A motor virtual welding detection method and system based on multi-modal signal fusion

Abstract

This invention relates to the field of motor defect detection technology, and provides a method and system for detecting motor cold solder joints based on multimodal signal fusion. The method includes: identifying hot spot locations in infrared images to determine the corresponding motor structure, acquiring real-time vibration spectrum data at the structure, comparing it with the vibration spectrum characteristics of the corresponding motor structure constructed from motor parameters, and determining whether the vibration matches its vibration spectrum characteristics to determine whether the motor has cold solder joint defects. By using multimodal data fusion of vibration and temperature for motor detection, the efficiency of motor detection is improved.

Description

Technical Field

[0001] This invention relates to the field of motor defect detection technology, and in particular to a method and system for detecting motor solder joint defects based on multimodal signal fusion. Background Technology

[0002] Poor soldering in motors is a common quality defect. Although it will not immediately render the motor unusable, it will gradually cause a series of problems as the motor runs, eventually leading to motor failure or even safety accidents.

[0003] Existing technologies generally employ manual visual inspection, but manual inspection is highly dependent on personnel experience, cannot detect internal defects, and is inefficient. In the published document CN118706966A, an ultrasonic sensor is used to collect signals from motor welds, followed by filtering and noise reduction to remove environmental noise and interference signals. It can detect weld defects through the reflection and propagation characteristics of high-frequency sound waves. However, ultrasonic testing also relies on the experience and skills of the inspectors, making it prone to misjudgments and missed detections. Furthermore, the complex signal processing and analysis process increases the difficulty and time required for inspection, and ultrasonic testing equipment is expensive. In the published document CN112729529A, an acoustic sensor is used to collect noise from the motor's idling. The sound segment is divided into multiple audio segments, and a two-receptive-field network is used to analyze and extract defect features. The defect status of the motor is then determined based on these features. However, this solution only processes the sound signal and cannot comprehensively detect defects by combining other signals, resulting in low accuracy. Summary of the Invention

[0004] This invention provides a method for detecting poor solder joints in motors based on multimodal signal fusion, which solves the problem of low efficiency in motor defect detection in the prior art.

[0005] The first aspect of this invention provides a method for detecting poor solder joints in motors based on multimodal signal fusion, comprising:

[0006] Obtain motor parameters and motor structure, and construct vibration spectrum characteristics corresponding to each motor structure based on the motor parameters;

[0007] Infrared images of the motor are acquired in real time, and hotspot locations are identified in the infrared images. When a hotspot is identified in the infrared images, vibration spectrum data of the corresponding motor structure is obtained based on the location of the hotspot.

[0008] Determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, determine that the motor structure has a poor welding defect.

[0009] Optionally, whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure further includes:

[0010] The characteristic amplitude at the peak is extracted from the vibration spectrum data, and the temperature difference between the hot spot temperature and the motor ambient temperature is identified based on the infrared image. It is then determined whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of poor welding. If so, it is determined that the motor structure has a poor welding defect.

[0011] Optionally, after obtaining the vibration spectrum data of the corresponding motor structure based on the hotspot location, the method further includes:

[0012] Identify the peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics to obtain non-false solder characteristic spectrum data. Determine whether it overlaps with the spectrum data that matches the vibration spectrum characteristics. If so, correct the false solder characteristic spectrum data with the non-false solder characteristic spectrum data. Generate the corresponding heating situation based on the non-false solder characteristic frequency spectrum data and correct the temperature difference value.

[0013] Optionally, determining whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure further includes:

[0014] The system acquires motor audio data and temperature distribution data, converts the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors, inputs them into a preset neural network model, and outputs the probability of poor soldering and defect classification results.

[0015] The second aspect of this application provides a motor cold solder joint detection system based on multimodal signal fusion, comprising:

[0016] The feature construction module is used to obtain motor parameters and motor structure, and construct vibration spectrum features corresponding to each motor structure based on the motor parameters.

[0017] The vibration data acquisition module is used to acquire the infrared image of the motor in real time and identify the hot spot location in the infrared image; when a hot spot is identified in the infrared image, the vibration spectrum data of the corresponding motor structure is acquired according to the location of the hot spot.

