Discharging screening guiding system of LCD display screen

Abstract

The invention discloses a discharging screening guiding system of an LCD (liquid crystal display) screen, and relates to the technical field of LCD screens, a data acquisition module acquires multi-modal parameters, an adaptive detection module comprehensively analyzes the screen through pre-classification, feature extraction and defect detection, an adaptive screening module sets a judgment threshold according to the type of the screen and performs weighted processing on a defect result, and the defect result is screened according to the judgment threshold. And the discharging guiding module generates different signals according to the judgment result, after the qualified products leave the factory and the unqualified products are subjected to repair cost evaluation, reworking or scrapping treatment is determined, so that the problem that detail defects cannot be effectively identified due to the fact that the difference of different types of screens in structure and performance is not considered in the discharging screening guiding of the existing LCD display screen is solved. According to the invention, comprehensive and accurate detection and reasonable discharging guidance of the LCD display screen are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of LCD (Liquid Crystal Display) screens, and in particular to a material discharging screening and guiding system for LCD screens. Background Art

[0002] The production of LCD displays involves multiple precise process steps, including thin film deposition, photolithography, and packaging. Improper operation in any single step can introduce various defects, leading to uneven display quality, bright and dark spots, color shift, light leakage, and other quality issues, preventing the displays from meeting factory standards. Therefore, implementing outgoing material screening guidance at the final stage of the production line helps identify quality defects in LCD displays and eliminate unqualified displays.

[0003] Different types of LCD screens have significant differences in structure, pixel density, and display quality. High-resolution displays (such as 4K or higher) have higher pixel counts per inch and a lower tolerance for spatial defects. However, existing technologies often use uniform standards to screen different types of LCD screens, failing to account for the differences in structure and performance requirements of different screen types. This results in an inability to effectively identify detailed defects in LCD screens, impacting display quality and quality, and potentially even leading to the release of substandard products into the market.

[0004] Therefore, a discharging screening and guiding system for an LCD display screen is needed to solve the above problems. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention discloses a discharge screening and guidance system for LCD display screens, which realizes comprehensive and accurate detection of LCD display screens and reasonable discharge guidance through the collaborative work of multiple modules.

[0006] The present invention adopts the following technical solutions:

[0007] A discharging screening and guiding system for an LCD display screen, comprising:

[0008] Data acquisition module, used to collect multimodal parameters of the display screen to be discharged;

[0009] The adaptive detection module includes a pre-classification unit, a feature extraction unit, and a defect detection unit. The pre-classification unit is used to identify the screen type of the display screen to be discharged. The feature extraction unit is used to extract the spatial features, temporal features, and physical features of the display screen to be discharged. The defect detection unit is used to perform spatial defect detection, temporal defect detection, and physical defect detection on the display screen to be discharged.

[0010] The adaptive screening module sets the threshold for determining defects in the display screens to be shipped based on the screen type. It also performs adaptive weighting on the different defect detection results according to the degree of impact of different defects on the quality of the display screens to be shipped. When the weighted scores of all defect detection results are lower than the threshold, the product is determined to be qualified; otherwise, it is determined to be unqualified.

[0011] The discharge guidance module generates a discharge path guidance signal when the display screen to be discharged is judged to be qualified. When the display screen to be discharged is judged to be unqualified, the repair cost of the unqualified product is evaluated. When the repair cost is lower than the scrap loss, it is judged to be a rework product and a rework path guidance signal is generated. Otherwise, it is judged to be a scrap product and a scrap processing guidance signal is generated.

[0012] Furthermore, the data acquisition module obtains the specification parameters of the display screen to be discharged by scanning, and dynamically adjusts the flexible sensor array based on the obtained specification parameters to collect the multimodal parameters of the display screen to be discharged. The specification parameters include the size, resolution, brightness and contrast of the display screen, and the multimodal parameters include visual images, electrical parameters, optical parameters and physical parameters. The flexible sensor array dynamically adjusts the arrangement, sampling frequency and measurement range of the sensors in the array through a microcontroller.

