A production management method and system based on TFT color LCD panel inspection

Abstract

The present invention discloses a production management method and system based on TFT color liquid crystal panel detection, and relates to the technical field of liquid crystal panel production management. It includes detecting raw materials to obtain real-time raw material detection data; obtaining various parameter items of the TFT color liquid crystal panel production line, and determining the adjustable parameter items of the equipment; obtaining the detection data of the finished product, the product detection data of the finished product; creating an adjustment model, the adjustment model is used to analyze the real-time detection data and output the adjustment parameters of the corresponding adjustable parameter items as parameter presets. The present invention creates an adjustment model, uses historical raw material detection data and production line adjustment data as the basis to create an intelligent model, and when the raw material parameter changes are detected later, the adjustment model can generate corresponding production parameters, which can weaken the problem of changes in production yield caused by differences in raw materials of different specifications, improve the risk resistance of the production line, and improve the fault tolerance of the production line.

Description

Technical Field

[0001] The present invention relates to the technical field of liquid crystal panel production management, and in particular to a production management method and system based on TFT color liquid crystal panel detection. Background Art

[0002] TFT color LCD panel is one of the most widely used display technologies. It occupies an important position in consumer electronics, automotive displays, industrial control, medical equipment and other fields with its good display performance and wide application scenarios. In order to carry out efficient production of TFT color LCD panels, most of the production lines are currently equipped with exclusive production management systems.

[0003] For example, a liquid crystal display production line operation management system based on the Internet of Things with publication number CN117132064A analyzes the processing equipment of the liquid crystal display production line from both external and internal perspectives to understand the processing safety level of the processing equipment in the overall liquid crystal display production line, so as to manage it reasonably and in a targeted manner. It also intuitively understands the situation of each liquid crystal display production line through information feedback and text warnings, which helps operation and management personnel to manage the liquid crystal display production lines that need to be controlled reasonably and in a targeted manner according to different control levels, so as to ensure the processing quality and efficiency of the liquid crystal display production line, and at the same time helps to improve the management rationality and operation safety of the liquid crystal display production line.

[0004] The above-mentioned and similar management systems are all aimed at the operation of the product line itself, but in the process of LCD production, the batches, suppliers and types of raw materials will change. In the current production process, these changes are almost unavoidable, because the difference in raw materials will cause the production parameters of the entire LCD production line to be adjusted to meet production needs. However, when purchasing raw materials, there are fixed raw material standards. Therefore, even if the raw materials from different suppliers and different batches are different, the difference in raw materials is low and meets the procurement needs. However, during production, the production parameters need to be adapted according to the differences in raw materials from different batches, so as to ensure the yield of production and avoid other production problems related to product quality. At present, the adjustment of these parameters is generally done by manual discovery and parameter adjustment after a period of production, and almost every batch of raw materials needs to be adjusted. The efficiency of the entire parameter adjustment is low, which will have a certain impact on the quality of production. Summary of the invention

[0005] The object of the present invention is to provide a production management method and system based on TFT color liquid crystal panel detection to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a production management method and system based on TFT color liquid crystal panel detection, including detecting raw materials to obtain real-time raw material detection data;

[0007] Obtain various parameters of the TFT color LCD panel production line and determine the adjustable parameters of the equipment;

[0008] Acquire test data of a finished product, product test data of the finished product;

[0009] Creating an adjustment model, wherein the adjustment model is used to analyze the real-time detection data and output adjustment parameters corresponding to the adjustable parameter items as parameter presets;

[0010] The method for creating the adjustment model includes:

[0011] S1: Acquire various test data of historical raw materials as historical raw material data, where the historical raw material data includes multiple data items, and each data item includes a test item corresponding to the historical raw material;

[0012] S2: Obtaining parameter data of each adjustable parameter item when using historical raw materials for production and historical finished product data of finished products produced using corresponding historical raw materials;

[0013] S3: Taking the historical raw material data item of one of the raw material items in the historical raw materials as a benchmark, determining the matching relationship between each detection item and each adjustable parameter item based on a matching method, and a single adjustable parameter item matches at least one detection item;

[0014] S4: Using the historical raw material data item of one of the raw material items in the historical raw materials as the basic data, based on the visualization method, a comparison chart of the detection item data matching the same adjustable parameter item and the detection item data as the basic data is obtained, and each comparison chart corresponds to the parameter data of the adjustable parameter item of the corresponding raw material item and the historical finished product data;

[0015] S5: Based on the convolutional neural network, the comparison graph, the parameter data of the adjustable parameter items associated with the comparison graph, and the historical finished product data are used as training data to train and obtain an adjustment model.