[0018] The defect detection module is used to determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, it is determined that the motor structure has a poor welding defect.

[0019] Optionally, the defect detection module further includes determining whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure.

[0020] The characteristic amplitude at the peak is extracted from the vibration spectrum data, and the temperature difference between the hot spot temperature and the motor ambient temperature is identified based on the infrared image. It is then determined whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of poor welding. If so, it is determined that the motor structure has a poor welding defect.

[0021] Optionally, after obtaining the vibration spectrum data of the corresponding motor structure based on the hotspot location, the defect detection module further includes:

[0022] Identify the peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics to obtain non-false solder characteristic spectrum data. Determine whether it overlaps with the spectrum data that matches the vibration spectrum characteristics. If so, correct the false solder characteristic spectrum data with the non-false solder characteristic spectrum data. Generate the corresponding heating situation based on the non-false solder characteristic frequency spectrum data and correct the temperature difference value.

[0023] Optionally, the defect detection module, in determining whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure, further includes:

[0024] The system acquires motor audio data and temperature distribution data, converts the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors, inputs them into a preset neural network model, and outputs the probability of poor soldering and defect classification results.

[0025] A third aspect of this application provides a method and apparatus for detecting poor solder joints in motors based on multimodal signal fusion, the apparatus comprising a processor and a memory:

[0026] The memory is used to store program code and transmit the program code to the processor;

[0027] The processor is used to execute, according to the instructions in the program code, a method for detecting motor solder joint defects based on multimodal signal fusion as described in any of the first aspects of the present invention.

[0028] The fourth aspect of this application provides a computer-readable storage medium for storing program code for executing a method for detecting motor solder joint defects based on multimodal signal fusion as described in any of the first aspects of this invention.

[0029] As can be seen from the above technical solutions, the present invention has the following advantages: hot spot locations are identified in infrared images to determine the corresponding motor structure, and real-time vibration spectrum data at the structure is obtained. This data is then compared with the vibration spectrum characteristics of the corresponding motor structure constructed from motor parameters to determine whether the vibration matches its vibration spectrum characteristics and whether the motor has a poor soldering defect. Multimodal data fusion of vibration and temperature is used for motor detection, which improves the efficiency of motor detection. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of a method for detecting poor solder joints in motors based on multimodal signal fusion;

[0032] Figure 2 This is a structural diagram of a motor solder joint detection system based on multimodal signal fusion. Detailed Implementation

[0033] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0034] This invention provides a method for detecting poor solder joints in motors based on multimodal signal fusion, which solves the problem of low efficiency in motor defect detection in the prior art.

[0035] Please see Figure 1 , Figure 1 This is the first flowchart of a method for detecting poor solder joints in motors based on multimodal signal fusion, provided in an embodiment of the present invention.

[0036] S100: Obtain motor parameters and motor structure, and construct vibration spectrum characteristics corresponding to each motor structure based on the motor parameters;

[0037] It should be noted that motor parameters include the motor power supply frequency, switching frequency, and the number of rotor slots. Motors are structurally divided into stator, rotor, and power devices. Different motor structures exhibit different vibration characteristics due to poor soldering. For example, poor soldering of stator winding leads or internal connections results in vibration characteristics similar to a three-phase current imbalance fault. The frequencies in the vibration spectrum are integer multiples of the power supply frequency. When the resistance of a phase increases due to poor soldering, the current in that phase decreases, disrupting the three-phase current balance. The interaction of positive and negative magnetic fields generates a pulsating force that is an integer multiple of the power supply frequency. This force acts on the stator core and frame, exciting vibrations. The vibration amplitude fluctuates periodically at twice the power supply frequency. A prominent peak will appear at the corresponding frequency. When the power device is poorly soldered to the heat sink or PCB, it will affect the switching characteristics of the inverter, thus exhibiting vibration characteristics related to the switching frequency. The frequency in the vibration spectrum characteristics is the switching frequency of several kilohertz. This poor soldering causes the temperature of the power device to rise sharply during operation, changing the switching characteristics of the device or causing the protection circuit to activate, thereby distorting the output PWM waveform. When there is a poor soldering in the rotor, the frequency in the vibration spectrum characteristics is proportional to the number of rotor slots and the slip frequency. Broken or poorly soldered conductors will cause rotor current imbalance, generating an asymmetrical magnetic field and exciting vibration. Each structure in the motor has a corresponding vibration spectrum characteristic when there is a poor soldering.

