An online detection device and method for product defects in the manufacture of closed compressors

Abstract

The present invention discloses an online detection device and method for product defects in hermetic compressor manufacturing. The online detection device uses a multi-channel time-frequency-spatial feature fusion deep learning algorithm to fuse the time-frequency features of the vibration signals of the hermetic compressor's entire housing in three directions. By learning and extracting time-frequency and spatial features, it solves the problem of identifying and classifying hermetic compressor manufacturing defects, enabling real-time feedback of whole-machine defects to the front-end component processing and assembly links in intelligent compressor manufacturing, and establishing an intelligent closed-loop for hermetic compressor manufacturing. Furthermore, the online detection device is highly intelligent and can automatically complete the entire defect detection process at the production rhythm of the compressor manufacturing assembly line. It also collects and provides feedback on data such as the type and proportion of product defects, the number of defective products, and the product defect rate during the manufacturing process, significantly improving the efficiency of hermetic compressor production line manufacturing.

Description

Technical Field

[0001] The present invention relates to the technical fields of compressor manufacturing quality inspection, vibration signal measurement, digital signal processing, and online product defect detection. In particular, it relates to a device and method for intelligently detecting product defects on a production line during the manufacturing process of a closed compressor. The device can identify defective compressors in real time and automatically classify them according to the defect type. Background Art

[0002] In recent years, with the continuous development of society and the improvement of people's living standards, the use of refrigeration equipment has become increasingly widespread, and the variety of refrigeration equipment has become more diverse. Examples include refrigerators, water dispensers, and tea and wine storage cabinets for everyday household use, various types of ice freezers and ice makers in shopping malls, and medical refrigerators for preserving vaccines and blood products in medical facilities. However, as the operating environments of refrigeration equipment become increasingly complex and energy efficiency requirements continue to rise, stricter quality requirements are being placed on the core component of refrigeration equipment: the hermetic refrigeration compressor. Currently, China has become a major global manufacturer of refrigeration compressors, accounting for over 70% of global production of hermetic refrigeration compressors alone, and the refrigeration compressor manufacturing industry is widely distributed.

[0003] The hermetic refrigeration compressor is the most complex and difficult-to-manufacture heart component in refrigeration systems and equipment. It has numerous internal parts, and during the manufacturing process on the production line, it is easy for defects to occur after the compressor is completed due to problems such as unqualified processing precision of parts and improper assembly. The shell of the hermetic refrigeration compressor is a steel plate with a thickness of about 2mm, and all parts are sealed inside the shell. In addition, the above-mentioned defects are often not obvious, and it is very difficult to rely on the external features of the shell to identify product defects. As a result, online detection of defects in the compressor has become a technology that needs to be broken through. At present, online detection of hermetic refrigeration compressors mainly relies on manual sensory experience such as listening and touching. It can only detect defective products with particularly obvious vibrations, and it is even more impossible to judge the type of product defects. The defect type can only be judged by post-dissection.

[0004] In summary, the method of identifying and classifying defects in the complete closed refrigeration compressor product, especially the online detection technology and equipment for product defects suitable for compressor automated production line manufacturing, has become a technical problem that needs to be urgently solved in the refrigeration compressor manufacturing industry. Summary of the Invention

[0005] In order to solve the technical bottleneck of automatic detection of defects of complete machine products during the manufacturing process of a closed refrigeration compressor production line, the present invention provides an online detection device and method for defects of closed compressor products.

[0006] The technical solutions of the present invention are as follows:

[0007] An online detection device for product defects in the manufacture of closed compressors comprises a mechanical system, wherein the mechanical system comprises a compressor to be inspected, a press base plate, a press conveyor, a press lifting cylinder, a main detection bracket, a code scanner and an auxiliary bracket; the press base plate is arranged on the press conveyor, and a press lifting cylinder is arranged below the press base plate; the main detection bracket is arranged on one side of the press conveyor and is located at the inspection point of the compressor to be inspected; a sensor telescopic chain is installed on the top of the main detection bracket, a three-axis acceleration sensor is installed below the sensor telescopic chain, and the three-axis acceleration sensor is connected to the sensor electromagnetic seat, a detection point proximity switch is installed in the middle of the main detection bracket, and a press power connector is provided below the main detection bracket; the auxiliary bracket is located in front of the detection point, and the code scanner is installed on the auxiliary bracket, and the code scanner is used to scan the compressor information QR code on the outside of the casing of the compressor to be inspected.

