A fan main shaft defect detection system

Abstract

The present invention discloses a ventilator main shaft defect detection system, which relates to the technical field of intelligent detection and includes an adjustment module, an analysis module, and a construction module. The system calculates the effective pixel ratio, adjusts the shooting information, collects the operation data of the main shaft, obtains the first main shaft defect through a first analysis, obtains the second main shaft defect through a second analysis, obtains the defect combination through association, calculates the probability of the defect combination, and establishes a first mapping relationship. The present invention ensures the image quality and the accuracy of defect detection by automatically adjusting the shooting parameters, matches the light source angle in combination with the main shaft material model, improves the pertinence and reliability of detection, detects main shaft defects from multiple dimensions, improves the accuracy of defect detection, combines image analysis and operation data analysis, can more comprehensively detect main shaft defects, avoids the limitations of a single detection method, and quickly screens out high-probability defect combinations through probability analysis, thereby improving the speed of defect diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent detection, and in particular to a ventilator main shaft defect detection system. Background Art

[0002] In recent years, fan spindle defect detection technology has developed in a more intelligent and automated direction. By introducing artificial intelligence, big data and Internet of Things technologies, real-time monitoring and automatic diagnosis of spindle defects can be achieved, and multiple detection technologies can be combined, such as the integration of ultrasonic detection and electromagnetic detection technology, to improve the accuracy and reliability of detection.

[0003] At present, a Chinese invention patent with publication number CN107402255B discloses an ultrasonic array online detection system for surface defects of a fan main shaft. The system sends detection instructions to an ultrasonic excitation receiving device through a router. The ultrasonic excitation receiving device controls the multiplexer to select and excite the piezoelectric array elements in the piezoelectric sensor array in turn, and receives the echo signal from the piezoelectric array element. The echo signal is then wirelessly transmitted back to the computer via the router to achieve detection. However, the related art does not adjust the angle of the light source according to the reflection problem of the main shaft model, nor does it adjust the shooting parameters through roughly taken photos, which can easily lead to incorrect recognition of machine vision and is not conducive to the accuracy of detection. The main shaft defects are not divided into surface defects and functional defects for distinction and identification, which is not conducive to the comprehensiveness of detection. The probabilities of main shafts that are likely to appear at the same time are not combined and associated based on historical data, and the detection is simplified through the combined probability, which is not conducive to the simplicity and intuitiveness of detection. Summary of the Invention

[0004] The technical problem solved by the present invention is: the related technology does not adjust the angle of the light source according to the reflection problem of the spindle model, nor does it adjust the shooting parameters through roughly taken photos, which easily leads to incorrect recognition of machine vision, which is not conducive to the accuracy of detection. The defects of the spindle are not divided into surface defects and functional defects for differentiation and identification, which is not conducive to the comprehensiveness of detection. The probabilities of spindles that are prone to appear at the same time are not combined and associated based on historical data, and the detection is simplified through the combined probability, which is not conducive to the simplicity and intuitiveness of detection.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a ventilator main shaft defect detection system, comprising an adjustment module, an analysis module and a construction module;

[0006] The adjustment module obtains the spindle material model, matches the corresponding red light source lamp bead angle according to the spindle material model, sends a start shooting signal, obtains a finished product image, sets the pixels belonging to the spindle in the finished product image as valid pixels, calculates the valid pixel ratio, and adjusts the shooting information according to the valid pixel ratio to obtain a first image. After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends a first power to the ventilator to collect the operation data of the spindle;

[0007] The analysis module preprocesses the first image to obtain a standard defect image, performs a first analysis on the preprocessed first image and the standard defect image, obtains a first main shaft defect based on the first analysis result, and performs a second analysis on the operating data to obtain a second main shaft defect;

[0008] The construction module associates the first spindle defect with the second spindle defect to obtain a defect combination, retrieves historical defect data to obtain each defect combination, calculates the defect combination probability, sets a probability threshold, filters the same defect combination probability according to the probability threshold, and establishes a first mapping relationship between the defect combination that meets the conditions and the defect combination probability.

[0009] As a preferred solution of the fan main shaft defect detection system described in the present invention, the adjustment module obtains the main shaft material model, calls the red light source database, inputs the main shaft material model into the red light source database, matches the lamp bead angle of the red light source corresponding to the main shaft material model, and the red light source adjusts the lamp bead angle according to the lamp bead angle of the red light source corresponding to the main shaft material model. When the adjustment is completed, the red light source sends an adjustment completion signal to the adjustment module.