[0038] S200 acquires infrared images of the motor in real time and identifies hotspot locations in the infrared images; when a hotspot is identified in the infrared images, it acquires the vibration spectrum data of the corresponding motor structure based on the location of the hotspot.

[0039] It should be noted that infrared cameras can be used to periodically or in real-time capture images of the motor to obtain infrared images. Under normal operating conditions, the motor temperature distribution should be uniform. Infrared images can provide a global and intuitive temperature distribution map. Infrared imaging has strong positioning capabilities and can directly detect solder joints that are overheating due to high resistance. If the temperature of a certain part is significantly higher than the surrounding area, it can be considered a suspected hot spot. When a localized hot spot is located at a solder joint and its adjacent wires and terminals, it can be determined that there is a cold solder joint in the stator winding or power lead. Current flows through the high-resistance contact surface of the cold solder joint, generating a large amount of heat. This heat cannot dissipate quickly enough, causing the local temperature of the solder joint, wire insulation, and terminal metal to rise sharply, much higher than other parts of the motor. Infrared thermal imagers can intuitively and accurately locate the solder joint with abnormally high temperature. When the chip body is overheated, it can be considered as heat generation from a cold solder joint between the power device and the heat sink. This cold solder joint affects the chip's heat dissipation path, and the cold solder joint forms a thermal barrier between the chip and the heat sink. The heat generated inside the chip due to switching losses and conduction losses cannot be conducted to the heat sink through the normal path, causing heat to accumulate inside the chip. In the infrared imaging of the motor during operation, the heat sink temperature is uneven. The temperature of the faulty chip is lower because it cannot dissipate heat, while the heat sink temperature of the normally functioning chip is higher. When the entire motor winding is uniformly overheated, it can be regarded as a poor solder joint between the rotor bars and the end ring. Since the heat comes from the motor's own reduced efficiency, the poor solder joint causes the entire motor to heat up, rather than a local hot spot. The resistance of the rotor circuit increases due to the poor solder joint, which leads to an increase in slip under the same load. More electrical energy is not converted into mechanical energy. In order to output the same power, the motor needs to absorb more electrical energy from the grid, which leads to an increase in stator current, thus causing the stator and rotor to overheat in general.

[0040] Based on the hotspot distribution in the infrared image, we can determine whether there is an overheated motor component structure. Then, we can obtain the current vibration data of the motor structure by using vibration sensors such as accelerometers to obtain vibration spectrum data.

[0041] S300: Determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, determine that the motor structure has a poor welding defect.

[0042] It should be noted that the vibration to be identified in this embodiment is caused by electromagnetic force. The interaction of the stator and rotor magnetic fields generates rotational torque, and also generates radial and tangential electromagnetic force waves. Factors that cause magnetic field imbalance, such as current imbalance, will generate abnormal electromagnetic force waves, thereby exciting vibration. Therefore, the vibration spectrum characteristics of each motor structure in step S100 are actually caused by poor welding at different motor structures. When the characteristics in the actual vibration spectrum data are consistent with the vibration spectrum characteristics of the corresponding motor structure, that is, the amplitude peak value meets the characteristics in the aforementioned step S100, the poor welding defect of the motor structure can be determined by signal fusion of temperature location and vibration spectrum characteristics. If there are abnormal hot spots in the infrared image, but the vibration characteristics cannot match, the heat generation may be caused by other reasons, such as loose fastening bolts or cooling fan failure, which generates mechanical friction heat or local poor heat dissipation, and the vibration has no specific electromagnetic characteristics.