[0013] Furthermore, the pre-classification unit performs spatiotemporal alignment of the collected visual images, electrical parameters, optical parameters and physical parameters according to timestamps and spatial positions, and uses a multimodal fusion analysis model to analyze the collected multimodal parameters to identify the screen type of the display screen to be discharged. The feature extraction unit adaptively adjusts the extraction degree of spatial features, temporal features and physical features based on the screen type, and uses edge nodes to distribute the extraction of spatial features, temporal features and physical features of the display screen to be discharged. The defect detection unit uses a self-supervised learning framework to perform spatial defect detection, temporal defect detection and physical defect detection.

[0014] Furthermore, the multimodal fusion analysis model uses a multi-branch neural network to extract feature vectors corresponding to different modal parameters, and inputs them into a classifier to identify the screen types corresponding to different modal feature vectors. Then, the attention mechanism is used to perform weighted fusion on the identified screen types to obtain the screen type of the display screen to be discharged, and the screen types include at least organic light-emitting diode screens, curved screens and liquid crystal screens.

[0015] Furthermore, the feature extraction unit space adopts a gating mechanism to control the activation ratio of the spatial feature channel, the temporal feature channel and the physical feature channel. When extracting spatial features, the edge node adopts image segmentation and principal component analysis to extract the pixel arrangement, brightness distribution and color distribution characteristics of the visual image, and adopts edge detection to identify the edge and contour characteristics of the display screen to be discharged. When extracting temporal features, the edge node extracts the response time and harmonic dynamic characteristics of the display screen to be discharged by performing time-frequency transformation on the current and voltage signals, and obtains the dynamic contrast and afterimage characteristics through the brightness change curve. When extracting physical features, the edge node establishes a temperature gradient field feature model through thermal imaging, and obtains the curvature of the display screen to be discharged by measuring the curvature radius and curvature angle.

[0016] Furthermore, the self-supervised learning framework uses parallel computing to perform spatial defect detection, temporal defect detection and physical defect detection. When performing spatial defect detection, the U-Net segmentation network is used to annotate the defect area at the pixel level and output the defect area ratio. When performing temporal defect detection, the LSTM time series prediction model is used to obtain the comparison deviation between the predicted response curve and the actual measurement value. When performing physical defect detection, the stress distribution is simulated based on finite element analysis, and the potential cracking risk rate is calculated based on the measured vibration data.

[0017] Furthermore, the adaptive screening module stores the characteristic information of all known defects by constructing a dynamic defect library, and sets defect judgment thresholds for different types of screens based on the quality requirements of different screen types and known defect data, and uses an improved hierarchical analysis method to adaptively weight different defect detection results. The improved hierarchical analysis method converts subjective experience weight distribution judgment into objective quantitative weight distribution through fuzzy mathematics and dynamic feedback mechanism.

[0018] Furthermore, the working method of the improved hierarchical analysis method includes the following steps:

[0019] S1. Construct a three-layer indicator system, which includes the target layer, the criterion layer, and the solution layer. The target layer is the overall quality of the display screen, the criterion layer includes spatial defects, timing defects, and physical defects, and the solution layer is the degree of impact of different defects on the quality of the output display screen;

[0020] S2. Use triangular fuzzy numbers to quantify the indicators of the criterion layer and the solution layer, and construct dynamic confidence intervals for different screen types, different production batches and different quality requirements;

[0021] S3. Constructing a fuzzy judgment matrix based on the quantified triangular fuzzy numbers and dynamic confidence intervals, and performing weight calculation using an extended analysis method, wherein the extended analysis method comprehensively analyzes the influence weights of different indicators;

[0022] S4. The dynamic defect library updates the defect feature vector in real time. When an update of the defect type is detected, the weight redistribution is triggered.

[0023] Furthermore, the discharge guidance module adopts a multi-point operation control model to respectively generate a factory path guidance signal, a rework path guidance signal and a scrap processing guidance signal. The multi-point operation control model adopts a distributed operation node structure to independently control and decide on the factory decision node, the rework decision node and the scrap decision node. The factory decision node determines the optimal factory path of qualified products based on the preset factory layout and the shipping process information library, and generates a factory path guidance signal according to the coding rules. The rework decision node determines the optimal rework path of qualified products based on the preset factory layout and the rework process information library, and generates a rework path guidance signal according to the coding rules. The scrap decision node determines the optimal rework path of the rework products based on the defects of the rework products and the workload of the rework location, and generates a rework path guidance signal according to the coding rules. The scrap decision node determines the optimal scrap path of the scrap products based on the scrap processing information library and the scrap processing area, and generates a scrap processing guidance signal according to the coding rules.