[0008] Furthermore, an online detection device for product defects in the manufacture of closed compressors further includes a measurement and control system, the measurement and control system including an industrial control computer, a display, an electrical cabinet, an Ethernet bus, and a data acquisition and controller; the industrial control computer uses a barcode scanner as an input device, and the industrial control computer uses a display as an output device; the data acquisition and controller implements data acquisition and real-time control of the online detection device, and exchanges data with the industrial control computer via the Ethernet bus; a vibration signal acquisition module, a pulse signal output module, a digital output module, and a digital input module are installed on the chassis of the data acquisition and controller;

[0009] The vibration signal acquisition module has four high-speed vibration signal acquisition channels, three of which collect vibration signals in the I direction, J direction, and K direction respectively. The vibration signals in the above three directions are measured by a triaxial acceleration sensor;

[0010] The pulse signal output module can output a 0-10V high-frequency response voltage signal, providing a pulse signal for the variable frequency starter of the variable frequency compressor to achieve speed adjustment of the variable frequency compressor;

[0011] The digital output module controls the external actuator through the intermediate relay, which can achieve effective isolation.

[0012] The digital output module controls the start and stop of the compressor under test, the start and stop of the press conveyor, and the switch of the main power supply of the online detection device through intermediate relays and contactors;

[0013] The digital output module controls the operation of the press lifting cylinder, press power connector, sensor electromagnetic seat, and defect alarm indicator through the intermediate relay;

[0014] The digital input module receives signals from the detection point proximity switch and the equipment abnormality alarm, and transmits the input signals to the data acquisition and controller.

[0015] Furthermore, the compressor power connector can automatically extend and retract. During testing, the compressor power connector extends and touches the triangular power socket of the compressor being tested, thereby providing power to the compressor being tested.

[0016] A method for detecting product defects online in the manufacture of a hermetic compressor comprises the following steps:

[0017] 1) When the test starts, the compressors manufactured on the assembly line are conveyed to the press conveyor. The online test device starts the press conveyor and transports the compressors to the test point.

[0018] 2) When the proximity switch at the detection point detects the compressor under inspection, the compressor conveyor stops, allowing the compressor under inspection to remain at the detection point. The scanner scans the compressor information QR code on the compressor housing and automatically records various production information of the compressor under inspection;

[0019] 3) The compressor lifting cylinder lifts the bottom plate of the compressor, the sensor electromagnetic seat is energized, the three-axis acceleration sensor is tightly connected to the shell of the compressor under test, the compressor power connector is extended and connected to the power triangle seat of the compressor under test, and the compressor under test is powered on and runs;

[0020] 4) The vibration signals of the compressor housing under test in three directions are sampled by a triaxial acceleration sensor. When the sampling time reaches the set time, the sampling of vibration data is stopped, the sensor electromagnetic seat is powered off, the compressor power connector is retracted, and the compressor lifting cylinder is lowered and returned to its original position;

[0021] 5) Analyze vibration signals in three directions using an embedded deep learning algorithm to determine whether the inspected compressor has defects. If so, the system classifies the defects accordingly, lights up the defect alarm indicator, and transports the defective product to the abnormal product area via the press conveyor. If not, the product is transported to the next process via the press conveyor.

[0022] 6) The test results are displayed and counted in real time, and the proportion and number of various defect types are counted and displayed in a pie-shaped distribution chart. At the same time, the production volume of the assembly line, the number of qualified products, the number of defective products, and the product defect rate are counted.