[0043] In this embodiment, the corresponding motor structure is determined by identifying hotspot locations in infrared images, and real-time vibration spectrum data at that structure is obtained. This data is then compared with the vibration spectrum characteristics of the corresponding motor structure constructed from motor parameters to determine whether the vibration matches its vibration spectrum characteristics and whether the motor has a poor soldering defect. Multimodal data fusion of vibration and temperature is used for motor detection, which improves the efficiency of motor detection.

[0044] The above is a detailed description of the first embodiment of the motor cold solder joint detection method based on multimodal signal fusion provided in this application. The following is a detailed description of the second embodiment of the motor cold solder joint detection method based on multimodal signal fusion provided in this application.

[0045] In this embodiment, a method for detecting poor weld joints in motors based on multimodal signal fusion is further provided. In step S300, determining whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure specifically includes:

[0046] The characteristic amplitude at the peak is extracted from the vibration spectrum data, and the temperature difference between the hot spot temperature and the motor ambient temperature is identified based on the infrared image. It is then determined whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of poor welding. If so, it is determined that the motor structure has a poor welding defect.

[0047] It should be noted that in the aforementioned steps, the frequency is determined to conform to the characteristics of the vibration spectrum by the time point of the peak in the vibration spectrum data, without considering the amplitude. However, in this embodiment, the amplitude can be determined after the frequency is determined, and only the characteristic amplitude at the peak of the frequency that conforms to the characteristics of the vibration spectrum is considered.

[0048] The temperature rise in the motor is caused by the resistance value of the poor solder joint. The local overheating is determined by the power consumption of the poor solder joint itself. The current flowing through the poor solder joint is the total current of the circuit. Although the actual heat generation power of the poor solder joint is related to the heat generation power and heat dissipation capacity, this heat dissipation can be ignored in the actual operating environment of the motor. In this embodiment, it can be simplified to the higher the heat generation power, the higher the temperature rise, and the degree of overheating depends on the resistance value of the poor solder joint. The root cause of motor vibration is the imbalance of electromagnetic force. Poor soldering will cause the rotor bars to break, which directly causes the rotor resistance to be asymmetrical. When the rotating magnetic field cuts the rotor, the current induced in the intact bars and broken bars is completely different. The electromagnetic force is asymmetrical, and the vibration amplitude is directly proportional to the magnitude of the unilateral magnetic pull of this fluctuation, that is, the degree of electromagnetic force imbalance. The higher the resistance value in the poor solder joint, the more serious the imbalance and the more violent the vibration. Therefore, by identifying whether the resistance value of the poor solder joint corresponding to the temperature rise is consistent with the resistance value of the poor solder joint corresponding to the current difference under the vibration amplitude, it can be determined whether the current motor defect is a poor solder joint, and the defect can be located.

[0049] After obtaining the vibration spectrum data of the corresponding motor structure based on the hot spot location, the process also includes: identifying the peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics, obtaining non-false solder characteristic spectrum data, determining whether it overlaps with the spectrum data that conforms to the vibration spectrum characteristics, and if so, correcting the false solder characteristic spectrum data with the non-false solder characteristic spectrum data; generating the corresponding heating situation based on the non-false solder characteristic frequency spectrum data, and correcting the temperature difference value.

[0050] It should be noted that the detected vibrations include not only those caused by electromagnetic force imbalance due to poor soldering defects, but also mechanical vibrations, such as those caused by direct mechanical problems like rotor imbalance, bearing defects, and misalignment. Both types of vibrations will be detected during vibration testing, and the mechanical and electromagnetic vibrations will be superimposed. In the vibration spectrum data, peaks matching the vibration spectrum characteristics can be identified, and the frequencies of the remaining peaks can be identified to determine if they match the peaks matching the vibration spectrum characteristics. For example, the vibration spectrum data may contain peaks with frequency matching the spectral characteristics of rotor poor soldering. At 0.5s, 1.0s, and 1.5s, the peak amplitudes are 2mm, 1.7mm, and 2mm, respectively. In contrast, the peak amplitudes in the non-soldering characteristic spectrum data appear at 0.7s, 0.9s, 1.1s, and 1.3s, with a peak amplitude of 0.3mm. Based on the frequency of the non-soldering characteristic spectrum data, it can be determined that there is overlap with the spectrum data that conforms to the vibration spectrum characteristics. Therefore, the soldering characteristic spectrum data is corrected by adjusting the peak amplitudes at 0.5s, 1.0s, and 1.5s to 1.7mm, thus obtaining the amplitude of vibration caused solely by electromagnetic force.

