxiliary gear .

[0050] Frequency domain twice the meshing frequency amplitude characteristics A2 (2f) m ) is A2(f) in 2f m The amplitude at that point.

[0051] Among them, peak means peak value, which refers to the maximum absolute value of the vibration time-domain signal within the set analysis period; rms is the root mean square value, which is the quantification of the average energy level of the vibration time-domain signal within the analysis period; the time-domain peak factor is the ratio of instantaneous impact to average energy, and its core function is to amplify the instantaneous anomalies caused by the fault.

[0052] Specifically, the signal preprocessing workflow significantly improves the effectiveness and consistency of multi-source signals through targeted denoising, synchronization, conversion, and normalization. After filtering and denoising, the torque signal eliminates electromagnetic and mechanical interference while retaining the true load fluctuation characteristics. The angle difference calculation eliminates sampling deviations through the synchronization angle signal, accurately reflecting the synchronicity of the main and auxiliary gear transmissions. The time-domain to frequency-domain conversion of the vibration signal focuses on the meshing frequency characteristics, distinguishing the vibration anomalies of the main and auxiliary gears. Normalization eliminates dimensional differences between different signals, allowing feature data to participate in analysis on a unified scale. This entire preprocessing logic not only removes invalid noise but also extracts core features, avoiding model analysis biases caused by cluttered original signals. It provides clean, consistent, and focused input for the multimodal model, improving judgment accuracy from the data source.

[0053] The multimodal analysis model is trained through the following steps: S04-1 The controller acquires at least 1500 sets of sample data pre-set, of which 700 sets are qualified gear pairs and 800 sets are faulty gear pairs.

[0054] S04-2, The controller performs a second preprocessing on each group of sample data.

[0055] S04-3. The controller inputs the second preprocessed sample data into the multimodal analysis model to obtain the trained multimodal analysis model.

[0056] Specifically, by constructing a sample set covering multiple types of defects, standardizing feature annotation, and employing scientific training methods, the reliability and generalization ability of the multimodal analysis model are ensured. The large-scale sample set (including qualified and various fault samples) covers common defects in gear pairs, avoiding overfitting due to a single sample type. Feature annotation associates the actual state with the fault type, providing a clear objective for model learning. The random forest algorithm combined with cross-validation optimizes parameters, enabling the model to stably capture the correlation between multiple features and meshing states. The final trained model possesses both high accuracy and adaptability to the feature differences of different defects, solving the problems of missed detections and high false positive rates in traditional single-threshold judgments, and providing accurate algorithmic support for gear pair state determination.

[0057] In this embodiment, in step S04-1, the types of faulty gear pairs include main gear installation eccentricity, secondary gear tooth surface wear, tooth pitch error, and tooth surface scratches; the number of each type is not less than 150 groups; in step S04-2, the preprocessed sample data is labeled with real state labels and fault type labels.

[0058] Specifically, its dynamic feature weight optimization mechanism effectively solves the problem of insufficient adaptability of multimodal models to different batches of gear pairs. Optimization is triggered by accumulating detection samples, and feature weights are iteratively updated using samples with known states. This allows the model to automatically adapt to differences in feature distribution caused by variations in material, module, and manufacturing process of the gear pairs (such as baseline shifts in vibration characteristics between different batches of gears). Controlling the learning rate and stopping conditions avoids drastic fluctuations in weights, ensuring model stability. This closed-loop design of detection-optimization-iteration maintains long-term detection accuracy, reduces model maintenance costs, and enhances the practicality of the method in mass production.

[0059] Specifically, step S4 includes: S41. The controller inputs the preprocessed data obtained in step S3 into the trained multimodal analysis model.

[0060] S42. Calculate the comprehensive judgment value S using a linear weighting formula. The current weighting formula is: ; w1, w2, w3, w4, w5, and w6 are weighting coefficients; in this embodiment, w1=0.2, w2=0.2, w3=0.1, w4=0.2, w5=0.1, w6=0.2, and the sum of all weighting coefficients is 1.

[0061] For example, the characteristic value of a certain gear pair under test is T f =0.15, Δθ f =0.08, a 1t =0.2, a 1f =0.12, a 2t =0.18, a 2f =0.15, then S=0.2×0.15+0.2×0.08+0.1×0.2+0.2×0.12+0.1×0.18+0.2×0.15=0.15. This value reflects the overall qualification of the gear pair meshing state. The smaller the value, the better the meshing state.