[0010] As a preferred solution of the fan main shaft defect detection system described in the present invention, wherein: after receiving the adjustment completion signal, the adjustment module sends a start shooting signal to the industrial camera;

[0011] The industrial camera takes pictures of the spindle to obtain the finished product picture;

[0012] The adjustment module sets the pixels belonging to the main axis in the finished image as valid pixels and counts the number of valid pixels and the total number of pixels;

[0013] Calculate the ratio of the number of effective pixels to the total number of pixels and set it as the effective pixel ratio;

[0014] The adjustment module adjusts the shooting information according to the effective pixel ratio to obtain a first image;

[0015] The shooting information includes a shooting position and a magnification, wherein the shooting position is represented as the three-dimensional coordinates of the center of the industrial camera set;

[0016] After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends the first power to the ventilator;

[0017] The operating data of the main shaft is collected, wherein the operating data includes vibration data and sound intensity data.

[0018] As a preferred solution of the fan main shaft defect detection system described in the present invention, wherein: the analysis module preprocesses the first image, and the preprocessing includes filtering processing and grayscale processing;

[0019] Acquire standard defect images, which include standard scratch images, standard rust images, standard crack images, standard wear images and standard excessive bending images, and the corresponding first spindle defect categories are scratches, rust, cracks, wear and excessive bending.

[0020] As a preferred solution of the fan main shaft defect detection system according to the present invention, the analysis module performs a first analysis on the preprocessed first image and the standard defect image, and the analysis logic of the first analysis includes:

[0021] Extracting shape features of each standard defect image, recorded as first features; extracting shape features of the preprocessed first image, recorded as second features; calculating similarities between the second features and each first feature, recorded as first similarities; setting the first value as a similarity threshold; comparing the first similarity with the first value; retaining the corresponding standard defect image when the first similarity is greater than or equal to the first value; otherwise, deleting the corresponding standard defect image;

[0022] The principal axis defect corresponding to the retained standard defect image is set as the first principal axis defect.

[0023] As a preferred solution of the fan main shaft defect detection system described in the present invention, it comprises: obtaining standard dynamic defects, which include improper installation of the standard main shaft and damage to the standard coupling, and corresponding second main shaft defects include improper installation of the main shaft and damage to the coupling, and obtaining historical vibration data and historical sound intensity data corresponding to historical standard dynamic defects.

[0024] As a preferred solution of the fan main shaft defect detection system according to the present invention, the analysis module performs a second analysis on the vibration data and the sound intensity data, and the analysis logic of the second analysis includes:

[0025] Perform high-frequency noise removal and normalization processing on historical vibration data and vibration data, and perform outlier removal and logarithm conversion processing on sound intensity data and historical sound intensity data;

[0026] Calculating the variance value of the vibration data after high-frequency noise removal and normalization and the variance value of the historical vibration data, calculating the intensity change rate of the sound intensity data after outlier removal and logarithm transformation and the intensity change rate of the historical sound intensity data, setting the variance value of the historical vibration data and the intensity change rate of the historical sound intensity data as a vibration feature vector and an intensity feature vector, fusing the vibration feature vector and the intensity feature to obtain a fused feature vector, inputting the fused feature vector and the corresponding standard dynamic defect into a machine learning model for training to obtain a dynamic defect model;

[0027] A fused eigenvector corresponding to the variance value of the vibration data and the intensity change rate of the sound intensity data is obtained, recorded as a third eigenvector, and the third eigenvector is input into the dynamic defect model to obtain the corresponding second main shaft defect.

[0028] As a preferred solution of the fan main shaft defect detection system described in the present invention, the calculation logic of the defect combination probability includes:

[0029] The total number of defect combinations is counted, the number of each identical defect combination is counted, and the ratio of the number of each identical defect combination to the total number of defect combinations is calculated, which is recorded as the defect combination probability.

[0030] As a preferred solution of a fan main shaft defect detection system described in the present invention, wherein: the construction module sets the second value as a probability threshold, compares the probability of each defect combination with the second value, and when the defect combination probability is greater than or equal to the second value, the corresponding defect combination probability and the corresponding defect combination are retained, otherwise the corresponding defect combination probability and the corresponding defect combination are deleted.

[0031] As a preferred solution of the fan main shaft defect detection system described in the present invention, the construction module establishes a first mapping relationship between the retained defect combinations and the corresponding defect combination probabilities, inputs any first defect into the first mapping relationship, and obtains the corresponding second defects and the corresponding defect combination probabilities, or inputs any second defect into the first mapping relationship to obtain the corresponding first defects and the corresponding defect combination probabilities.

