A Speech Recognition-Based Management System and Method for Recording Defective Products in Home Appliance Production

Abstract

This invention discloses a speech recognition-based management system and method for recording defective products in home appliance production, relating to the field of home appliance production quality management technology. This system, combined with speech recognition technology, firstly improves the efficiency of recording defective home appliances. After determining an appliance to be defective, it performs closed-loop verification of the defective product's defective features using a multimodal evidence chain. It quantifies the completeness of the defective appliance's recorded speech by comprehensively and intuitively matching the defective appliance's voice. When the recorded defect is found to be incomplete, it uses dynamic parameter coupling technology and a dynamic review mechanism to eliminate ambiguity and guide quality inspectors to accurately complete the defective features. By forming a three-in-one defective product management system of "efficient speech recognition recording – virtual prototype simulation verification – actual operation data matching," it achieves a qualitative leap from "post-inspection" to "pre-prediction."

Description

Technical Field

[0001] This invention relates to the field of quality management technology in home appliance production, and more specifically, to a voice recognition-based management system and method for recording defective products in home appliance production. Background Technology

[0002] In the field of quality management in home appliance manufacturing, traditional defective product management systems generally face technical bottlenecks such as low data entry efficiency, crude verification mechanisms, and a lack of proactive quality control. These bottlenecks manifest themselves as follows:

[0003] Traditional methods rely on quality inspectors manually filling out paper records or typing electronic forms, resulting in low data entry efficiency and significant human error. Furthermore, quality inspectors typically test appliances visually or with simple instruments, lacking quantitative analysis and correlative evidence of defects, and prone to missing defects. Direct data entry can also lead to missing defects, impacting subsequent targeted diagnostic analysis of defective products. For more complex defects hidden within the product (such as loose internal solder joints), traditional testing methods often require disassembly, a cumbersome and time-consuming process.

[0004] Based on the above, this invention proposes a voice recognition-based management system and method for recording defective products in home appliance manufacturing. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a voice recognition-based management system and method for recording and managing defective products in home appliance manufacturing.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A voice recognition-based management system for recording and managing defective products in home appliance manufacturing, including

[0008] The quality inspection voice analysis unit analyzes and extracts keywords from the voices spoken by quality inspectors when they inspect defective home appliances.

[0009] The initial defect judgment unit marks the appliance as an initial defect judgment appliance when it extracts keywords contained in the defect keyword library from the voice, and generates an appliance log at the same time.

[0010] The virtual prototype operation unit converts the defect characteristics of home appliance logs into implantable parameters and inputs them into the virtual prototype model, collecting various intuitive feedback data from the virtual prototype model;

[0011] The comprehensive intuitive matching unit is used to determine the comprehensive intuitive matching degree of the initially judged home appliance;

[0012] The defective product entry analysis unit is used to determine whether a defective product has been entered into the inventory or to initiate the entry review process.

[0013] The data entry review execution unit selects review features after the data entry review step is started, checks the review features, and determines whether to enter the review features.

[0014] Furthermore, the method for determining the overall intuitive matching degree of the home appliance in the initial assessment is as follows: control the home appliance in the initial assessment to operate in actual operation, collect various intuitive feedback data of the home appliance during operation, further determine the intuitive matching degree of various intuitive feedback data, sum the intuitive matching degrees of various intuitive feedback data and take the average value to calculate the overall intuitive matching degree.

[0015] Furthermore, the process of determining whether to enter defective products into the database or to initiate a review process is as follows: If the overall visual matching degree of the initially determined appliance is higher than the threshold matching degree, the appliance initially determined to be a defective appliance is directly marked as a defective appliance, and the appliance log of the defective appliance is simultaneously entered into the defective product database.

[0016] Initially, it was determined that the overall intuitive matching degree of the home appliances was not higher than the threshold matching degree, so the data entry and review process was initiated.

[0017] Further, select the review features. The specific steps are as follows: run multiple defect addition simulations on the controlled virtual prototype model, obtain the comprehensive intuitive matching degree corresponding to each defect addition simulation, and when the comprehensive intuitive matching degree corresponding to a defect addition simulation is higher than the threshold matching degree, mark the new defect feature generated by the defect addition simulation as the review feature.

[0018] Furthermore, the re-inspection features are checked and it is determined whether to record them: quality inspectors are arranged to check the re-inspection features of the initially judged home appliances. Whenever a keyword contained in the defect keyword library is extracted from the quality inspector's voice, the defect feature corresponding to the keyword is recorded in the home appliance log of the initially judged home appliance.