[0016] Furthermore, the parameter presets are fine-tuned to obtain final parameters, which are input into the TFT color LCD panel production line, and corresponding finished products are obtained based on the corresponding raw materials. After the corresponding finished products are inspected, the inspection data, final parameters and real-time raw material inspection data are input into the adjustment model for repeated learning to improve the model function and realize self-iteration of the model.

[0017] Furthermore, the matching method includes:

[0018] N1: Get the detection items and adjustable parameter items respectively;

[0019] N2: According to the historical raw material data items of the raw material items used as the benchmark, a benchmark matching relationship is established, in which each detection item has a matching relationship with each adjustable parameter item;

[0020] N3: based on the data changes of the various detection items of the remaining historical raw material data items and the data changes of the adjustable parameter items, eliminate the non-matching relationship between the various detection items and the various adjustable parameter items;

[0021] N4: After repeated N3 processing, the final matching relationship between each detection item and each adjustable parameter item is obtained.

[0022] Furthermore, the matching method includes:

[0023] Q1: Obtain the influence rate of each adjustable parameter item on each detection item of the color LCD panel;

[0024] Q2: The influence relationship between the adjustable parameter items and the detection items of the removal influence rate in the preset standard;

[0025] Q3: Based on the influence rate of the remaining adjustable parameter items retained in Q2, each adjustable parameter item is assigned an influence weight to establish a matching relationship with the detection item.

[0026] Furthermore, the visualization method comprises:

[0027] M1: Obtain detection items matched by the adjustable parameter items according to the matching relationship, and construct an impact group for each detection item matched by the adjustable parameter item;

[0028] M2: Obtain the value of the detection item in the historical raw material data item of the raw material item corresponding to the basic data, and construct a rectangular range to divide the reference range;

[0029] M3: Divide the change range on the benchmark range in proportion to the change relationship between the values ​​of each test item in the impact group and the values ​​of the test items corresponding to the basic data;

[0030] M4: Obtain a rectangular image containing the variation range and the reference range as a comparison image.

[0031] Furthermore, in M2, a circular range is inscribed in the rectangular range, and a suitable figure is selected as the inscribed figure of the circular range according to the number of detection items matched by a single adjustable parameter item, and the variable range and the reference range are constructed based on the edges or corners of the inscribed figure.

[0032] Furthermore, the specific training method of the adjustment model in S5 includes:

[0033] P1: Matrix image, divide the contrast image according to pixel size to obtain a matrix image;

[0034] P2: Matrix assignment, assign different values ​​to each variation range and reference range to obtain a numerical matrix diagram. The variation range and reference range corresponding to the same test item have the same value assigned to them.

[0035] P3: Depth convolution, select appropriate convolution kernels for multi-layer convolution, and after the convolution is completed, strip the convolved numerical matrix into a numerical string;

[0036] P4: Store the value strings to obtain the adjustment model, where each value string corresponds to a specific value of an adjustable parameter item.

[0037] A production management system based on TFT color liquid crystal panel detection uses the above-mentioned production management method based on TFT color liquid crystal panel detection.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The production management method and system based on TFT color liquid crystal panel detection can create an intelligent model based on historical raw material detection data and production line adjustment data through the creation of an adjustment model, and can allow the adjustment model to generate corresponding production parameters when raw material parameter changes are subsequently detected, thereby eliminating the need to manually adjust production line parameters based on raw material parameter changes. This can reduce the problem of changes in production yield caused by differences in raw materials provided by different batches and suppliers, improve the risk resistance of the production line, and improve the fault tolerance of the production line.

[0040] At the same time, in the process of creating the adjustment model, a convolutional neural network capable of high-precision image processing is used. According to the characteristics of the convolutional neural network, the polygonal basic image shape and the regional design of the variation range and the reference range can make the detection item of each parameter more obvious than the comparison graph features, thereby improving the model accuracy and the accuracy of subsequent recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0042] Figure 2 generating a schematic diagram for a regulation model of the present invention;

[0043] Figure 3 It is a schematic diagram of the production process of the present invention;

[0044] Figure 4 This is a schematic diagram of the initial matching of sample 1 of the present invention;

[0045] Figure 5 This is a matching schematic diagram of sample 2 of the present invention;

[0046] Figure 6 This is a matching schematic diagram of sample 3 of the present invention;

[0047] Figure 7 It is a matching schematic diagram of sample 4 of the present invention;