[0032] The beneficial effects of the present invention are as follows: by automatically adjusting the shooting parameters, the image quality and the accuracy of defect detection are ensured; the light source angle is matched in combination with the spindle material model to improve the pertinence and reliability of detection; by combining image analysis and operation data analysis, spindle defects are detected from multiple dimensions; standard defect images are used as a reference to improve the accuracy of defect detection; through historical data and probability analysis, potential defects are predicted, the prediction ability and reliability of the system are improved; a mapping relationship between defect combination and probability is established, and data support is provided for subsequent maintenance and decision-making; by combining image analysis and operation data analysis, spindle defects can be detected more comprehensively, avoiding the limitations of a single detection method; high-probability defect combinations are quickly screened out through probability analysis, and the speed of defect diagnosis is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of the basic flow of a fan main shaft defect detection system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0035] Example, see Figure 1 , as one embodiment of the present invention, provides a fan main shaft defect detection system, including an adjustment module, an analysis module and a construction module;

[0036] The adjustment module obtains the spindle material model, matches the corresponding red light source lamp bead angle according to the spindle material model, sends a start shooting signal, obtains a finished product image, sets the pixels belonging to the spindle in the finished product image as valid pixels, calculates the valid pixel ratio, and adjusts the shooting information according to the valid pixel ratio to obtain a first image. After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends a first power to the ventilator to collect the operation data of the spindle;

[0037] The analysis module preprocesses the first image to obtain a standard defect image, performs a first analysis on the preprocessed first image and the standard defect image, obtains a first main shaft defect based on the first analysis result, and performs a second analysis on the operating data to obtain a second main shaft defect;

[0038] The construction module associates the first spindle defect with the second spindle defect to obtain a defect combination, retrieves historical defect data to obtain each defect combination, calculates the defect combination probability, sets a probability threshold, filters the same defect combination probability according to the probability threshold, and establishes a first mapping relationship between the defect combination that meets the conditions and the defect combination probability.

[0039] The present invention ensures image quality and the accuracy of defect detection by automatically adjusting shooting parameters, matches the light source angle in combination with the spindle material model, improves the pertinence and reliability of detection, detects spindle defects from multiple dimensions in combination with image analysis and operation data analysis, uses standard defect images as a reference, improves the accuracy of defect detection, predicts potential defects through historical data and probability analysis, improves the system's prediction ability and reliability, establishes a mapping relationship between defect combinations and probabilities, and provides data support for subsequent maintenance and decision-making. In combination with image analysis and operation data analysis, it can detect spindle defects more comprehensively, avoid the limitations of a single detection method, and quickly screen out high-probability defect combinations through probability analysis, thereby improving the speed of defect diagnosis.

[0040] The adjustment module obtains the spindle material model, calls the red light source database, inputs the spindle material model into the red light source database, matches the lamp bead angle of the red light source corresponding to the spindle material model, and the red light source adjusts the lamp bead angle according to the lamp bead angle of the red light source corresponding to the spindle material model. When the adjustment is completed, the red light source sends an adjustment completion signal to the adjustment module.

[0041] In specific implementation, by matching the spindle material model with the corresponding red light source lamp bead angle, it is possible to optimize the detection of spindles of different materials. Different materials have different reflection and absorption characteristics of light. By adjusting the lamp bead angle, the best image quality is ensured, thereby improving the accuracy of defect detection. The optimized light source angle reduces light interference or insufficient reflection caused by material characteristics, avoiding misjudgment or missed defects. There are many spindle material models, such as steel, aluminum alloy, titanium alloy, etc. The best lamp bead angle is matched for each material model, for example: steel: 30°, aluminum alloy: 45°, titanium alloy: 60°, the lamp bead angle adjustment accuracy can reach ±0.5, from inputting the material model to completing the adjustment, the average time taken is less than 5 seconds, the optimized light source angle shortens the detection time by about 15%, improves the overall detection efficiency, reduces the misjudgment rate to less than 2%, and significantly improves the accuracy of detection.