[0024] Furthermore, in the process of executing the outgoing material screening instructions, full-link quality traceability is achieved by embedding blockchain hash values, and SPC control charts are used to display the distribution and frequency of detection defects in real time.

[0025] The beneficial effects of the present invention are:

[0026] 1. The present invention collects multimodal parameters, including visual images, electrical parameters, optical parameters and physical parameters, through the data acquisition module, which can comprehensively obtain the information of the display screen to be discharged, greatly improve the accuracy of defect detection, and reduce missed detection and misjudgment.

[0027] 2. This invention first uses a pre-classification unit to identify the screen type, allowing subsequent feature extraction and defect detection to be tailored to the specific characteristics of each screen type. This avoids potential misjudgments caused by standardized testing, significantly improving the accuracy of defect detection for different LCD display types. The feature extraction unit then extracts features from three dimensions: spatial, temporal, and physical. The defect detection unit also performs multi-dimensional detection, enabling comprehensive detection of all possible defects on the display, further enhancing detection accuracy.

[0028] 3. This invention sets defect thresholds based on screen type, fully accounting for the varying tolerances for defects across different screen types. This ensures screening results are more realistic. Adaptive weighting is performed based on the impact of different defects on display quality, enabling a more scientific assessment of the overall quality of the display. This prevents a single defect from significantly impacting the screening results, thus improving the rationality of the screening process.

[0029] 4. The present invention generates corresponding path guidance signals according to the judgment results of the display screen, and directly guides qualified products to leave the factory. After evaluating the repair cost of unqualified products, it decides whether to rework or scrap them, making the entire discharge process clearer and more efficient. By evaluating the repair cost of unqualified products and giving priority to rework, it can reduce production costs and improve resource utilization while ensuring product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the overall system architecture of the present invention;

[0031] Figure 2 It is a schematic diagram of the overall flow of the system of the present invention;

[0032] Figure 3 Schematic diagram of the improved hierarchical analysis method in the present invention. DETAILED DESCRIPTION

[0033] The following is a combination of the embodiments of the present invention Figure 1 To the attached Figure 3 The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0034] The embodiment of the present invention discloses a discharging screening and guidance system for LCD display screens. This system realizes the automated quality control of the entire discharging process of LCD display screens through four core modules: multimodal data acquisition, adaptive detection, dynamic screening, and intelligent guidance. Each module works together to ensure efficient and accurate screening of qualified products, reworked products, and scrapped products. Figure 1 As shown, the system includes:

[0035] The data acquisition module is used to collect the visual images, electrical parameters, optical parameters, and physical parameters of the display screen to be discharged. It uses a high-resolution industrial camera to capture screen surface images, supports multi-angle shooting, uses a multimeter or dedicated test equipment to measure voltage, current, response time, etc., uses a spectrum analyzer to detect brightness, contrast, color temperature, color gamut coverage, etc., and uses sensors to measure size, thickness, weight, surface flatness, etc., to structure multimodal data sets for processing by subsequent modules.

[0036] The adaptive detection module is used to perform detailed inspection and analysis of the display screen. This module is divided into three sub-units. The pre-classification unit pre-identifies the type of display screen based on the collected multimodal parameters. This step can quickly distinguish different types of displays, such as LED and OLED, providing a basis for subsequent inspection.

[0037] The feature extraction unit extracts the spatial features of the display screen (such as whether there are scratches or cracks on the surface of the display screen), temporal features (such as whether the display effect is stable, response time, etc.) and physical features (such as strength, weight, etc.) from the multimodal parameters. These features are very important for determining whether the display screen is qualified; the defect detection unit performs defect detection based on the extracted features, which is divided into three categories: spatial defect detection: checks whether there are visible defects on the surface of the display screen, such as scratches, bubbles, color difference, etc.; temporal defect detection: checks the display effect of the display screen to see whether there are problems such as color difference and uneven brightness; physical defect detection: checks whether the physical performance of the display screen meets the standards, such as unqualified size, abnormal weight, etc.

[0038] The adaptive screening module sets an appropriate defect threshold based on the display type. Different display types have varying tolerances for defects, necessitating customized criteria. All defect detection results are weighted based on their impact on display quality. Each defect is weighted differently, with severe defects having a greater impact and minor defects receiving a smaller weight. Ultimately, the weighted score is used to comprehensively assess the display's quality. If the weighted score falls below the set threshold, the display is deemed acceptable; otherwise, it is deemed unacceptable.