[0051] The heat generated by mechanical vibration is generally frictional heat. This heat generation is related to the magnitude of the friction force; that is, the greater the vibration amplitude, the greater the friction force and the greater the heat generation. The heat generated by mechanical vibration is calculated based on experience or a preset ratio. The corresponding heat generation power is calculated using the amplitude of the non-fraudulent solder joint characteristic spectrum data. Then, the corresponding value is subtracted from the temperature difference value. The corrected temperature difference value and vibration amplitude can improve the accuracy of the previous steps in determining whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of a fraudulent solder joint.

[0052] In step S300, determining whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure specifically includes: acquiring motor audio data and temperature distribution data, converting the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors and inputting them into a preset neural network model, and outputting the probability of false welding and the defect classification results.

[0053] It should be noted that motors will produce noise during operation, which is normal sound from rotation. However, if abnormal vibration occurs, the motor's audio data can be obtained by placing a radio receiver near the motor. The peaks and amplitudes of the audio data can also reflect the motor's operating status. The raw signals of the collected vibration spectrum data, motor audio data, and temperature distribution data are preprocessed, such as by filtering and denoising. The preset neural network model can use support vector machine (SVM) or random forest network. The model is first trained with a large number of normal solder joints and various types of poor solder joint samples to learn the complex mapping relationship between poor solder joint defects and multimodal features. After the multimodal data is input into the model, it can output the probability and location of poor solder joints.

[0054] The above is a detailed description of a motor solder joint detection method based on multimodal signal fusion based on the first aspect of this application. The following is a detailed description of an embodiment of a motor solder joint detection system based on multimodal signal fusion based on the second aspect of this application.

[0055] Please see Figure 2 , Figure 2 This is a structural diagram of a motor solder joint detection system based on multimodal signal fusion. This embodiment provides a motor solder joint detection system based on multimodal signal fusion, including:

[0056] Feature construction module 10 is used to obtain motor parameters and motor structure, and construct vibration spectrum features corresponding to each motor structure based on the motor parameters;

[0057] The vibration data acquisition module 20 is used to acquire the infrared image of the motor in real time and identify the hot spot location in the infrared image; when a hot spot is identified in the infrared image, the vibration spectrum data of the corresponding motor structure is acquired according to the hot spot location.

[0058] The defect detection module 30 is used to determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, it is determined that the motor structure has a poor welding defect.

[0059] In the defect detection module 30, whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure specifically includes:

[0060] The characteristic amplitude at the peak is extracted from the vibration spectrum data, and the temperature difference between the hot spot temperature and the motor ambient temperature is identified based on the infrared image. It is then determined whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of poor welding. If so, it is determined that the motor structure has a poor welding defect.

[0061] In the defect detection module 30, after obtaining the vibration spectrum data of the corresponding motor structure based on the hot spot location, it also includes:

[0062] Identify the peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics to obtain non-false solder characteristic spectrum data. Determine whether it overlaps with the spectrum data that matches the vibration spectrum characteristics. If so, correct the false solder characteristic spectrum data with the non-false solder characteristic spectrum data. Generate the corresponding heating situation based on the non-false solder characteristic frequency spectrum data and correct the temperature difference value.

[0063] In the defect detection module 30, determining whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure specifically includes:

[0064] This application also provides a method and device for detecting motor weld defects based on multimodal signal fusion, comprising a processor and a memory: the memory is used to store program code and transmit the program code to the processor; the processor is used to execute the above-mentioned method for detecting motor weld defects based on multimodal signal fusion according to the instructions in the program code. The method acquires motor audio data and temperature distribution data, converts the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors, inputs them into a preset neural network model, and outputs the probability of weld defects and defect classification results.