[0023] Furthermore, the specific process of step 5) defect judgment and classification is as follows:

[0024] 5.1) Vibration signal acquisition: A triaxial acceleration sensor samples vibration signals in three directions of the compressor housing under inspection, and then automatically intercepts 5-10 working cycles from the sampled data as raw analysis data;

[0025] 5.2) Signal Denoising and Enhancement: The original analysis data is parsed into spectral signals of different frequency bands using modal decomposition methods, and the spectral signals of production line interference noise are removed;

[0026] 5.3) Signal reconstruction image: Reconstruct the remaining spectrum signal after removing the interference noise of the production line, and then convert the reconstructed vibration signal into an image signal to facilitate subsequent feature extraction by deep convolutional neural network;

[0027] 5.4) Multi-layer convolution feature extraction: The reconstructed images in the three directions are subjected to multi-layer convolution and pooling according to their respective channels to extract defect features. Each convolution and pooling network structure consists of a convolution layer, a batch normalization layer, and a pooling layer. Each convolution and pooling network structure is connected in series to achieve feature learning and extraction of the reconstructed images in the three directions;

[0028] 5.5) Defect Feature Fusion: After extracting features from the reconstructed images in the three directions through convolutional pooling, the defect features are first expanded according to their respective channels using a flattening layer. The defect features from the three channels are then fused and spliced ​​through a fully connected layer to achieve the fusion of defect feature information from the three directions.

[0029] 5.6) Defect classification output: Use the classification layer to decide the category of product defects and finally output the classification results.

[0030] Furthermore, in step 5), the defect identification and classification adopts a multi-channel time-frequency fusion deep learning algorithm to fuse the time-frequency characteristics of the vibration signals in three directions of the closed compressor, automatically learn the time-frequency characteristics and spatial characteristics of the defects, and effectively improve the detection accuracy of the online detection device.

[0031] The present invention can identify and classify defective products manufactured on the automated production line of hermetic compressors, solve a bottleneck in the intelligent manufacturing process of hermetic compressors, realize real-time feedback of whole machine defects to the front-end parts processing and assembly links in intelligent manufacturing of compressors, and establish an intelligent closed loop for the manufacture of hermetic compressors. The online detection device is composed of a product to be detected, a mechanical system, and a measurement and control system. The product to be detected refers to a hermetic compressor that has been completed on the production line and includes the compressor to be detected, the compressor waiting to be detected, and the compressor that has been detected. The measurement and control system is mainly composed of an industrial control computer, a display, a data acquisition and controller, a three-axis acceleration sensor, etc. The three-axis acceleration sensor can collect vibration signals in three directions of the outer shell of the compressor. The mechanical system is the main part of the online detection device, which mainly includes the bottom plate of the press to be detected, the press conveyor, the press lifting cylinder, the main detection bracket, the auxiliary bracket, etc. The auxiliary bracket is equipped with a barcode scanner, which can scan the press information QR code on the outside of the shell of the compressor to be detected, so as to realize the automatic acquisition of the product information of the compressor to be detected.

[0032] The measurement and control system of the present invention uses an industrial control computer as its data processing core to identify and classify defects in products manufactured on a closed-circuit compressor production line. The industrial control computer uses a display as an output device to implement application software operations, including system parameter settings, defective product identification and classification results, and production data statistics. The industrial control computer uses a barcode scanner as an input device to automatically collect various product information about the inspected compressors. A data acquisition and controller (DAC) collects data and performs real-time control of the online detection device, exchanging data with the DAC via an Ethernet bus. The DAC chassis is equipped with a vibration signal acquisition module, a pulse signal output module, a digital output module, and a digital input module. The vibration signal acquisition module has four high-speed vibration signal acquisition channels, three of which collect vibration signals from three directions of the compressor housing. The vibration signals are measured by a triaxial accelerometer. The pulse signal output module outputs a 0-10V high-frequency voltage signal, providing a pulse signal for the variable-frequency starter of the variable-frequency compressor, enabling speed adjustment of the variable-frequency compressor. The digital output module controls external actuators via intermediate relays and contactors. The digital input module receives signals from the detection point proximity switch and the equipment abnormality alarm.

[0033] The design and development of the application software of the present invention is written in a graphical editing language, and the generated program is in block diagram format, suitable for the rapid development of Windows-based applications. The online detection device application software is no longer limited to simple data acquisition and equipment control, but has added functions such as system parameter setting, product information scanning and input, model self-learning, product defect detection, data management, and interface display. The application software data management adopts a relational database management system.