[0039] The discharge guidance module generates corresponding guidance signals based on the adaptive screening module's judgment results: Conforming products: When a display screen is judged as conforming, a factory routing guidance signal is generated, directing the display screen to the normal factory process. Defective products: When a display screen is judged as defective, the system evaluates the cost of repair. If the repair cost is lower than the scrap loss, the display screen is considered reworked and a rework routing guidance signal is generated; otherwise, the display screen is considered scrapped and a scrapping guidance signal is generated.

[0040] The output end of the data acquisition module is connected to the input end of the adaptive detection module, the output end of the adaptive detection module is connected to the input end of the adaptive screening module, and the output end of the adaptive screening module is connected to the input end of the discharge guidance module.

[0041] As attached Figure 2 As shown, the overall system process includes the following steps:

[0042] Step 1: Multimodal sensors synchronously collect screen parameters;

[0043] Step 2: Adaptive detection of the display screen to be discharged, using a pre-classification unit to identify the screen type, a feature extraction unit to extract spatial, temporal, and physical features, and a defect detection unit to perform multi-dimensional defect recognition;

[0044] Step 3: Adaptive screening of display screens to be discharged is achieved through dynamic threshold determination and weighted scoring;

[0045] Step 4: Generate factory, rework or scrap path signals based on the quality results to provide discharge guidance.

[0046] The data acquisition module obtains the specifications of the display screen to be discharged through scanning and dynamically adjusts the configuration of the flexible sensor array based on these specifications. Specifically, it includes the following steps:

[0047] First, the module scans and acquires the key specifications of the display, including: the physical dimensions of the screen, such as length and width; the number of pixels on the screen, typically expressed as horizontal and vertical pixels (resolution); the intensity of light emitted by the display (brightness), and the difference in brightness between the brightest and darkest levels of the display (contrast). Based on the acquired specifications, the flexible sensor array dynamically adjusts its configuration, including: adjusting the sensor layout according to the size and resolution of the display to accommodate displays of different sizes and resolutions; adjusting the frequency of sensor data acquisition based on the operating characteristics of the display (such as the image update rate or the speed at which the displayed content changes) to capture accurate multimodal data; and adjusting the sensor's measurement range based on parameters such as the brightness and contrast of the display to ensure that relevant data can be accurately captured.

[0048] Flexible sensor arrays are used not only to capture traditional visual images but also to capture multiple other parameters. These include: capturing images of the display screen through visual sensors (such as cameras or image sensors); capturing the display's electrical characteristics, such as current and voltage; capturing optical characteristics, such as brightness, spectral distribution, and reflectivity; and environmental factors, such as temperature, humidity, and pressure, that can affect the display's performance. A microcontroller controls and adjusts the sensor array's operating mode in real time, ensuring dynamic adjustment of the array's behavior based on the display's specifications, thereby achieving efficient and accurate data acquisition.

[0049] In summary, the data acquisition module can flexibly collect various multimodal data related to display performance according to the specifications of different display screens by automatically adjusting the configuration of the flexible sensor array for further analysis or application.

[0050] The pre-classification unit first aligns the data of different modalities (visual images, electrical parameters, optical parameters and physical parameters) in time and space through timestamps and spatial positions. This ensures that the data collected by different sensors have a corresponding relationship at the same time and spatial position, which helps to improve the accuracy of the analysis. A multimodal fusion model is used to analyze various types of data, and the screen type of the display screen to be discharged is identified by fusing the features of different modalities. Multimodal fusion can effectively integrate different types of signals and information, thereby improving classification accuracy. The multimodal fusion analysis model uses a multi-branch neural network, and each branch processes input data of different modalities (such as visual images, electrical signals, etc.). Each branch extracts the feature vectors of different modalities, and then identifies them through a classifier. Finally, these recognition results are weighted and fused through the attention mechanism to output the final screen type. This method can identify different types of display screens, including OLED screens, curved screens, and LCD screens.

[0051] The feature extraction unit adaptively adjusts its spatial, temporal, and physical feature extraction strategies based on the screen type. This helps extract more effective and relevant features for different display types. Edge nodes are used to extract the display's spatial, temporal, and physical features through distributed computing. This edge computing approach helps reduce the burden on the central processing unit and improve processing efficiency.