[0065] The fourth aspect of this application provides a computer-readable storage medium for storing program code for executing the above-described method for detecting motor solder joint defects based on multimodal signal fusion.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting poor solder joints in motors based on multimodal signal fusion, characterized in that... include: The motor parameters and motor structure are obtained. The motor parameters include the motor power supply frequency, switching frequency, and the number of rotor slots in the motor. The motor structure consists of a stator, a rotor, and power devices. Vibration spectrum characteristics corresponding to each motor structure are constructed based on the motor parameters. The system acquires infrared images of the motor in real time and identifies hotspot locations within these images. When a hotspot is identified in the infrared image, the system obtains vibration spectrum data of the corresponding motor structure based on the hotspot location. It then identifies peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics to obtain non-fractured weld characteristic spectrum data. The system determines whether this data overlaps with spectrum data that matches the vibration spectrum characteristics. If so, the non-fractured weld characteristic spectrum data is used to correct the non-fractured weld characteristic spectrum data. Based on the non-fractured weld characteristic spectrum data, the system generates corresponding heating data and corrects for temperature differences. Determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, determine that the motor structure has a poor welding defect. The determination of whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure further includes: extracting the characteristic amplitude at the peak based on the vibration spectrum data, identifying the temperature difference between the hot spot temperature and the motor ambient temperature based on the infrared imaging image, and determining whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of cold solder joint. If so, it is determined that the motor structure has a cold solder joint defect.

2. The method for detecting motor solder joint defects based on multimodal signal fusion according to claim 1, characterized in that, The determination of whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure specifically includes: The system acquires motor audio data and temperature distribution data, converts the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors, inputs them into a preset neural network model, and outputs the probability of poor soldering and defect classification results.

3. A motor solder joint detection system based on multimodal signal fusion, characterized in that, include: The feature construction module is used to obtain motor parameters and motor structure. The motor parameters include motor power frequency, switching frequency, and the number of rotor slots in the motor. The motor structure consists of stator, rotor, and power devices. Vibration spectrum features corresponding to each motor structure are constructed based on the motor parameters. The vibration data acquisition module is used to acquire infrared images of the motor in real time and identify hotspot locations in the infrared images. When a hotspot is identified in the infrared images, the module acquires the vibration spectrum data of the corresponding motor structure based on the hotspot location, identifies peak frequencies in the vibration spectrum data that do not match the vibration spectrum characteristics, obtains non-false weld characteristic spectrum data, and determines whether it overlaps with spectrum data that matches the vibration spectrum characteristics. If so, the false weld characteristic spectrum data is corrected using the non-false weld characteristic spectrum data. The module also generates the corresponding heating information based on the non-false weld characteristic spectrum data and corrects the temperature difference values. The defect detection module is used to determine whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure. If so, it is determined that the motor structure has a poor welding defect. The determination of whether the vibration spectrum data conforms to the vibration spectrum characteristics of the corresponding motor structure further includes: extracting the characteristic amplitude at the peak based on the vibration spectrum data, identifying the temperature difference between the hot spot temperature and the motor ambient temperature based on the infrared imaging image, and determining whether the characteristic amplitude at the peak and the temperature difference value conform to the correlation of cold solder joint. If so, it is determined that the motor structure has a cold solder joint defect.

4. The motor cold solder joint detection system based on multimodal signal fusion according to claim 3, characterized in that, The defect detection module, which determines whether the vibration spectrum data matches the vibration spectrum characteristics of the corresponding motor structure, specifically includes: The system acquires motor audio data and temperature distribution data, converts the vibration spectrum data, motor audio data, and temperature distribution data into feature vectors, inputs them into a preset neural network model, and outputs the probability of poor soldering and defect classification results.

5. A motor solder joint defect detection device based on multimodal signal fusion, characterized in that, The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute, according to the instructions in the program code, a method for detecting motor solder joint defects based on multimodal signal fusion as described in any one of claims 1-2.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code for executing the motor cold solder joint detection method based on multimodal signal fusion as described in any one of claims 1-2.

Images