[0034] By adopting the above technology, compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1) This detection device can realize online detection of defects in the complete sealed compressor product during the production line manufacturing process. It can not only determine whether the product has defects, but also classify its defect type. At the same time, the defects of the complete machine can be fed back to the front-end component processing and assembly links in real time, establishing an intelligent closed loop for the manufacture of closed sealed compressors.

[0036] 2) The defect recognition and classification method of the detection device adopts a multi-channel time-frequency fusion deep learning algorithm to fuse the time-frequency characteristics of the vibration signals of the closed compressor in three directions, automatically learn the time-frequency and spatial characteristics of the defects, and effectively improve the detection accuracy of the online detection device.

[0037] 3) The detection device is highly intelligent and can automatically complete the entire defect detection process under the production rhythm of the compressor manufacturing line. It can also collect statistics and feedback on data such as the type and proportion of product defects, the number of defective products, the number of qualified products, the total production volume, and the product defect rate during the manufacturing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the external structure of the device of the present invention;

[0039] Figure 2 It is a hardware framework diagram of the control system of the device of the present invention;

[0040] Figure 3 It is a functional framework diagram of the application software of the device of the present invention;

[0041] Figure 4 It is a framework diagram of the defect identification and classification method of the device of the present invention;

[0042] Figure 5 It is a test flow chart of the device of the present invention;

[0043] In the figure: 1-industrial control computer, 2-display, 3-electrical cabinet, 4-detection main bracket, 5-sensor telescopic chain, 6-three-axis acceleration sensor, 7-sensor electromagnetic base, 8-inspected compressor, 9-pressor information QR code, 10-pressor base plate, 11-code scanner, 12-auxiliary bracket, 13-pressor conveyor, 14-pressor lifting cylinder, 15-pressor power connector, 16-detection point proximity switch, 17-Ethernet bus, 18-data acquisition and controller, 19-vibration signal acquisition module, 20-pulse signal output module, 21-digital output module, 22-digital input module, 23-frequency drive, 24-contactor, 25-intermediate relay, 26-main power supply of online detection device, 27-defect alarm indicator, 28-equipment abnormality alarm. DETAILED DESCRIPTION

[0044] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited thereto:

[0045] The external structure of the online detection device of this embodiment is as follows Figure 1 As shown, it can identify and classify defective products manufactured on the automated production line of hermetic compressors, solve a bottleneck in the intelligent manufacturing process of hermetic compressors, realize real-time feedback of whole machine defects to the front-end parts processing and assembly links in intelligent manufacturing of compressors, and establish an intelligent closed loop for hermetic compressor manufacturing.

[0046] The online detection device consists of a product to be detected, a mechanical system and a measurement and control system. The product to be detected refers to the enclosed compressor that has been completed on the production line, including the compressor to be detected 8, as well as compressors waiting for detection on other production lines and compressors that have been detected. The measurement and control system is installed in the electrical cabinet 3, and mainly includes an industrial control computer 1, a display 2, and components related to data acquisition and control. The mechanical system is the main part of the online detection device, mainly including the bottom plate 10 of the press to be detected, the press conveyor 13, the press lifting cylinder 14, the main detection bracket 4, the auxiliary bracket 12 and other parts. The compressor to be detected 8 is installed on the top of the press bottom plate 10, and the press bottom plate 10 can be conveyed in a flow-through manner on the press conveyor 13. The main inspection bracket 4 is located at the position where the inspected compressor 8 is to be inspected. A compressor power connector 15 is installed below it. The compressor power connector 15 can automatically extend and retract. During inspection, the compressor power connector 15 extends to contact the triangular power socket of the inspected compressor 8, providing power to the inspected compressor 8. When the inspection is completed, the compressor power connector 15 retracts to disconnect the power supply to the inspected compressor 8. A detection point proximity switch 16 is installed in the middle of the main inspection bracket 4. The detection point proximity switch 16 uses an infrared sensing proximity switch. When it detects that the inspected compressor 8 has reached the inspection point, it transmits a signal to the online detection device, which stops the movement of the compressor conveyor 13. A sensor telescopic chain 5 is installed on the top of the main detection bracket 4, and a three-axis acceleration sensor 6 is installed below the sensor telescopic chain 5. The three-axis acceleration sensor 6 is connected to the sensor electromagnetic seat 7. When the compressor 8 under test begins to be tested, the compressor lifting cylinder 14 lifts the compressor base 10 and the compressor 8 under test to a certain position. After the sensor electromagnetic seat 7 is powered on, the three-axis acceleration sensor 6 is tightly connected to the shell of the compressor 8 under test through the sensor electromagnetic seat 7. The auxiliary bracket 12 is located in front of the detection point, and the code scanner 11 is installed on the auxiliary bracket 12. When the compressor 8 under test approaches or reaches the detection point, the code scanner 11 scans the compressor information QR code 9 on the outside of the shell of the compressor 8 under test, realizing automatic acquisition of the product information of the compressor under test.