[0052] Spatial feature extraction: Image segmentation and principal component analysis (PCA) are used to extract pixel arrangement, brightness distribution, and color distribution features from visual images. Edge detection algorithms are used to identify the edges and contours of the display. These features are crucial for subsequent defect detection and screen recognition.

[0053] Timing feature extraction: By performing time-frequency transformation on the current and voltage signals, the screen's response time and harmonic dynamic characteristics are extracted. Dynamic contrast and image sticking characteristics are obtained from the brightness change curve.

[0054] Physical feature extraction: Thermal imaging technology establishes a temperature gradient field to further extract the temperature distribution characteristics of the display. In addition, measuring the curvature radius and curvature angle can help evaluate the shape deformation of the display, especially for curved screens.

[0055] A self-supervised learning framework is used to detect spatial, temporal and physical defects through parallel computing.

[0056] Spatial defect detection: Using the U-Net segmentation network for pixel-level annotation, it can accurately identify defective areas on the display and output the defect area ratio.

[0057] Timing defect detection: Using the LSTM (Long Short-Term Memory) time series prediction model, the deviation between the predicted response curve and the actual measurement value is compared to identify the timing problems of the screen.

[0058] Physical defect detection: Simulates stress distribution based on finite element analysis (FEA) and calculates potential cracking risks based on vibration data.

[0059] The Adaptive Inspection Module combines multimodal data analysis, deep learning, edge computing, and self-supervised learning to efficiently detect defects across a wide range of display types. It not only accurately identifies screen types but also efficiently detects spatial, temporal, and physical defects, thereby improving display quality control.

[0060] The adaptive screening module stores the characteristic information of all known defects by establishing a dynamic defect inventory, and sets defect judgment thresholds for different types of screens in combination with the quality requirements and known defect data of different screen types. The dynamic defect inventory is used to store the characteristic information of all known defects. This defect information includes the type, location, size, shape, brightness, color and other characteristics of the defect. These data are compiled based on historical test results or pre-collected quality data, and can provide a basis for subsequent defect judgment. According to the different quality requirements of different screen types (such as LED screens, OLED screens, etc.), the defect judgment threshold is set dynamically. For example, some screen types may have a higher tolerance for small-scale defects, while other screen types have a lower tolerance for defects. By analyzing historical defect data and the quality standards of screen types, adaptive defect detection standards can be set for different types of screens.

[0061] The improved analytic hierarchy process (AHP) performs a multi-dimensional, hierarchical weighted analysis of defect detection results to determine the final importance score for each defect. This method assigns different weights to different defect types, allowing for a more accurate assessment of their impact on screen quality. The improved approach optimizes the traditional AHP method, specifically employing a more scientific and efficient strategy for weight assignment and judgment matrix processing. While the traditional AHP method may rely too heavily on human experience, the improved AHP method incorporates fuzzy mathematics and a dynamic feedback mechanism, resulting in a more objective and precise weight assignment.

[0062] Fuzzy mathematics is an effective method for dealing with uncertainty and ambiguity. In this system, fuzzy mathematics is used to quantify subjective experience, helping to resolve the ambiguity in judgment that can arise in traditional methods. Through fuzzy mathematics, subjective judgments can be transformed into clearer quantitative data, avoiding human bias. A dynamic feedback mechanism enables the system to adaptively adjust judgment criteria and weightings based on actual test results. As test data accumulates, the system continuously optimizes its weighting and defect threshold settings through feedback, improving overall test accuracy and reliability.

[0063] Through the above method, the system can weight the results of different types of defect detection, assigning different weights to each defect based on its importance and the requirements of the screen type. This adaptive weighting process can more accurately identify which defects have a greater impact on screen quality, thereby improving screening efficiency and accuracy.

[0064] In summary, this adaptive screening module, by combining a dynamic defect inventory, an improved hierarchical analysis method, fuzzy mathematics, and a dynamic feedback mechanism, can provide more accurate defect judgment criteria for different types of screens, and optimize defect detection results through adaptive weighting, thereby improving overall detection accuracy and screen quality control effects.