[0062] S43. Result Determination and Output: When S≤0.3, the controller determines that the gear pair meshing state is qualified and outputs a qualified signal and various characteristic values.

[0063] When S>0.3, it is judged as unqualified, and an unqualified signal, S value and the two features that contribute the most are output.

[0064] During the training phase of the multimodal analysis model, at least 1500 sets of sample data are preset, including 700 sets of qualified gear pairs and 800 sets of faulty gear pairs. The distribution of their comprehensive judgment value S is calculated, with 0.3 being the optimal dividing point derived from this distribution. The S value distribution of qualified samples: The S values ​​of the 700 qualified gear pairs (meshing clearance 0.1-0.3mm, tooth surface roughness Ra≤1.6μm) are concentrated in the range of 0-0.25—because the characteristic values ​​of qualified samples are all within the normal range, even after weighting, the S value will not exceed 0.25. The S value distribution of faulty samples: The S values ​​of the 800 faulty samples (including eccentricity, wear, scratches, etc.) are concentrated in the range of 0.35-1.0—faults will cause at least 1-2 core characteristics to increase significantly, such as a when the main gear is eccentric. 1f ≥0.4, weight 0.2, single feature contribution reaches 0.08; superimposed rotation angle difference Δθ f With a value ≥0.5, a weight of 0.2, and a contribution of 0.1, the S-value of two features alone contributes 0.18. After adding other features, the S-value is very likely to exceed 0.35. 0.3 is in the gap between the maximum value of qualified samples (0.25) and the minimum value of faulty samples (0.35), which can avoid misjudging qualified samples as unqualified samples and also avoid misjudging faulty samples as qualified samples.

[0065] Example 3: In this embodiment, the acquisition component also includes a second torque sensor, which is disposed between the power coupling 363 and the power shaft seat 364, and is used to acquire the time domain signal T2(t) of the auxiliary gear torque.

[0066] This embodiment provides another detection method for the gear pair detection device described in Embodiment 1, including the following steps: S100: The controller controls the first detection assembly 2 to approach and abut against the positioning main gear, the controller controls the main shaft assembly 4 to drive the main gear to rotate, and the controller controls the second detection assembly 3 to drive the auxiliary gear to rotate and adjust its height to approach and mesh with the main gear.

[0067] S200 During the meshing process, the controller receives the analysis data collected by the acquisition component. The analysis data includes: the main gear torque time domain signal T1(t), the secondary gear torque time domain signal T2(t), the main gear angle time domain signal θ1(t), the secondary gear angle time domain signal θ2(t), the main gear vibration time domain signal a1(t), and the secondary gear vibration time domain signal a2(t).

[0068] S300 and the controller perform a third preprocessing of the analysis data to obtain preprocessed data.

[0069] S400: The controller inputs the preprocessed data into the trained multimodal analysis model and outputs a comprehensive judgment value S. When S≤0.3, the gear pair meshing state is deemed qualified; when S>0.3, it is deemed unqualified.

[0070] Specifically, step S100 includes: S101. The main gear is driven to rotate at a low speed of 20~40r / min through the main shaft assembly 4, and the auxiliary gear is driven to rotate through the second detection assembly 3. The rotation speed of the auxiliary gear is consistent with that of the main gear. The auxiliary gear is driven to approach and mesh with the main gear through the second detection assembly 3.

[0071] S102. An axial pressure of 20~30N is applied to the auxiliary gear through the first detection assembly 2, and an axial pressure of 70~100N is applied to the main gear through the second detection assembly 3.

[0072] S103, the main gear and auxiliary gear continue to rotate through the first detection assembly 2 and the second detection assembly 3, and the rotational speed of the main gear and auxiliary gear is increased to 150~200 r / min.

[0073] Specifically, step S300 includes: S301. The controller denoises the main gear torque time-domain signal T1(t) using a Kalman filter to obtain the denoised torque signal T1'(t); the specific process is the same as step S31 in Example 2, and will not be described in detail here. The controller denoises the secondary gear torque time-domain signal T2(t) using a Kalman filter to obtain the denoised torque signal T2'(t).

[0074] S302 is the same as step S32 in Example 2, and will not be described in detail here.

[0075] S303 is the same as step S33 in Example 2, and will not be described in detail here.

[0076] S304. Perform feature normalization to obtain the following eigenvalue set: Main gear torque ripple characteristics .