[0047] The hardware framework of the measurement and control system of the online detection device of this embodiment is shown in the figure below. Figure 2As shown, the measurement and control system uses an industrial control computer 1 as its data processing core to identify and classify defects in products manufactured on the hermetic compressor production line. The industrial control computer 1 uses a display 2 as an output device to implement application software operations, including system parameter settings, defective product identification and classification results, and production data statistics. The industrial control computer 1 uses a barcode scanner 11 as an input device to automatically collect various product information about the inspected compressor 8. A data acquisition and controller 18 performs data acquisition and real-time control of the online detection device, exchanging data with the industrial control computer 1 via an Ethernet bus 17. The chassis of the data acquisition and controller 18 is equipped with a vibration signal acquisition module 19, a pulse signal output module 20, a digital output module 21, and a digital input module 22. The vibration signal acquisition module 19 has four high-speed vibration signal acquisition channels, three of which collect vibration signals in the I, J, and K directions, respectively. The vibration signals in these three directions are measured by a triaxial accelerometer 6. The pulse signal output module 20 can output a 0-10V high-frequency response voltage signal, providing a pulse signal for the variable frequency starter 23 of the variable frequency compressor to adjust the speed of the variable frequency compressor. The digital output module 21 controls the external actuator through the intermediate relay 25, which can effectively isolate the compressor. The digital output module 21 controls the start and stop of the inspected compressor 8, the start and stop of the press conveyor 13, and the switching of the main power supply 26 of the online detection device through the intermediate relay 25 and the contactor 24. The digital output module 21 controls the operation of the press lifting cylinder 14, the press power connector 15, the sensor electromagnetic seat 7, and the defect alarm indicator 27 through the intermediate relay 25. The digital input module 22 receives signals from the detection point proximity switch 16 and the equipment abnormality alarm 28, and transmits the input signals to the data acquisition and controller 18.

[0048] The design and development of the application software of the online detection device of this embodiment is written in a graphical editing language. The generated program is in the form of a block diagram, which is suitable for rapid development of window-based applications. The functional block diagram of the application software is as follows: Figure 3 As shown in the figure, this measurement and control application software is no longer limited to simple data acquisition and equipment control. It also adds functions such as system parameter setting, product information scanning and input, model self-learning, product defect detection, data management, and interface display. The application software uses MySQL, a relational database management system from MySQL AB, which is characterized by its open source nature.

[0049] The functions of the online detection device application software of the present invention include system setting a, information scanning b, model self-learning c, defect detection d, data management e and interface display f. The details of each functional module are as follows:

[0050] (1) System settings a mainly complete user authority settings a1 and user password management a2, as well as system parameters a3, etc.; user authority is divided into production administrators and ordinary operators. Production administrators have the highest authority and can use all functions of the application software, including user authority settings a1, system parameter settings a3, etc.; ordinary operators can only use part of the functions of the application software to complete the entire online detection process, and cannot use advanced functions such as user authority settings a1, system parameter settings a3, etc.

[0051] (2) Information scanning b mainly completes the information input of the two-dimensional information code of the compressor 8 shell on the production line, including the compressor number, compressor steel grade, production batch, production line number, inspector information, etc.