[0065] The improved analytic hierarchy process combines fuzzy mathematics with dynamic feedback mechanism to achieve the reasonable allocation of defect judgment weights. Figure 3 As shown, its workflow includes the following steps:

[0066] S1: Construct a three-tier indicator system

[0067] Target layer: This layer mainly focuses on the comprehensive quality of the display screen, and its goal is to evaluate the quality level of the entire display screen.

[0068] The criteria layer includes three main defect types: Spatial defects: Physical problems on the display surface or structure, such as dead pixels and bright spots. Timing defects: Timing synchronization issues that may occur during the operation of the display, usually involving refresh rate, etc. Physical defects: Hardware damage and production defects of the display.

[0069] Solution layer: This layer evaluates the impact of different defects on display quality. The severity of each defect affects the final quality evaluation.

[0070] S2: Quantization using triangular fuzzy numbers

[0071] Triangular fuzzy numbers are used to quantify metrics at the criterion and solution levels. They can express the ambiguity and uncertainty of each metric. For example, the impact of a defect on screen quality cannot be simply quantified; it can be expressed as "slight impact," "moderate impact," or "severe impact."

[0072] Dynamic confidence intervals are established for each indicator based on different screen types, production batches, and quality requirements. This helps adjust judgment criteria in real time and improves adaptability.

[0073] S3: Constructing a fuzzy judgment matrix and performing weight calculation

[0074] The fuzzy judgment matrix is ​​constructed by combining the quantified triangular fuzzy numbers and dynamic confidence intervals to express the relationship and relative importance between various indicators.

[0075] The extended analysis method further calculates the weight of the fuzzy judgment matrix and comprehensively analyzes the influence weights of different indicators. By calculating the relative importance of each indicator, the degree of influence of each defect type on the overall quality of the display is determined.

[0076] S4: Real-time update and weight redistribution of dynamic defect database

[0077] Known defect feature vectors are updated in real time. When new defect types are detected, the defect library will be updated.

[0078] Once a new defect type is detected, the system automatically recalculates the defect weight distribution and adjusts the display quality assessment results. This ensures that the system always makes judgments based on the latest defect data.

[0079] Through the above steps, the improved hierarchical analysis method is combined with fuzzy mathematics and dynamic feedback mechanism, so that the defect detection process can adaptively adjust the weights and quantify the impact of different defects on screen quality in an objective manner. This method not only improves the accuracy of defect detection, but also can update and optimize the evaluation criteria in real time according to changes in production conditions, thereby improving the quality control level of the display production process. The discharge guidance module uses a multi-point operation control model to generate factory path guidance signals, rework path guidance signals, and scrap processing guidance signals respectively. Its core function is to control different decision nodes through distributed operation to generate path guidance signals for factory, rework, and scrap processing respectively. It can make the best decision for different product states (such as qualified products, rework products, or scrap products) and generate path guidance signals accordingly.

[0080] The multi-point computing control model uses a distributed computing node structure, managing different decisions through multiple independent control nodes. Each decision node corresponds to a different task: the factory decision node is responsible for determining the path of qualified products; the rework decision node is responsible for determining the path of reworked products; and the scrap decision node is responsible for determining the path of scrapped products.

[0081] The shipping decision node uses factory layout information and a shipping process database to determine the optimal shipping route for qualified products. It then generates a routing guidance signal based on specific coding rules to ensure that qualified products can be shipped smoothly. The rework decision node uses the rework process database and factory layout information to determine the optimal rework route for reworked products and generates a corresponding rework route guidance signal. This ensures the efficiency and accuracy of the rework process. The scrap decision node determines whether rework is necessary based on the defect data of the reworked products and the workload at the rework location. If necessary, it uses the scrap disposal database to determine the scrap disposal route for the scrapped products. The generated scrap route guidance signal helps the product enter the scrap disposal process.

[0082] All decision nodes (shipping, rework, and scrapping) generate guidance signals based on coded rules. These signals help factory operators and automation systems understand the status of products in the production process and guide them to the appropriate path, reducing errors and waste.

[0083] In summary, this system can effectively guide products along different paths during the production process, ensuring efficient and accurate material flow on the production line. This multi-point computing control model can significantly enhance the intelligence of production management and optimize overall factory efficiency.

[0084] In the process of executing the outgoing material screening guidelines, the combination of embedding blockchain hash values ​​and using SPC control charts can significantly improve the accuracy of product quality traceability and the visualization of process control.