[0048] Figure 8 It is a schematic diagram of partial matching results obtained based on four samples of the present invention;

[0049] Fig. 9 A in FIG. 1 is a comparison diagram of detection parameters corresponding to two detection items of the present invention;

[0050] Fig. 9 B in FIG. 1 is a comparison diagram of detection parameters corresponding to the three detection items of the present invention;

[0051] Fig. 9 C in FIG. 1 is a comparison chart of detection parameters corresponding to the four detection items of the present invention;

[0052] Fig. 9 D in FIG. 1 is a comparison diagram of detection parameters corresponding to the five detection items of the present invention;

[0053] Fig.10 A in the figure is a schematic diagram of bidirectional parameter adjustment of the present invention;

[0054] Fig.10 B is a schematic diagram of multi-directional parameter adjustment of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In the production process of TFT color liquid crystal panels, different batches and raw materials from different suppliers need to be quality tested before production to ensure that the quality of the raw materials meets the regulations. Due to the different production conditions of different batches and different suppliers, there will be slight differences in the parameters of the raw materials. Although these slight differences will not affect the final production to a large extent, in order to cope with the slight changes in the raw materials, the entire TFT color liquid crystal panel production line needs to adjust the corresponding parameters, thereby offsetting various production problems caused by changes in the quality of the raw materials. Since the core display unit of the TFT color liquid crystal panel is small, these tiny errors will be more obvious in the yield rate. The basic materials that currently make up the TFT color liquid crystal panel include: 1. Glass substrate. High-quality transparent glass substrate is the basis for building TFT color liquid crystal panels. Special glass materials such as non-alkali glass are usually used to ensure high transparency and good electrochemical stability; 2. Liquid crystal material. Liquid crystal is the core material of TFT color liquid crystal panel, which is responsible for the optical properties of the display. There are many types of liquid crystal materials. Their molecular structure and arrangement determine the display effect and viewing angle of the display; 3. Thin film transistor, which is a micro switch that controls the arrangement of liquid crystal at each pixel, is usually made of silicon material. The design and manufacturing quality of TFT directly affects the display effect and resolution; 4. Color filter, color filter is used to generate red, green and blue primary colors, and is a key component to achieve color display. It is usually located on a glass substrate, and color resin is accurately added through photolithography technology; 5. Polarizer, a component used to control the polarization state of light passing through the liquid crystal layer, is usually installed on the outside of the glass substrate. Polarizer is essential for improving display quality and contrast; 6. Packaging materials, including sealants and adhesives used to seal liquid crystal materials, these materials need to have good optical clarity and chemical stability to prevent liquid crystal leakage or chemical deterioration; 7. Backlight unit, in non-self-luminous TFT color liquid crystal panels, the backlight unit is essential to provide a uniform light source, which usually includes components such as LED lights, light guide plates, diffusers and reflective films; 8. Conductive materials, such as indium tin oxide (ITO) are used to make transparent electrodes, which are key to controlling the direction and optical properties of liquid crystals.

[0057] It should be noted that in order to specify the subtle changes in the parameters of the above raw materials, examples of possible changes in some parameters of each parameter are given below. The glass substrate may have thickness changes, and the general thickness tolerance is within 0.05mm, and its flatness tolerance is within 10μm; liquid crystal materials mainly consider the changes in dielectric constant, viscosity and photoelectric properties, and the viscosity fluctuation range is usually ±5%, and the reaction time fluctuation of photoelectric properties may be within 1 millisecond; thin film transistors have differences in switching speed and threshold voltage, which are generally 10 nanoseconds and 0.2 volts respectively; color filters change in color saturation and transmittance, and the transmittance change is usually ±2%, and there are also changes in color difference (ΔE value). The change in ΔE value should be less than 1; the transmittance of the polarizer is generally within 1%, the polarization degree fluctuates within 0.1%, and the overall is maintained at 99.8%; there are differences in viscosity and curing time of packaging materials, the viscosity fluctuates within 10%, and the curing time fluctuates within 5%; the brightness fluctuation range of the backlight unit is 5%, and the color temperature fluctuation range is usually ±100K; the resistivity of the conductive material is ±10%, and the transparency is within 2%; although the above-mentioned numerical deviations are relatively small, since the TFT color LCD panel as a display device pursues a more perfect display effect, the stability of its raw material parameters is crucial, but it is unavoidable both in theory and in practice, so it can only be offset from the production parameters.

[0058] like Figure 1-Figure 10 As shown, the present invention provides a technical solution: a production management method and system based on TFT color liquid crystal panel detection, including detecting raw materials to obtain real-time raw material detection data.