[0019] Furthermore, after the data entry and review steps are completed, the overall intuitive matching degree of the initially judged home appliance is determined again. If the overall intuitive matching degree of the initially judged home appliance is higher than the threshold matching degree, the initially judged home appliance is directly marked as a defective home appliance, and the home appliance log of the defective home appliance is simultaneously entered into the defective product database.

[0020] Furthermore, since the initial assessment indicated that the overall visual matching degree of the home appliances was not higher than the threshold matching degree, the data entry and review process was restarted.

[0021] Furthermore, the steps of the voice recognition-based method for recording and managing defective products in home appliance manufacturing are as follows:

[0022] Step 1: Quality Inspection Voice Collection;

[0023] Step 2: Initial assessment of defective products;

[0024] Step 3: Defect matching analysis of defective products;

[0025] Step 4: Re-inspection of defects in defective products.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The system and method of this invention combine speech recognition technology to first improve the efficiency of recording defective home appliances. After determining that a home appliance is defective, it performs closed-loop verification of the defective features through a multimodal evidence chain. By comprehensively and intuitively matching the data, it quantifies whether the recorded defective home appliance speech is complete. When the recorded defective home appliance speech is found to be incomplete, it combines dynamic parameter coupling technology and a dynamic review mechanism to eliminate ambiguity and guide quality inspectors to accurately complete the defective features. By forming a three-in-one defective product management system of "efficient speech recognition recording - virtual prototype simulation verification - actual operation data matching", it achieves a qualitative change from "post-inspection" to "pre-prediction". Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the system of the present invention;

[0029] Figure 2 Flowchart for inputting features for review. Detailed Implementation

[0030] Example 1: Refer to Figures 1-2 The home appliance manufacturing defective product entry and management system based on voice recognition includes a quality inspection voice parsing unit, a defective product initial judgment unit, a virtual prototype operation unit, a comprehensive and intuitive matching unit, a defective product entry and analysis unit, and an entry review execution unit.

[0031] The quality inspection speech parsing unit collects the speech of quality inspectors in real time when they inspect defective home appliances. Simultaneously, it performs keyword parsing and extraction on the speech. Keyword parsing and extraction involves: first, preprocessing the speech using noise reduction techniques (e.g., adaptive filters like spectral subtraction to remove background noise from the production line, such as from stamping machines and conveyor belts); end-point detection (based on short-time energy and zero-crossing rate to detect speech start and end points and avoid invalid audio processing); and feature extraction (converting the audio signal into Mel-frequency cepstral coefficients (MFCC) or linear predictive cepstral coefficients (LPCC) features). The second step is speech recognition using deep learning models, employing an end-to-end Transformer architecture (e.g., Wav2Vec). 2.0 or Conformer), industry-customized optimization: fine-tuning using voice data from the home appliance quality inspection field (such as 3000 hours of defect description voice), introducing language models (such as LM based on BERT) to improve the accuracy of professional terminology recognition), the third step, selecting keywords for parsing and extraction algorithms, including keyword matching algorithms: exact matching: achieving fast keyword retrieval with O(n) complexity based on Trie trees; fuzzy matching: using Levenshtein distance to calculate approximate keywords; semantic understanding enhancement: using Word2Vec or GloVe pre-trained word vectors to calculate semantic similarity between words, for example: associating "gaps too large" with "gaps exceeding the standard" in the keyword library (semantic similarity 0.85) .

[0032] The initial defect judgment unit extracts keywords from the defect keyword library from the speech (the construction process of the defect keyword library: Step 1, Initial library construction: 1. Industry standard integration: GB / T 2828.1-2012 "Sampling Inspection Procedures for Counting"; Defect classification standards for the home appliance industry (such as "Safety of Household and Similar Electrical Appliances" GB). 4706); 2. Enterprise historical data mining: Extract high-frequency defect descriptions from 100,000+ quality inspection records in the past 3 years. Example: Extract keywords from "oil stains seeping from the bottom of the compressor": compressor, bottom, oil stains; 3. Expert knowledge collection: Invite 10+ senior quality inspectors to organize typical defect descriptions and construct a defect-cause-solution triplet (e.g., "shell deformation → injection temperature too high → reduce mold temperature by 5℃"); Step 2: Thesaurus hierarchical structure design, such as a three-level classification system, first-level classification: appearance defects, etc.; second-level classification: surface defects, etc.; third-level keywords: scratches, dents, bubbles, sand holes, etc.), mark the appliance as the initial judgment appliance, and generate the appliance log at the same time. The appliance log includes appliance ID (each appliance ID is unique), appliance batch (i.e., the production batch of the appliance), and defect characteristics (defect characteristics include defect location, defect type, defect size, etc.).