[0042] After receiving the adjustment completion signal, the adjustment module sends a start shooting signal to the industrial camera;

[0043] The industrial camera takes pictures of the spindle to obtain the finished product picture;

[0044] The adjustment module sets the pixels belonging to the main axis in the finished image as valid pixels and counts the number of valid pixels and the total number of pixels;

[0045] Calculate the ratio of the number of effective pixels to the total number of pixels and set it as the effective pixel ratio;

[0046] The adjustment module adjusts the shooting information according to the effective pixel ratio to obtain a first image;

[0047] The shooting information includes a shooting position and a magnification, wherein the shooting position is represented as the three-dimensional coordinates of the center of the industrial camera set;

[0048] After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends the first power to the ventilator;

[0049] The operating data of the main shaft is collected, wherein the operating data includes vibration data and sound intensity data.

[0050] In the specific implementation, by calculating the effective pixel ratio, the clarity and integrity of the main shaft part in the captured image are ensured, thereby improving the accuracy of defect detection. The shooting information (such as shooting position and magnification) is adjusted according to the effective pixel ratio to further optimize the image quality and ensure that the defects of the main shaft can be clearly captured. The threshold of the effective pixel ratio is set to 0.8, and the three-dimensional coordinate adjustment range of the collection center of the industrial camera is ±10mm. The adjustment accuracy can reach 0.1mm. The average time from sending the start shooting signal to completing the shooting is less than 2 seconds. The average time from completing the shooting to adjusting the shooting parameters and reshooting is less than 3 seconds. The average time from starting the detection to completing the shooting and adjustment is less than 5 seconds. By optimizing the shooting parameters, the defect detection rate is increased by 25%. The vibration data acquisition frequency is 1000Hz, and the sound intensity data acquisition frequency is 500Hz.

[0051] The analysis module preprocesses the first image, wherein the preprocessing includes filtering and grayscale processing;

[0052] Acquire standard defect images, which include standard scratch images, standard rust images, standard crack images, standard wear images and standard excessive bending images, and the corresponding first spindle defect categories are scratches, rust, cracks, wear and excessive bending.

[0053] The analysis module performs a first analysis on the preprocessed first image and the standard defect image. The analysis logic of the first analysis includes:

[0054] Extracting shape features of each standard defect image, recorded as first features; extracting shape features of the preprocessed first image, recorded as second features; calculating similarities between the second features and each first feature, recorded as first similarities; setting the first value as a similarity threshold; comparing the first similarity with the first value; retaining the corresponding standard defect image when the first similarity is greater than or equal to the first value; otherwise, deleting the corresponding standard defect image;

[0055] The principal axis defect corresponding to the retained standard defect image is set as the first principal axis defect.

[0056] In specific implementation, the shape features of the image and operational data (such as vibration data and sound intensity data) are combined to evaluate the health status of the main shaft from multiple dimensions, improving the data integrity and diagnostic accuracy. The similarity threshold is set to 0.8. Combining image data and operational data, the accuracy of defect diagnosis is improved by about 30%.

[0057] Obtain standard dynamic defects, which include improper installation of a standard main shaft and damage to a standard coupling. The corresponding second main shaft defects include improper installation of the main shaft and damage to the coupling. Obtain historical vibration data and historical sound intensity data corresponding to historical standard dynamic defects.

[0058] The analysis module performs a second analysis on the vibration data and the sound intensity data, and the analysis logic of the second analysis includes:

[0059] Perform high-frequency noise removal and normalization processing on historical vibration data and vibration data, and perform outlier removal and logarithm conversion processing on sound intensity data and historical sound intensity data;

[0060] Calculating the variance value of the vibration data after high-frequency noise removal and normalization and the variance value of the historical vibration data, calculating the intensity change rate of the sound intensity data after outlier removal and logarithm transformation and the intensity change rate of the historical sound intensity data, setting the variance value of the historical vibration data and the intensity change rate of the historical sound intensity data as a vibration feature vector and an intensity feature vector, fusing the vibration feature vector and the intensity feature to obtain a fused feature vector, inputting the fused feature vector and the corresponding standard dynamic defect into a machine learning model for training to obtain a dynamic defect model;

[0061] A fused eigenvector corresponding to the variance value of the vibration data and the intensity change rate of the sound intensity data is obtained, recorded as a third eigenvector, and the third eigenvector is input into the dynamic defect model to obtain the corresponding second main shaft defect.

[0062] In specific implementation, by fusing the variance value of vibration data and the intensity change rate of sound intensity data, a comprehensive feature vector is formed, which can more comprehensively reflect the operating status of the spindle. The fused feature vector is trained using a machine learning model, which can automatically learn and identify the characteristic patterns of spindle defects, thereby improving the accuracy of defect detection. Combining vibration data and sound intensity data, the health status of the spindle can be evaluated from multiple dimensions, avoiding the limitations of a single data source.