[0077] Torque ripple characteristics of the auxiliary gear .

[0078] Rotation angle difference characteristic △θ f ; Peak factor a in the time domain of the main gear 1t Frequency domain amplitude characteristics at 1 times meshing frequency (a) 1f ; Peak factor a in the time domain of the auxiliary gear 2t Frequency domain twice the meshing frequency amplitude characteristic a 2f The calculation process is the same as in Example 2, and will not be described in detail here.

[0079] Step S400 includes: S401. Input the preprocessed data obtained in step S300 into the multimodal analysis model in a fixed order.

[0080] S402. Calculate the comprehensive judgment value S using the linear weighting formula. The linear weighting formula is: ; w1, w2, w3, w4, w5, w6, and w7 are weighting coefficients; in this embodiment, w1=0.18, w2=0.17, w3=0.18, w4=0.12, w5=0.15, w6=0.1, and w7=0.1. The sum of all weighting coefficients is 1.

[0081] For example, the characteristic value of a certain gear pair under test is T 1f =0.12, T 2f =0.35, Δθ f =0.08, a 1t =0.15, a 1f =0.10, a 2t =0.12, a 2f =0.18, then S=0.18×0.12+0.17×0.35+0.18×0.08+0.12×0.15+0.15×0.10+0.1×0.12+0.1×0.18=0.156. This value reflects the overall qualification of the gear pair meshing state. The smaller the value, the better the meshing state.

[0082] S403. Result Determination and Output: When S≤0.3, the controller determines that the gear pair meshing state is qualified and outputs a qualified signal and various characteristic values; When S>0.3, it is judged as unqualified, and an unqualified signal, S value and the two features that contribute the most are output.

[0083] Specifically, after adding the torque signal of the auxiliary gear, the detection method significantly improves the accuracy of identifying and locating faults such as abnormal load on the auxiliary gear, while the detection time increases very little, thus balancing accuracy and efficiency. In particular, for defects that are prone to misjudgment, such as slight tooth surface wear of the auxiliary gear, this embodiment can accurately identify them through the slight fluctuation of T2f, reducing the misjudgment rate.

[0084] Example 4: This embodiment provides a gear pair detection device, which includes all the structures of the gear pair detection device described in Embodiment 1.

[0085] like Figures 10-12As shown, in this embodiment, a housing 5 is also provided on the base 1, which covers the first detection assembly 2, the second detection assembly 3, and the main shaft assembly 4. An automatic sliding door 51, a door opening / closing driver 52, and a door opening / closing control panel 53 are provided on the housing 5. Specifically, a first guide rail 54 and a second guide rail 55 are provided at the opening of the housing 5. The first guide rail 54 is located at the upper edge of the opening, and the second guide rail 55 is located at the lower edge of the opening. The top of the automatic sliding door 51 is slidably mounted on the first guide rail 54 via a slider, and the bottom of the automatic sliding door 51 is movably mounted on the second guide rail 55 via rollers. The door opening / closing driver 52 is installed on the housing 5. In this embodiment, the door opening / closing driver 52 is a cylinder with electromagnetic induction. The door opening / closing driver 52 is connected to the automatic sliding door 51, and the door opening / closing control panel 53 is connected to the door opening / closing driver 52 to control the door opening / closing driver 52 to drive the automatic sliding door 51 to slide along the first guide rail 54 and the second guide rail 55, thereby opening and closing the housing 5.

[0086] Specifically, the embodiment significantly improves the environmental adaptability and operational safety of the gear pair detection device by adding an outer shell 5 and a matching automatic sliding door 51 structure to the base 1. The outer shell 5 completely protects the first detection assembly 2, the second detection assembly 3, and the main shaft assembly 4, effectively isolating dust, moisture, and airflow interference from the external environment, preventing these impurities from adhering to the sensor surface or entering the transmission mechanism and affecting detection accuracy. It also blocks debris that may be generated during gear rotation during detection and reduces the transmission of external vibrations (such as workshop floor vibration) to the core detection components, creating a stable internal environment for the accurate acquisition of signals such as torque, angle, and vibration. The automatic sliding door 51 achieves smooth sliding through the cooperation of guide rails, sliders, and rollers. Combined with the linkage control of the door opening and closing control panel 53 and the driver, manual opening and closing of the door is not required, making operation convenient and avoiding device positioning deviations caused by manual contact. It is especially suitable for the automated flow requirements in batch detection scenarios.