[0059] Obtain various parameter items of the TFT color LCD panel production line and determine the adjustable parameter items of the equipment.

[0060] Get the inspection data of finished products and the product inspection data of finished products.

[0061] Create an adjustment model, which is used to analyze real-time detection data and output adjustment parameters corresponding to adjustable parameter items as parameter presets.

[0062] Methods for creating a reconciliation model include:

[0063] Step 1: Acquire various test data of historical raw materials as historical raw material data, the historical raw material data includes multiple data items, and each data item includes a test item corresponding to the historical raw material;

[0064] Step 2: Obtain parameter data of each adjustable parameter item when using historical raw materials for production and historical finished product data of finished products produced using corresponding historical raw materials;

[0065] Step 3: Using the historical raw material data item of one of the raw material items in the historical raw materials as a benchmark, determine the matching relationship between each detection item and each adjustable parameter item based on the matching method, and a single adjustable parameter item matches at least one detection item;

[0066] Step 4: Using the historical raw material data item of one of the raw material items in the historical raw materials as the basic data, a comparison chart of the detection items matching the same adjustable parameter item and the detection item data as the basic data is obtained based on the visualization method, and each comparison chart corresponds to the parameter data of the adjustable parameter item of the corresponding raw material item and the historical finished product data;

[0067] Step 5: Based on the convolutional neural network, the comparison graph, the parameter data of the adjustable parameter items associated with the comparison graph, and the historical finished product data are used as training data to train and obtain an adjustment model.

[0068] It should be noted that in order to specify the adjustable parameters and match the above content, the alignment and placement accuracy are adjusted during the production process, and high-precision placement equipment is used to adapt to slight changes in the substrate size. The pressure and temperature are adjusted during the lamination process to adapt to changes in the thickness and flatness of the substrate; injection volume adjustment: the injection volume and speed are adjusted according to changes in the viscosity of the liquid crystal material; curing temperature and time adjustment: the temperature and time of UV curing or thermal curing are adjusted to match the curing characteristics of the liquid crystal material; the voltage and current parameters in the drive circuit are adjusted to adapt to changes in the threshold voltage and switching speed of the thin film transistor, and the deposition thickness and process parameters are adjusted according to the electrical characteristics of the thin film transistor layer; the deposition amount and uniformity of the colorant are adjusted to adapt to changes in the color transmittance and color difference of the color filter, and the accuracy of the alignment equipment is improved to adapt to the tolerance of the filter size; according to the thickness of the polarizer Adjust the cutting equipment and alignment process according to the degree and transmittance, adjust the pressure and temperature during the pasting process to adapt to the physical properties of the polarizer; adjust the mixing ratio and curing conditions according to the viscosity and curing speed of the packaging material, adapt to the changes in the mechanical strength and hardness of the material, and adjust the pressure and curing time; adjust the current and driving strategy according to the luminous efficiency and color temperature fluctuations of the backlight unit to maintain the consistency of brightness and color temperature, adjust the layout of the backlight unit light guide plate and diffuser to meet the requirements of brightness uniformity; adjust the thickness or formula of the conductive material such as indium tin oxide layer to adapt to the changes in resistivity, optimize the coating process and post-processing steps to maintain the transparency standard; all of the above parameters can be manually adjusted on the production line. Traditionally, experienced workers are required to adjust them one by one to complete the optimization of the production line. Such operations must be repeated every time the raw materials are updated.

[0069] As a specific implementation method, Figure 2As shown, the parameter presets are fine-tuned to obtain the final parameters, which are then input into the TFT color LCD panel production line. The corresponding finished products are obtained based on the corresponding raw materials. After the corresponding finished products are inspected, the inspection data, final parameters and real-time raw material inspection data are input into the adjustment model for repeated learning to improve the model function and realize the self-iteration of the model.

[0070] As a specific implementation method, Figure 4-Figure 8 As shown, the matching methods include:

[0071] Step 1: Get the detection items and adjustable parameter items respectively;

[0072] Step 2: According to the historical raw material data items of the raw material items used as the benchmark, a benchmark matching relationship is established, in which each detection item has a matching relationship with each adjustable parameter item;

[0073] Step 3: Based on the data changes of each detection item of the remaining historical raw material data items and the data changes of the adjustable parameter items, eliminate the non-matching relationship between each detection item and each adjustable parameter item;

[0074] Step 4: After repeated N3 processing, the final matching relationship between each detection item and each adjustable parameter item is obtained.