[0063] The calculation logic of the defect combination probability includes:

[0064] The total number of defect combinations is counted, the number of each identical defect combination is counted, and the ratio of the number of each identical defect combination to the total number of defect combinations is calculated, which is recorded as the defect combination probability.

[0065] In specific implementation, by calculating the probability of defect combinations, the frequency of occurrence of specific defect combinations can be more accurately predicted, allowing preventive measures to be taken in advance. Based on the probability of defect combinations, priority is given to defect combinations with high probability, optimizing the allocation of maintenance resources. Through probabilistic analysis, potential defect combinations can be discovered in a timely manner, reducing the occurrence of equipment failures and improving the overall reliability of the system. Over a period of time, a total of 1,000 defect combinations were detected, and the defect combination of "cracks and improper spindle installation" appeared 150 times, with a probability of 15%.

[0066] The construction module sets the second value as the probability threshold, compares each defect combination probability with the second value, and when the defect combination probability is greater than or equal to the second value, retains the corresponding defect combination probability and the corresponding defect combination; otherwise, deletes the corresponding defect combination probability and the corresponding defect combination.

[0067] In specific implementation, by setting the probability threshold, the system can automatically screen out high-probability defect combinations, help maintenance personnel prioritize these high-risk defect combinations, improve maintenance efficiency, and take preventive maintenance measures by identifying high-probability defect combinations in advance, reduce the occurrence of equipment failures, and improve the overall reliability of the system. Screening defect combinations based on the probability threshold provides a scientific basis for maintenance decisions and helps decision makers formulate more reasonable maintenance plans. The second value is set to 0.1 (10%).

[0068] The construction module establishes a first mapping relationship between the retained defect combinations and the corresponding defect combination probabilities, inputs any first defect into the first mapping relationship, obtains the corresponding second defects and the corresponding defect combination probabilities, or inputs any second defect into the first mapping relationship, obtains the corresponding first defects and the corresponding defect combination probabilities.

[0069] In specific implementation, by establishing the first mapping relationship between defect combinations and defect combination probabilities, the system can quickly and accurately identify and predict high-probability defect combinations, provide a scientific basis for maintenance decisions based on historical data and probability analysis, and reduce the uncertainty of human judgment.

[0070] The present invention ensures image quality and the accuracy of defect detection by automatically adjusting shooting parameters, matches the light source angle in combination with the spindle material model, improves the pertinence and reliability of detection, detects spindle defects from multiple dimensions in combination with image analysis and operation data analysis, uses standard defect images as a reference, improves the accuracy of defect detection, predicts potential defects through historical data and probability analysis, improves the system's prediction ability and reliability, establishes a mapping relationship between defect combinations and probabilities, and provides data support for subsequent maintenance and decision-making. In combination with image analysis and operation data analysis, it can detect spindle defects more comprehensively, avoid the limitations of a single detection method, and quickly screen out high-probability defect combinations through probability analysis, thereby improving the speed of defect diagnosis.

[0071] It will be understood by those skilled in the art that embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Wherein the storage medium is implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device that implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that modifications or equivalent replacements of the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A fan main shaft defect detection system, characterized in that: It includes adjustment module, analysis module and construction module; The adjustment module obtains the spindle material model, matches the corresponding red light source lamp bead angle according to the spindle material model, sends a start shooting signal, obtains a finished product image, sets the pixels belonging to the spindle in the finished product image as valid pixels, calculates the valid pixel ratio, and adjusts the shooting information according to the valid pixel ratio to obtain a first image. After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends a first power to the ventilator to collect the operation data of the spindle; The analysis module preprocesses the first image to obtain a standard defect image, performs a first analysis on the preprocessed first image and the standard defect image, obtains a first main shaft defect based on the first analysis result, and performs a second analysis on the operating data to obtain a second main shaft defect; The construction module associates the first spindle defect with the second spindle defect to obtain a defect combination, retrieves historical defect data to obtain each defect combination, calculates the defect combination probability, sets a probability threshold, filters the same defect combination probability according to the probability threshold, and establishes a first mapping relationship between the defect combination that meets the conditions and the defect combination probability.

2. A fan main shaft defect detection system according to claim 1, characterized in that: The adjustment module obtains the spindle material model, calls the red light source database, inputs the spindle material model into the red light source database, matches the lamp bead angle of the red light source corresponding to the spindle material model, and the red light source adjusts the lamp bead angle according to the lamp bead angle of the red light source corresponding to the spindle material model. When the adjustment is completed, the red light source sends an adjustment completion signal to the adjustment module.