[0087] Furthermore, the overall reliability and adaptability of the testing device have been enhanced: the enclosed protection of the outer casing 5 allows the device to adapt to complex production environments such as dusty and humid workshops, extending the service life of core components; the automated control of the automatic sliding door 51 can be deeply integrated with the testing process (such as automatic door opening during gear clamping and automatic door closing after testing starts), reducing manual intervention and minimizing testing errors caused by human error (such as environmental interference caused by untimely door closing), while improving the standardization and efficiency of the testing process. Overall, the addition of the outer casing 5 and the automatic sliding door 51, while retaining the core testing capabilities of the device in Example 1, compensates for its shortcomings in environmental protection and automated operation, making the device more suitable for the high-precision and high-stability testing requirements of industrial mass production.

[0088] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0089] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0090] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that they are in indirect contact through an intermediate medium. Furthermore, "above," "over," or "on top" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," or "beneath" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0091] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0092] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A gear pair detection device, characterized in that, include: Base (1), first detection assembly (2), second detection assembly (3), spindle assembly (4), acquisition component and controller; The first detection assembly (2), the second detection assembly (3), and the spindle assembly (4) are mounted on the base (1); the acquisition component is mounted on the second detection assembly (3) and the spindle assembly (4), the controller is mounted on the base (1), and the acquisition component is connected to the controller; The spindle assembly (4) is used to place the main gear and drive the main gear to rotate, and the first detection assembly (2) is used to position the main gear; The second inspection assembly (3) has a secondary gear inspection station, which is used to drive the secondary gear to rotate and apply pressure to the secondary gear in the axial direction, and drive the secondary gear to approach and mesh with the main gear; The base (1) is filled with concrete, and the outer side of the base (1) has a counterweight installation groove (11). The acquisition component is used to acquire the torque time-domain signal, angle time-domain signal, and vibration time-domain signal of the main gear and the auxiliary gear; the controller has a pre-trained multimodal analysis model, which is used to analyze the data acquired by the acquisition component to determine whether the meshing state of the main gear and the auxiliary gear is qualified.

2. The gear pair detection device as described in claim 1, characterized in that, The data acquisition components include a torque sensor, a first circular grating, a second circular grating, a first vibration sensor, and a second vibration sensor; The torque sensor, the first circular grating and the first vibration sensor are mounted on the spindle assembly (4), and the second circular grating and the second vibration sensor are mounted on the second detection assembly (3).

3. The gear pair detection device as described in claim 1, characterized in that, The first detection assembly (2) includes a first column (21), a first slide plate (22), a first drive mechanism (23), and a first bearing (24); The first column (21) is fixed to the base (1), and the first slide plate (22) is slidably disposed on the first column (21); the first bearing (24) is mounted on the first slide plate (22) and is located above the main shaft assembly (4); the first drive mechanism (23) is disposed on the first column (21), and the first drive mechanism (23) is connected to the first slide plate (22) to drive the first slide plate (22) to slide up and down, thereby causing the first bearing (24) to move closer to and away from the main gear.

4. The gear pair detection device as described in claim 2, characterized in that, The second detection assembly (3) includes a second column (31), a column drive mechanism (32), a second slide plate (33), a second drive mechanism (35), a power head (36), a third slide plate (37), a third drive mechanism (38), and a second shaft seat (39); The second column (31) is movably mounted on the base (1), the column drive mechanism (32) is mounted on the base (1), and the column drive mechanism (32) is connected to the second column (31) to drive the second column (31) to move closer to and away from the spindle assembly (4). The second slide plate (33) and the third slide plate (37) are slidably mounted on the second column (31). The second slide plate (33) is located above the third slide plate (37). The power head (36) is located on the second slide plate (33), and the second bearing seat (39) is located on the third slide plate (37). The second drive mechanism (35) and the third drive mechanism (38) are mounted on the second column (31). The second drive mechanism (35) is connected to the second slide plate (33) to drive the second slide plate (33) to slide up and down, causing the power head (36) to move closer to and away from the auxiliary gear. The third drive mechanism (38) is connected to the third slide plate (37) to drive the third slide plate (37) to slide up and down, causing the second bearing seat (39) to adjust the height of the auxiliary gear. The acquisition component is located on the power head (36).