[0075] It should be noted that the logic of the above method can be simply summarized as taking historical raw material data items of raw material items with sample one as the benchmark. These raw material data items are assumed to have six items, namely item 1 to item 6, and there are assumed to be three corresponding parameter items. Then, in sample one, each data item is matched with a parameter item, thereby building a matching basis. Subsequently, irrelevant parameters are proposed based on the matching changes of parameters. For example, in sample two, compared with the detection items of sample one, only item 1 has a numerical improvement. Among the parameter items, parameter two has been increased, parameter three has been decreased, and parameter one remains unchanged. Then it can be clearly seen that the change of item one will bring about changes in parameters two and three, and then it is determined that parameter one has a matching relationship with parameters two and three, which is for further confirmation of other detection items. Regarding the matching relationship with the parameters, in sample three, both item 1 and item 2 have changed, and parameter two has increased, parameter three has increased, and parameter one has remained unchanged. Without considering other factors, item 1 and item 2 are both unrelated to parameter one, but are related to parameter two and parameter three. In sample four, item 2 and item 4 have increased, parameter one and parameter three have changed, and both item 2 and item 4 are unrelated to parameter two. Then, by combining the judgments of the four samples, we can basically judge the matching relationship between item 1 and item 2. Sample 3 determines the range of item 2 and parameter two and parameter three. Sample four removes the matching relationship between item 2 and parameter two. Considering that the probability that the effects of different detection items on parameters can offset each other does not exist, item 2 is related to parameter three, and the correlation of item 4 has not yet been clarified.

[0076] As a specific implementation, the matching method also has an alternative solution which includes:

[0077] Step 1: Obtain the influence rate of each adjustable parameter item on each detection item of the color LCD panel;

[0078] Step 2: Remove the influence relationship between the adjustable parameter items and the detection items in the preset standard;

[0079] Step 3: Based on the influence rate of the remaining adjustable parameter items retained in the second step, the influence weight of each adjustable parameter item is assigned to establish a matching relationship with the detection item.

[0080] As a specific implementation method, Fig. 9 As shown in AD, the visualization methods include:

[0081] The first step: obtain the detection items matched by the adjustable parameter items according to the matching relationship, and construct an impact group for each detection item matched by the adjustable parameter item; the second step: obtain the detection item values ​​in the historical raw material data items of the raw material items corresponding to the basic data, and construct a reference range divided within a rectangular range; the third step: divide the variation range on the reference range in proportion according to the variation relationship between the values ​​of each detection item in the impact group and the values ​​of the detection item corresponding to the basic data; the fourth step: obtain a rectangular image containing the variation range and the reference range as a comparison chart.

[0082] As a specific implementation method, Fig. 9 As shown in AD, in the second step, a circular range is inscribed in the rectangular range, and a suitable figure is selected as the inscribed figure of the circular range according to the number of detection items matched by a single adjustable parameter item, and the variable range and the reference range are constructed based on the edges or corners of the inscribed figure.

[0083] It should be noted that the figure Fig. 9 As shown in Figures AD, there are fan-shaped areas of the same size. These areas are the reference ranges, and the remaining shaded ranges are the variable ranges. The area of ​​the variable range can be larger or smaller than the reference range. In a few cases, it will be the same as the reference range. The design of these polygons, variable ranges, and reference ranges can make the detection items of each parameter more obvious compared to the graph features, providing an image basis for subsequent deep convolution.

[0084] As a specific implementation method, as shown in the figure, the specific training method of the adjustment model in S5 includes: matrixing the image, dividing the comparison image according to the pixel size to obtain a matrix image; matrix assignment, assigning different values ​​to each variation range and reference range to obtain a numerical matrix image, and the variation range and reference range corresponding to the same detection item are assigned the same value; deep convolution, selecting a suitable convolution kernel for multi-layer convolution, and after the convolution is completed, striping the convolved numerical matrix into a numerical string; storing the numerical string to obtain an adjustment model, each numerical string corresponds to a specific numerical value of an adjustable parameter item.

[0085] It should be noted that there is no obvious difference between the deep convolution model used above and the establishment of conventional image recognition models on the market. The only difference is that this model does not need to analyze the color of the image, and the images processed are all grayscale images, which are less difficult to process. You can refer to mature image recognition models built using convolutional neural networks such as LeNet-5, AlexNet, VGGNet and GoogLeNet, and the principles are the same.