[0033] Building a virtual prototype model; Step 1: Collect basic data of the home appliance (Design data: Import product CAD drawings (.STEP / .IGES format), extract key dimensional parameters (such as refrigerator cabinet thickness, air conditioner heat exchanger pipe diameter). Process data: Obtain production parameters from the MES system (such as injection molding temperature curve, welding pressure) to simulate defect generation mechanisms (such as shrinkage rate of plastic parts). Material library access: Directly use the software's built-in material library (such as SimScale's "Home Appliance Material Database"), which includes parameters such as thermal conductivity and elastic modulus of common materials such as ABS plastic and aluminum alloy); Step 2: Template-based rapid modeling (including component drag-and-drop generation (such as in SimScale Home Appliance Edition, drag prefabricated modules such as "compressor", "evaporator", and "condenser" from the left component library to the work area to automatically generate the refrigeration system framework. Example: When modeling a refrigerator, call the "cabinet module" to automatically generate the insulation layer and inner liner structure, with parameters including insulation material thickness (50). Step 3: Multiphysics coupling and boundary condition setting (e.g., taking refrigerators as an example, the integrated thermal-fluid-solid coupling of refrigerators); Step 4: Intelligent solver optimization (automatic mesh generation: the software automatically generates hexahedral meshes according to the model complexity (e.g., the refrigerator body mesh size is 5mm, and the compressor internal fine mesh is 2mm), and repairs deformed elements through mesh quality inspection tools. Reduced-order model acceleration: for repetitive simulations (e.g., analysis of different batches of products), a reduced-order model (ROM) is generated, and the calculation speed is increased by more than 10 times); Step 5: Verification and optimization of virtual prototype model (experimental data import: laboratory test data (e.g., the measured value of refrigerator energy consumption is 1.2kWh / 24h) is imported into the software and compared with the simulation results (1.28kWh / 24h); parameter inversion correction: the model parameters are adjusted through optimization algorithms (e.g., increasing the thickness of the insulation layer of the body), so that the simulation error is reduced from 6.7% to 2.1%).

[0034] The virtual prototype operation unit converts the defect features from the appliance logs into implantable parameters and inputs them into the virtual prototype model (e.g., if there is a 0.1mm micro-leak at the coordinates (X=120mm, Y=80mm, Z=50mm) of the air conditioner heat exchanger pipe, this feature is converted into {

[0035] "leakage_type": "orifice",

[0036] "diameter": 0.1e-3, / / Aperture (m)

[0037] "position": [0.12, 0.08, 0.05], / / Coordinates (m)

[0038] "flow_coefficient": 0.8,

[0039] "refrigerant": "R410A"

[0040] The converted parameters are input into the virtual prototype model. Then, the virtual prototype model is controlled to run in simulation. During the operation, various intuitive feedback data are collected from the virtual prototype model (intuitive feedback data are data directly reflected by the virtual prototype model or home appliance during operation, including but not limited to thermal field data (such as the temperature gradient of the refrigerator compartment and the temperature of the air conditioner compressor shell), flow field data (such as the flow velocity and pressure distribution of fluids such as refrigerant and air), structural field data (stress and strain distribution, deformation and displacement), and electromagnetic field data (electromagnetic radiation intensity, current and voltage distribution). The specific data included are limited according to the type of home appliance, which will not be elaborated in this application).