3. A fan main shaft defect detection system according to claim 2, characterized in that: After receiving the adjustment completion signal, the adjustment module sends a start shooting signal to the industrial camera; The industrial camera takes pictures of the spindle to obtain the finished product picture; The adjustment module sets the pixels belonging to the main axis in the finished image as valid pixels and counts the number of valid pixels and the total number of pixels; Calculate the ratio of the number of effective pixels to the total number of pixels and set it as the effective pixel ratio; The adjustment module adjusts the shooting information according to the effective pixel ratio to obtain a first image; The shooting information includes a shooting position and a magnification, wherein the shooting position is represented as the three-dimensional coordinates of the center of the set of industrial cameras; After the shooting is completed, the adjustment module sends a power-on start signal to the power supply device, and the power supply device sends the first power to the ventilator; The operating data of the main shaft is collected, wherein the operating data includes vibration data and sound intensity data.

4. A fan main shaft defect detection system according to claim 1, characterized in that: The analysis module preprocesses the first image, wherein the preprocessing includes filtering and grayscale processing; Acquire standard defect images, which include standard scratch images, standard rust images, standard crack images, standard wear images and standard excessive bending images, and the corresponding first spindle defect categories are scratches, rust, cracks, wear and excessive bending.

5. A fan main shaft defect detection system according to claim 4, characterized in that: The analysis module performs a first analysis on the preprocessed first image and the standard defect image. The analysis logic of the first analysis includes: Extracting shape features of each standard defect image, recorded as first features; extracting shape features of the preprocessed first image, recorded as second features; calculating similarities between the second features and each first feature, recorded as first similarities; setting the first value as a similarity threshold; comparing the first similarity with the first value; retaining the corresponding standard defect image when the first similarity is greater than or equal to the first value; otherwise, deleting the corresponding standard defect image; The principal axis defect corresponding to the retained standard defect image is set as the first principal axis defect.

6. The fan main shaft defect detection system according to claim 1, characterized in that: Obtain standard dynamic defects, which include improper installation of a standard main shaft and damage to a standard coupling. The corresponding second main shaft defects include improper installation of the main shaft and damage to the coupling. Obtain historical vibration data and historical sound intensity data corresponding to historical standard dynamic defects.

7. A fan main shaft defect detection system according to claim 5, characterized in that: The analysis module performs a second analysis on the vibration data and the sound intensity data, and the analysis logic of the second analysis includes: Perform high-frequency noise removal and normalization processing on historical vibration data and vibration data, and perform outlier removal and logarithm conversion processing on sound intensity data and historical sound intensity data; Calculating the variance value of the vibration data after high-frequency noise removal and normalization and the variance value of the historical vibration data, calculating the intensity change rate of the sound intensity data after outlier removal and logarithm transformation and the intensity change rate of the historical sound intensity data, setting the variance value of the historical vibration data and the intensity change rate of the historical sound intensity data as a vibration feature vector and an intensity feature vector, fusing the vibration feature vector and the intensity feature to obtain a fused feature vector, inputting the fused feature vector and the corresponding standard dynamic defect into a machine learning model for training to obtain a dynamic defect model; A fused eigenvector corresponding to the variance value of the vibration data and the intensity change rate of the sound intensity data is obtained, recorded as a third eigenvector, and the third eigenvector is input into the dynamic defect model to obtain the corresponding second main shaft defect.

8. The fan main shaft defect detection system according to claim 1, characterized in that: The calculation logic of the defect combination probability includes: The total number of defect combinations is counted, the number of each identical defect combination is counted, and the ratio of the number of each identical defect combination to the total number of defect combinations is calculated, which is recorded as the defect combination probability.

9. The fan main shaft defect detection system according to claim 1, characterized in that: The construction module sets the second value as the probability threshold, compares each defect combination probability with the second value, and when the defect combination probability is greater than or equal to the second value, retains the corresponding defect combination probability and the corresponding defect combination; otherwise, deletes the corresponding defect combination probability and the corresponding defect combination.

10. A fan main shaft defect detection system according to claim 9, characterized in that: The construction module establishes a first mapping relationship between the retained defect combinations and the corresponding defect combination probabilities, inputs any first defect into the first mapping relationship, obtains the corresponding second defects and the corresponding defect combination probabilities, or inputs any second defect into the first mapping relationship, obtains the corresponding first defects and the corresponding defect combination probabilities.

Images