[0085] Blockchain, through its distributed ledger technology, provides a secure, immutable record-keeping system. By embedding blockchain hash values ​​in the material screening guidelines, full-chain product quality traceability can be achieved, ensuring that every step of the product, from production and testing to shipment, is recorded and cannot be tampered with. Every time a product passes through a quality inspection stage, the system generates a hash value, recording the product's inspection data (such as defect type, inspection personnel, timestamp, etc.). The hash value serves as a unique identifier for this data and is stored on the blockchain. Once the data is embedded in the blockchain, it can never be modified or deleted. Even if problems arise later, every inspection step and its specific data can be accurately traced back to the time. Through blockchain technology, users or managers can query the production and inspection process of each product in real time to understand its quality, thereby providing transparent quality information to consumers or regulators.

[0086] SPC (Statistical Process Control) helps companies identify potential quality fluctuations and ensure product stability by monitoring and analyzing process data in real time. Using SPC control charts to display the distribution and frequency of detection defects in real time, inspectors can observe the status of quality control in the production process at any time by displaying SPC control charts in real time. When product quality deviates, the control chart can quickly reflect it and remind inspectors to take corrective measures. SPC control charts can help analyze the frequency and distribution patterns of defects. For example, by displaying the types of defects in different time periods and different production batches, the occurrence trends and potential causes of defects can be analyzed, so that targeted improvement measures can be taken. When the control chart shows an abnormality (for example, a data point exceeds the control limit), the system will automatically issue an alarm, prompting relevant personnel to make adjustments or stop production for inspection to prevent the outflow of unqualified products.

[0087] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is clearly not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A discharging screening and guiding system for an LCD display screen, characterized in that: include: Data acquisition module, used to collect multimodal parameters of the display screen to be discharged; The adaptive detection module includes a pre-classification unit, a feature extraction unit, and a defect detection unit. The pre-classification unit is used to identify the screen type of the display screen to be discharged. The feature extraction unit is used to extract the spatial features, temporal features, and physical features of the display screen to be discharged. The defect detection unit is used to perform spatial defect detection, temporal defect detection, and physical defect detection on the display screen to be discharged. The adaptive screening module sets the threshold for determining defects in the display screens to be shipped based on the screen type. It also performs adaptive weighting on the different defect detection results according to the degree of impact of different defects on the quality of the display screens to be shipped. When the weighted scores of all defect detection results are lower than the threshold, the product is determined to be qualified; otherwise, it is determined to be unqualified. The discharge guidance module generates a discharge path guidance signal when the display screen to be discharged is judged to be qualified. When the display screen to be discharged is judged to be unqualified, the repair cost of the unqualified product is evaluated. When the repair cost is lower than the scrap loss, it is judged to be a rework product and a rework path guidance signal is generated. Otherwise, it is judged to be a scrap product and a scrap processing guidance signal is generated.

2. The discharging screening and guiding system for LCD display screen according to claim 1, characterized in that: The data acquisition module obtains the specification parameters of the display screen to be discharged by scanning, and dynamically adjusts the flexible sensor array based on the obtained specification parameters to collect the multimodal parameters of the display screen to be discharged. The specification parameters include the size, resolution, brightness and contrast of the display screen, and the multimodal parameters include visual images, electrical parameters, optical parameters and physical parameters. The flexible sensor array dynamically adjusts the arrangement, sampling frequency and measurement range of the sensors in the array through a microcontroller.

3. The discharging screening and guiding system for LCD display screen according to claim 1, characterized in that: The pre-classification unit performs spatiotemporal alignment of the collected visual images, electrical parameters, optical parameters, and physical parameters according to timestamps and spatial positions, and uses a multimodal fusion analysis model to analyze the collected multimodal parameters to identify the screen type of the display screen to be discharged. The feature extraction unit adaptively adjusts the degree of extraction of spatial features, temporal features, and physical features based on the screen type, and uses edge nodes to distribute the extraction of spatial features, temporal features, and physical features of the display screen to be discharged. The defect detection unit uses a self-supervised learning framework to perform spatial defect detection, temporal defect detection, and physical defect detection.

4. The discharging screening and guiding system for LCD display screen according to claim 3, characterized in that: The multimodal fusion analysis model uses a multi-branch neural network to extract feature vectors corresponding to different modal parameters and input them into a classifier to identify the screen types corresponding to the different modal feature vectors. Then, an attention mechanism is used to perform weighted fusion on the identified screen types to obtain the screen type of the display screen to be discharged. The screen types include at least organic light-emitting diode screens, curved screens and liquid crystal displays.