[0052] (3) Model self-learning c mainly completes the self-learning and model retraining of the deep learning algorithm embedded in the online detection device, so that the embedded deep learning algorithm model can learn new defect types and new compressor product types, thereby improving the accuracy of the online detection device in identifying and classifying defective products, and improving the robustness and universality of the deep learning algorithm embedded in the online detection device;

[0053] (4) Defect detection d is divided into manual detection d1 and automatic detection d2. In the manual detection d1 state, the start, stop and speed of the inspected compressor 8 can be manually controlled, and the shell vibration signal is measured, and then the defect identification and classification are performed; in the automatic detection d2 state, the detection system will perform fully automatic online detection according to the set system parameters, and go through the two processes of defect identification d21 and defect classification d22 until the entire detection process is completed;

[0054] (5) Data management (e) is the management of online detection data, which includes data processing (e1), data storage (e2), and data query (e3). Recording and saving product data and operating status during online detection is an important task of the application software. By analyzing and processing the shell vibration data, it can be determined whether the inspected compressor product has defects. If there are defects, the defects can be classified according to their vibration characteristics.

[0055] (6) Interface display f mainly realizes vibration spectrum display f1, defect statistics display f2 and production report display f3. Among them, vibration spectrum display f1 mainly displays the vibration data in three directions collected by the three-axis acceleration sensor 6, and the spectrum diagram after time-frequency transformation. Defect statistics display f2 mainly counts the number of compressors with various defect types detected, as well as the corresponding defect ratio, and represents them with a pie-shaped distribution chart. Production report display f3 mainly displays various production information of the compressor assembly line, such as total production volume, qualified product volume, defective product volume, product defect rate, etc.

[0056] The online defect detection method for hermetic compressor products in this embodiment, and its defect identification and classification method are as follows Figure 4 As shown, the specific steps include:

[0057] (1) Vibration signal acquisition: The triaxial acceleration sensor 6 samples the vibration signals of the compressor 8 shell in three directions, and then automatically intercepts 5 working cycles from the sampled data as the original analysis data;

[0058] (2) Signal denoising and enhancement: The original analysis data is parsed into 8 spectrum signals of different frequency bands through the modal decomposition method, and the spectrum signal of the production line interference noise is removed. Figure 4 The spectrum signal in the middle dotted box is interference noise;

[0059] (3) Signal reconstruction image: Reconstruct the remaining spectrum signal after removing the interference noise of the production line, and then convert the reconstructed vibration signal into an image signal to facilitate the subsequent deep convolutional neural network for feature extraction;

[0060] (4) Multi-layer convolution feature extraction: The reconstructed images in the three directions are subjected to multi-layer convolution pooling according to their respective channels to extract defect features. Each layer of convolution pooling network structure consists of a convolution layer, a batch normalization layer, and a pooling layer. The convolution pooling network structures of each layer are serially connected in sequence to achieve feature learning and extraction of the reconstructed images in the three directions;

[0061] (5) Defect feature fusion: After the reconstructed images in three directions are extracted through convolution pooling, the defect features are first expanded according to their respective channels by the flattening layer, and then the defect features of the three channels are fused and spliced ​​through the fully connected layer to achieve the fusion of the defect feature information in the three directions;

[0062] (6) Defect classification output: Use the classification layer to decide the category of product defects and finally output the classification results.

[0063] The on-line detection method for defects of hermetic compressor products in this embodiment has the following detection process: Figure 5 As shown, the specific steps include:

[0064] (1) When the test starts, the compressors manufactured on the assembly line are conveyed to the press conveyor 13. The online test device starts the press conveyor 13 and transports the compressor 8 to the test point.