[0086] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. A production management method based on TFT color liquid crystal panel detection, characterized in that: include: Test raw materials to obtain real-time raw material test data; Obtain various parameters of the TFT color LCD panel production line and determine the adjustable parameters of the equipment; Acquire test data of a finished product, product test data of the finished product; Creating an adjustment model, wherein the adjustment model is used to analyze the real-time detection data and output adjustment parameters corresponding to the adjustable parameter items as parameter presets; The method for creating the adjustment model includes: S1: Acquire various test data of historical raw materials as historical raw material data, where the historical raw material data includes multiple data items, and each data item includes a test item corresponding to the historical raw material; S2: Obtaining parameter data of each adjustable parameter item when using historical raw materials for production and historical finished product data of finished products produced using corresponding historical raw materials; S3: Taking the historical raw material data item of one of the raw material items in the historical raw materials as a benchmark, determining the matching relationship between each detection item and each adjustable parameter item based on a matching method, and a single adjustable parameter item matches at least one detection item; S4: Using the historical raw material data item of one of the raw material items in the historical raw materials as the basic data, based on the visualization method, a comparison chart of the detection item data matching the same adjustable parameter item and the detection item data as the basic data is obtained, and each comparison chart corresponds to the parameter data of the adjustable parameter item of the corresponding raw material item and the historical finished product data; S5: Based on a convolutional neural network, the comparison graph and the parameter data of the adjustable parameter items associated with the comparison graph and the historical finished product data are used as training data to train and obtain an adjustment model; The visualization method comprises: M1: Obtain detection items matched by the adjustable parameter items according to the matching relationship, and construct an impact group for each detection item matched by the adjustable parameter item; M2: Obtain the value of the detection item in the historical raw material data item of the raw material item corresponding to the basic data, and construct a rectangular range to divide the reference range; M3: Divide the change range on the benchmark range in proportion to the change relationship between the values ​​of each test item in the impact group and the values ​​of the test items corresponding to the basic data; M4: Obtain a rectangular image containing the variation range and the reference range as a comparison image; In M2, a circular range is inscribed in the rectangular range, and a suitable figure is selected as the inscribed figure of the circular range according to the number of detection items matched by a single adjustable parameter item, and the variable range and the reference range are constructed based on the edges or corners of the inscribed figure.

2. A production management method based on TFT color liquid crystal panel detection according to claim 1, characterized in that: The parameter presets are fine-tuned to obtain final parameters, which are input into the TFT color liquid crystal panel production line. Corresponding finished products are obtained based on the corresponding raw materials. After the corresponding finished products are inspected, the inspection data, final parameters and real-time raw material inspection data are input into the adjustment model for repeated learning to improve the model function and realize self-iteration of the model.

3. The production management method based on TFT color liquid crystal panel detection according to claim 1, characterized in that: The matching method comprises: N1: Get the detection items and adjustable parameter items respectively; N2: According to the historical raw material data items of the raw material items used as the benchmark, a benchmark matching relationship is established, in which each detection item has a matching relationship with each adjustable parameter item; N3: based on the data changes of the various detection items of the remaining historical raw material data items and the data changes of the adjustable parameter items, eliminate the non-matching relationship between the various detection items and the various adjustable parameter items; N4: After repeated N3 processing, the final matching relationship between each detection item and each adjustable parameter item is obtained.

4. The production management method based on TFT color liquid crystal panel detection according to claim 1, characterized in that: The matching method comprises: Q1: Obtain the influence rate of each adjustable parameter item on each detection item of the color LCD panel; Q2: The influence relationship between the adjustable parameter items and the detection items of the removal influence rate in the preset standard; Q3: Based on the influence rate of the remaining adjustable parameter items retained in Q2, each adjustable parameter item is assigned an influence weight to establish a matching relationship with the detection item.

5. The production management method based on TFT color liquid crystal panel detection according to claim 1, characterized in that: The specific training method of the adjustment model in S5 includes: P1: Matrix image, divide the contrast image according to pixel size to obtain a matrix image; P2: Matrix assignment, assign different values ​​to each variation range and reference range to obtain a numerical matrix diagram. The variation range and reference range corresponding to the same test item have the same value assigned to them. P3: Depth convolution, select appropriate convolution kernels for multi-layer convolution, and after the convolution is completed, strip the convolved numerical matrix into a numerical string; P4: Store the value strings to obtain the adjustment model, where each value string corresponds to a specific value of an adjustable parameter item.

6. A production management system based on TFT color liquid crystal panel detection, characterized in that: A production management method based on TFT color liquid crystal panel detection as described in any one of claims 1 to 5 is used.

Images