[0041] The integrated intuitive matching unit synchronously controls the initially determined home appliance to operate in actual operation. During operation, various intuitive feedback data of the home appliance are collected to further determine the intuitive matching degree of various intuitive feedback data (the intuitive matching degree is the value obtained by calculating the cosine similarity between two intuitive feedback data of the same type; taking the temperature gradient of the refrigerator compartment as an example, the cosine similarity calculation is performed between the temperature gradient of the refrigerator compartment in the virtual prototype model and the temperature gradient of the refrigerator compartment during actual operation; specifically, the temperature change rate is obtained from five positions: the upper left corner, upper right corner, lower left corner, lower right corner, and center of the refrigerator compartment, as shown in the corresponding vector of the virtual prototype model). The actual vector corresponding to the refrigerator ,vector with vector All include the temperature change rate at 5 locations in the refrigerator compartment, vector with vector The formula for calculating the dot product is: Where n is the number of temperature measurement locations in the vector. and They are vectors with vector The element at the i-th position (i.e., the normalized rate of temperature change at the corresponding position), for example: , Then the dot product is Calculate the vectors respectively and The modulus (also called norm) is calculated using the following formula: , For example, vectors The modulus is ,vector Similarly, the modulus of cosine similarity can be obtained. According to the definition of cosine similarity, its calculation formula is: This allows for the determination of the overall intuitive matching degree of the initially judged home appliances.

[0042] The initial method for determining the overall intuitive matching degree of home appliances is as follows: sum the intuitive matching degrees of various intuitive feedback data and take the average value to calculate the overall intuitive matching degree.

[0043] In the defective product entry and analysis unit, if the overall intuitive matching degree of the initially judged home appliance is higher than the threshold matching degree (the threshold matching degree is a pre-set threshold), the initially judged home appliance is directly marked as a defective home appliance, and the home appliance log of the defective home appliance is simultaneously entered into the defective product database (the defective product database is an information database that stores the home appliance logs of all defective home appliances).

[0044] Initially, it was determined that the overall intuitive matching degree of the home appliances was not higher than the threshold matching degree, so the data entry and review process was initiated.

[0045] The input review execution unit, upon initiating the input review step, controls the virtual prototype model (converted from the input defect features into implantable parameters) to perform multiple defect addition simulation runs. Before each simulation run, new defect features (different from existing defect features in the appliance) are randomly generated and converted into implantable parameters, input into the virtual prototype model, and the virtual prototype model is then controlled to perform one simulation run; this simulation run is referred to as a defect addition simulation run. The unit obtains the comprehensive intuitive matching degree corresponding to each defect addition simulation run. After each simulation run, various intuitive feedback data from the virtual prototype model during the defect addition simulation run are collected to further determine the intuitive matching degree of these data, thereby determining the overall intuitive matching degree of each defect addition simulation run. The overall intuitive matching degree corresponding to the proposed run is calculated. When the overall intuitive matching degree corresponding to a defect addition simulation run is higher than the threshold matching degree, the new defect feature generated by the defect addition simulation run is marked as a re-examination feature (no processing is required if it is not higher). Quality inspectors are arranged to check the re-examination features of the initially judged home appliances (i.e., check whether the initially judged home appliances have re-examination features). Whenever a keyword contained in the defect keyword library is extracted from the quality inspector's voice, the defect feature corresponding to the keyword is recorded in the home appliance log of the initially judged home appliance (i.e., the home appliance log of the initially judged home appliance is updated). When the quality inspector finds that the initially judged home appliance has a re-examination feature, the re-examination feature can be extracted by speaking the keyword and recorded in the home appliance log.

[0046] After the data entry and review process is completed (i.e., after the quality inspector has checked all the review features), the overall visual matching degree of the initially judged appliance is determined again. If the overall visual matching degree of the initially judged appliance is higher than the threshold matching degree (the threshold matching degree is a pre-set threshold), the initially judged appliance is directly marked as a defective appliance, and the appliance log of the defective appliance is simultaneously entered into the defective appliance database (the defective appliance database is an information database that stores the appliance logs of all defective appliances).

[0047] If the overall intuitive matching degree of the home appliances is determined to be no higher than the threshold matching degree, the data entry and review process is initiated (and this process is repeated).

[0048] The aforementioned system, combined with speech recognition technology, firstly improves the efficiency of recording defective home appliances. After determining that a home appliance is defective, it performs closed-loop verification of the defective features through a multimodal evidence chain. By comprehensively and intuitively matching the data, it quantifies whether the recorded defective home appliance speech is complete. When the recorded defective home appliance speech is found to be incomplete, it uses dynamic parameter coupling technology and a dynamic review mechanism to eliminate ambiguity and guide quality inspectors to accurately complete the defect features.

[0049] Example 2: A method for recording and managing defective products in home appliance manufacturing based on voice recognition, the steps of which are as follows:

[0050] Step 1: Quality Inspection Voice Collection;

[0051] Step 2: Initial assessment of defective products;

[0052] Step 3: Defect matching analysis of defective products;

[0053] Step 4: Re-inspection of defects in defective products.