[0065] (2) When the detection point proximity switch 16 detects the inspected compressor 8, the online detection device stops the press conveyor 13, causing the inspected compressor 8 to remain at the detection point. The scanner 11 scans the compressor information QR code 9 on the housing of the inspected compressor 8 and automatically records various production information of the inspected compressor 8;

[0066] (3) The press lifting cylinder 14 lifts the press bottom plate 10, and the sensor electromagnetic seat 7 is energized, so that the three-axis acceleration sensor 6 is tightly connected to the shell of the inspected compressor 8. The press power connector 15 extends and connects to the power triangle seat of the inspected compressor 8, and the inspected compressor 8 is powered on and runs;

[0067] (4) The online detection device samples the vibration signals of the shell of the inspected compressor 8 in three directions through the three-axis acceleration sensor 6. When the sampling time reaches the set time, the sampling of vibration data is stopped, the sensor electromagnetic seat 7 is powered off, the compressor power connector 15 is retracted, and the compressor lifting cylinder 14 is lowered and returned to its original position;

[0068] (5) The online detection device uses the embedded deep learning algorithm to analyze the vibration signals in three directions to determine whether the inspected compressor 8 has defects. If there are defects, the corresponding defect classification is performed, the defect alarm indicator 27 lights up, and the defective product is transported to the abnormal product area through the press conveyor 13. If there are no defects, the product is transported to the next process through the press conveyor 13;

[0069] (6) The online detection device displays and counts the detection results in real time, counts the proportion and number of various defect types, and displays them in a pie-shaped distribution chart. At the same time, it counts the production volume of the assembly line, the number of qualified products, the number of defective products, and the product defect rate.

Claims

1. An online detection device for product defects in the manufacture of hermetic compressors, characterized in that: The invention comprises a mechanical system, wherein the mechanical system comprises a compressor (8) to be inspected, a press bottom plate (10), a press conveyor (13), a press lifting cylinder (14), a main inspection bracket (4), a code scanner (11) and an auxiliary bracket (12); the press bottom plate (10) is arranged on the press conveyor (13), and a press lifting cylinder (14) is arranged below the press bottom plate (10); the main inspection bracket (4) is arranged on one side of the press conveyor (13) and is located at the inspection point of the compressor (8) to be inspected; a sensor is installed on the top of the main inspection bracket (4) A sensor telescopic chain (5) is provided below the sensor telescopic chain (5), and a three-axis acceleration sensor (6) is provided below the sensor electromagnetic seat (7). A detection point proximity switch (16) is provided in the middle of the detection main bracket (4), and a compressor power connector (15) is provided below the detection main bracket (4); the auxiliary bracket (12) is located in front of the detection point, and the code scanner (11) is installed on the auxiliary bracket (12). The code scanner (11) is used to scan the compressor information QR code (9) on the outside of the shell of the compressor (8) to be inspected; The device also includes a measurement and control system, which includes an industrial control computer (1), a display (2), an electrical cabinet (3), an Ethernet bus (17), and a data acquisition and controller (18); the industrial control computer (1) uses a barcode scanner (11) as an input device, and the industrial control computer (1) uses a display (2) as an output device; the data acquisition and controller (18) realizes data acquisition and real-time control of the online detection device, and exchanges data with the industrial control computer (1) through the Ethernet bus (17); a vibration signal acquisition module (19), a pulse signal output module (20), a digital output module (21), and a digital input module (22) are installed on the chassis of the data acquisition and controller (18); The vibration signal acquisition module (19) has four high-speed vibration signal acquisition channels, three of which respectively acquire vibration signals in the I direction, the J direction, and the K direction. The vibration signals in the three directions are measured and obtained by the triaxial acceleration sensor (6); The pulse signal output module (20) can output a 0-10V high-frequency response voltage signal to provide a pulse signal for the variable frequency drive (23) of the variable frequency compressor, thereby adjusting the speed of the variable frequency compressor; The digital output module (21) controls the external actuator via the intermediate relay (25), and the intermediate relay (25) can achieve effective isolation. The digital output module (21) controls the start and stop of the compressor (8) to be inspected, the start and stop of the press conveyor (13), and the switch of the main power supply (26) of the online inspection device through the intermediate relay (25) and the contactor (24); The digital output module (21) controls the operation of the press lifting cylinder (14), the press power connector (15), the sensor electromagnetic seat (7), and the defect alarm indicator (27) through the intermediate relay (25); The digital input module (22) receives signals from the detection point proximity switch (16) and the equipment abnormality alarm (28), and transmits the input signals to the data acquisition and controller (18).