5. The discharging screening and guiding system for LCD display screen according to claim 3, characterized in that: The feature extraction unit space adopts a gating mechanism to control the activation ratio of the spatial feature channel, the temporal feature channel and the physical feature channel. When extracting the spatial feature, the edge node adopts image segmentation and principal component analysis to extract the pixel arrangement, brightness distribution and color distribution characteristics of the visual image, and adopts edge detection to identify the edge and contour characteristics of the display screen to be discharged. When extracting the temporal feature, the edge node extracts the response time and harmonic dynamic characteristics of the display screen to be discharged by performing time-frequency transformation on the current and voltage signals, and obtains the dynamic contrast and afterimage characteristics through the brightness change curve. When extracting the physical feature, the edge node establishes a temperature gradient field feature model through thermal imaging, and obtains the curvature of the display screen to be discharged by measuring the curvature radius and curvature angle.

6. The discharging screening and guiding system for LCD display screen according to claim 3, characterized in that: The self-supervised learning framework uses parallel computing to perform spatial defect detection, temporal defect detection, and physical defect detection. When performing spatial defect detection, a U-Net segmentation network is used to annotate defect areas at the pixel level and output the defect area ratio. When performing temporal defect detection, an LSTM time series prediction model is used to obtain the comparison deviation between the predicted response curve and the actual measurement value. When performing physical defect detection, stress distribution is simulated based on finite element analysis, and the potential cracking risk rate is calculated based on measured vibration data.

7. The discharging screening and guiding system for LCD display screen according to claim 1, characterized in that: The adaptive screening module builds a dynamic defect library to store the characteristic information of all known defects, sets defect judgment thresholds for different types of screens based on the quality requirements of different screen types and known defect data, and uses an improved hierarchical analysis method to adaptively weight different defect detection results. The improved hierarchical analysis method converts subjective experience weight distribution judgment into objective quantitative weight distribution through fuzzy mathematics and dynamic feedback mechanism.

8. The discharging screening and guiding system for LCD display screen according to claim 7, characterized in that: The working method of the improved analytic hierarchy process comprises the following steps: S1. Construct a three-layer indicator system, which includes the target layer, the criterion layer, and the solution layer. The target layer is the overall quality of the display screen. The criterion layer includes spatial defects, timing defects, and physical defects. The solution layer is the degree of impact of different defects on the quality of the output display screen. S2. Use triangular fuzzy numbers to quantify the indicators of the criterion layer and the solution layer, and construct dynamic confidence intervals for different screen types, different production batches and different quality requirements; S3. Constructing a fuzzy judgment matrix based on the quantified triangular fuzzy numbers and dynamic confidence intervals, and performing weight calculation using an extended analysis method, wherein the extended analysis method comprehensively analyzes the influence weights of different indicators; S4. The dynamic defect library updates the defect feature vector in real time. When an update of the defect type is detected, the weight redistribution is triggered.

9. The discharging screening and guiding system for LCD display screen according to claim 1, characterized in that: The discharge guidance module adopts a multi-point operation control model to generate a factory path guidance signal, a rework path guidance signal and a scrap processing guidance signal respectively. The multi-point operation control model adopts a distributed operation node structure to independently control and decide on the factory decision node, the rework decision node and the scrap decision node. The factory decision node determines the optimal factory path of qualified products based on the preset factory layout and the shipping process information library, and generates a factory path guidance signal according to the coding rules. The rework decision node determines the optimal rework path of qualified products based on the preset factory layout and the rework process information library, and generates a rework path guidance signal according to the coding rules. The scrap decision node determines the optimal rework path of the rework products based on the defects of the rework products and the workload of the rework location, and generates a rework path guidance signal according to the coding rules. The scrap decision node determines the optimal scrap path of the scrap products based on the scrap processing information library and the scrap processing area, and generates a scrap processing guidance signal according to the coding rules.

10. The discharging screening and guiding system for LCD display screen according to claim 1, characterized in that: In the process of executing the outgoing material screening instructions, full-link quality traceability is achieved by embedding blockchain hash values, and SPC control charts are used to display the distribution and frequency of detection defects in real time.

Images