[0054] The above method achieves a qualitative leap from "post-event inspection" to "pre-event prediction" by forming a three-in-one defective product management system of "efficient voice recognition input - virtual prototype simulation verification - actual operation data matching".

[0055] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0056] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0057] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0058] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0059] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0060] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0061] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 this application. 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.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A voice recognition-based management system for recording and managing defective products in home appliance manufacturing, characterized in that: include The quality inspection voice analysis unit analyzes and extracts keywords from the voices spoken by quality inspectors when they inspect defective home appliances. The initial defect judgment unit marks the appliance as an initial defect judgment appliance when it extracts keywords contained in the defect keyword library from the voice, and generates an appliance log at the same time. The virtual prototype operation unit converts the defect characteristics of the home appliance log into implantable parameters and inputs them into the virtual prototype model to control the virtual prototype model to perform simulation operation. During the operation, it collects various intuitive feedback data from the virtual prototype model. The intuitive feedback data is the data directly reflected by the virtual prototype model or the home appliance during the operation. The types of intuitive feedback data include, but are not limited to, thermal field data, flow field data, structural field data, and electromagnetic field data. The comprehensive intuitive matching unit is used to determine the comprehensive intuitive matching degree of the initially judged home appliance; The method for determining the overall intuitive matching degree of the initial assessment of the home appliance is as follows: control the home appliance to be assessed for actual operation, collect various intuitive feedback data of the home appliance during operation, further determine the intuitive matching degree of various intuitive feedback data, sum the intuitive matching degrees of various intuitive feedback data and take the average value to calculate the overall intuitive matching degree; the intuitive matching degree is the value obtained by calculating the cosine similarity of two intuitive feedback data of the same type. The defective product entry and analysis unit is used to determine, based on the comprehensive and intuitive matching degree of the initially judged home appliance, whether to enter the home appliance log of the initially judged home appliance into the defective product database or to initiate the entry review step. The data entry review execution unit selects review features after the data entry review step is started, checks the review features, and determines whether to enter the review features. Select the re-examination features. The specific steps are as follows: control the virtual prototype model to run multiple defect addition simulations, obtain the comprehensive intuitive matching degree corresponding to each defect addition simulation, and when the comprehensive intuitive matching degree corresponding to a defect addition simulation is higher than the threshold matching degree, mark the new defect feature generated by the defect addition simulation as the re-examination feature.

2. The home appliance manufacturing defective product entry and management system based on voice recognition according to claim 1, characterized in that, Based on the overall intuitive matching degree of the initially judged home appliance, should the home appliance log of the initially judged home appliance be entered into the defective product database or should the entry review step be initiated? If the overall intuitive matching degree of the initially judged home appliance is higher than the threshold matching degree, the initially judged home appliance is directly marked as a defective home appliance, and the home appliance log of the defective home appliance is entered into the defective product database at the same time. When it was initially determined that the overall intuitive matching degree of the home appliances was not higher than the threshold matching degree, the data entry and review process was initiated.

3. The home appliance manufacturing defective product entry and management system based on voice recognition according to claim 1, characterized in that, Check the re-inspection features and determine whether to record them: Arrange quality inspectors to check the re-inspection features of the initially judged home appliances. Whenever a keyword contained in the defect keyword library is extracted from the quality inspector's voice, the defect feature corresponding to the keyword is recorded in the home appliance log of the initially judged home appliance.

4. The home appliance manufacturing defective product entry and management system based on voice recognition according to claim 1, characterized in that, After the data entry and review steps are completed, the overall intuitive matching degree of the initially judged home appliance is determined again. If the overall intuitive matching degree of the initially judged home appliance is higher than the threshold matching degree, the initially judged home appliance is directly marked as a defective home appliance, and the home appliance log of the defective home appliance is simultaneously entered into the defective product database.

5. The home appliance manufacturing defective product entry and management system based on voice recognition according to claim 4, characterized in that, If the overall visual matching degree of the home appliances is determined to be no higher than the threshold matching degree, the data entry and review process will be restarted.

6. A method for recording and managing defective products in home appliance manufacturing based on speech recognition, applied to the speech recognition-based management system for recording defective products in home appliance manufacturing as described in any one of claims 1-5, characterized in that, The steps are as follows: Step 1: Quality Inspection Voice Collection; Step 2: Initial assessment of defective products; Step 3: Defect matching analysis of defective products; Step 4: Re-inspection of defects in defective products.

Images