2. The device for online detection of product defects in the manufacture of hermetic compressors according to claim 1, characterized in that: The compressor power connector (15) is automatically retractable. During testing, the compressor power connector (15) extends to contact the triangular power socket of the compressor (8) being tested, thereby providing power to the compressor (8) being tested.

3. The detection method of the on-line detection device for product defects in the manufacture of hermetic compressors according to any one of claims 1-2, characterized in that: The steps include: 1) When the test starts, the compressors manufactured on the assembly line are conveyed to the press conveyor (13), and the online test device starts the press conveyor (13) and transports the compressors (8) to the test point; 2) When the detection point proximity switch (16) detects the inspected compressor (8), the compressor conveyor (13) stops, causing the inspected compressor (8) to remain at the detection point, and the code scanner (11) scans the compressor information QR code (9) on the shell of the inspected compressor (8) to automatically record various production information of the inspected compressor (8); 3) The press lifting cylinder (14) lifts the press bottom plate (10), the sensor electromagnetic seat (7) is energized, so that the three-axis acceleration sensor (6) is tightly connected to the housing of the compressor (8) under test, the press power connector (15) extends and connects to the triangular power socket of the compressor (8) under test, and the compressor (8) under test is powered on and runs; 4) The vibration signals of the shell of the inspected compressor (8) in three directions are sampled by the triaxial acceleration sensor (6). When the sampling time reaches the set time, the sampling of the vibration data is stopped, the sensor electromagnetic seat (7) is powered off, the compressor power connector (15) is retracted, and the compressor lifting cylinder (14) is lowered and returned to its original position; 5) Analyzing the vibration signals in three directions using the embedded deep learning algorithm to realize defect recognition and classification of the inspected compressor (8), that is, determining whether the inspected compressor (8) has defects. If so, the corresponding defect classification is performed, the defect alarm indicator (27) lights up, and the defective product is transported to the abnormal product area through the press conveyor (13). If not, the defect is transported to the next process through the press conveyor (13); The specific process of defect identification and classification in step 5) is as follows: 5.1) Vibration signal acquisition: The triaxial acceleration sensor (6) samples the vibration signals of the shell of the compressor (8) under inspection in three directions, and then automatically intercepts 5-10 working cycles from the sampled data as the original analysis data; 5.2) Signal Denoising and Enhancement: The original analysis data is parsed into spectral signals of different frequency bands using modal decomposition methods, and the spectral signals of production line interference noise are removed; 5.3) Signal Reconstruction Image: Reconstruct the remaining spectrum signal after removing the production line interference noise, and then convert the reconstructed vibration signal into an image signal to facilitate subsequent feature extraction using a deep convolutional neural network; 5.4) Multi-layer convolution feature extraction: The reconstructed images in the three directions are subjected to multi-layer convolution and pooling according to their respective channels to extract defect features. Each convolution and pooling network structure consists of a convolution layer, a batch normalization layer, and a pooling layer. These convolution and pooling network structures are connected in series to achieve feature learning and extraction of the reconstructed images in the three directions. 5.5) Defect Feature Fusion: After extracting features from the reconstructed images in the three directions through convolutional pooling, the defect features are first expanded according to their respective channels using a flattening layer. The defect features from the three channels are then fused and spliced ​​through a fully connected layer to achieve the fusion of defect feature information from the three directions. 5.6) Defect classification output: Use the classification layer to determine the category of product defects and finally output the classification results; 6) The test results are displayed and counted in real time, and the proportion and number of various defect types are counted and displayed using a pie-shaped distribution chart. At the same time, the production volume of the assembly line, the number of qualified products, the number of defective products, and the product defect rate are counted.

4. The detection method of the on-line detection device for product defects in the manufacture of hermetic compressors according to claim 3, characterized in that: In step 5), the defect recognition and classification adopts a multi-channel time-frequency fusion deep learning algorithm to fuse the time-frequency features of the vibration signals in three directions of the closed compressor, and automatically learn the time-frequency and spatial features of the defects. Effectively improve the detection accuracy of online detection devices.

Images