5. A detection method for a gear pair detection device as described in any one of claims 1-4, characterized in that, Including the following steps: S1. Control the first detection assembly (2) to approach and abut the positioning main gear through the controller, control the main shaft assembly (4) to drive the main gear to rotate through the controller, and control the second detection assembly (3) to drive the auxiliary gear to rotate and adjust its height to approach and mesh with the main gear through the controller. S2. During the meshing process, the controller receives the analysis data collected by the acquisition component. The analysis data includes: the main gear torque time domain signal T(t), the main gear angle time domain signal θ1(t), the secondary gear angle time domain signal θ2(t), the main gear vibration time domain signal a1(t), and the secondary gear vibration time domain signal a2(t). S3. The controller performs the first preprocessing of the analysis data to obtain preprocessed data. S4. Then the controller inputs the preprocessed data into the trained multimodal analysis model and outputs the comprehensive judgment value S. Specifically, when S≤0.3, the gear pair meshing state is considered qualified; when S>0.3, it is considered unqualified.

6. The detection method as described in claim 5, characterized in that, Step S1 includes: S11. The controller controls the main shaft assembly (4) to drive the main gear to rotate at a low speed of 20~40r / min. The controller controls the second detection assembly (3) to drive the auxiliary gear to rotate. The rotation speed of the auxiliary gear is consistent with that of the main gear. The controller controls the second detection assembly (3) to drive the auxiliary gear to approach and mesh with the main gear. S12. The controller controls the first detection assembly (2) to apply an axial pressure of 20~30N to the auxiliary gear, and the controller controls the second detection assembly (3) to apply an axial pressure of 70~100N to the main gear. S13. The controller controls the first detection assembly (2) and the second detection assembly (3) to continue driving the main gear and the auxiliary gear to rotate, and increases the speed of the main gear and the auxiliary gear to 150~200r / min.

7. The detection method as described in claim 5, characterized in that, Step S3 includes: S31. The controller uses Kalman filtering to denoise the main gear torque time-domain signal T(t) to obtain the denoised torque signal T'(t). S32, The controller determines the transmission ratio between the main gear and the auxiliary gear. ; Z1 is the number of teeth on the main gear, Z2 is the number of teeth on the secondary gear, and the rotation angle difference signal Δθ(t) is calculated using a pre-set formula (Formula 1); the formula (Formula 1) is: ； S33. The controller removes low-frequency interference from the main gear vibration time-domain signal a1(t) and the auxiliary gear vibration time-domain signal a2(t) by using a 50Hz high-pass filter, and obtains the filtered signals a1'(t) and a2'(t). Perform a fast Fourier transform on a1'(t) and a2'(t) to obtain the frequency domain signal A1(f) of the main gear and the frequency domain signal A2(f) of the auxiliary gear. S34. Perform feature normalization to obtain the eigenvalue set: The feature set includes torque fluctuation feature T. f Angular difference characteristics △θ f Peak factor a in the time domain of the main gear 1t Frequency domain 1 times meshing frequency amplitude characteristics a 1f Peak factor a in the time domain of the auxiliary gear 2t And frequency domain twice the meshing frequency amplitude characteristic a 2f .

8. The detection method as described in claim 7, characterized in that, Step S4 includes: S41. The controller inputs the preprocessed data obtained in step S3 into the trained multimodal analysis model; S42. Calculate the comprehensive judgment value S using a linear weighting formula; the linear weighting formula is: ； w1, w2, w3, w4, w5, and w6 are pre-set weighting coefficients; S43. Result Determination and Output: When S≤0.3, the controller determines that the gear pair meshing state is qualified and outputs a qualified signal and various characteristic values; When S>0.3, it is judged as unqualified, and an unqualified signal, S value, and the two features that contribute the most in the feature value group are output.

9. The detection method as described in claim 5, characterized in that, The multimodal analysis model is trained through the following steps: S04-1. Obtain at least 1500 sets of sample data pre-set, of which 700 sets are qualified gear pairs and 800 sets are faulty gear pairs; S04-2, Perform a second preprocessing on each group of sample data; S04-3. Input the second preprocessed sample data into the multimodal analysis model to obtain the trained multimodal analysis model.

10. The detection method as described in claim 9, characterized in that, In S04-1, the types of faulty gear pairs include main gear installation eccentricity, secondary gear tooth surface wear, tooth pitch error, and tooth surface scratches; the number of each type is not less than 150 groups. In step S04-2, the sample data after the second preprocessing is labeled with a true state label and a fault type label.

Images