Mask frame posture detection method and device

Abstract

The present application belongs to the technical field of semiconductor wafer transmission and detection, and particularly relates to a mask iron frame posture detection method and device. The method comprises the following steps: after the sealed box is opened, the transport mechanism carries the frame mask workpiece or frameless single mask workpiece to cross the horizontal detection light path of the bilateral laser beam pair component; four independent function detection sections are generated according to the dynamic calibration of geometric size parameters, the number of light path conduction edges and light path shielding edges of each section is counted to construct a basic feature vector, the workpiece posture and the workpiece type are distinguished through a preset multi-branch judgment logic; the continuous waveform quantization parameters are extracted and matched with the standard posture reference waveform, the posture matching confidence index is calculated, and the multi-level confidence rule is used for hierarchical processing. The present application is adaptive to multi-batch tolerances, uses continuous waveform mode recognition instead of simple on-off quantity judgment, has online self-calibration and fault tracing capabilities, and effectively suppresses accidental interference misjudgment.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor wafer transport and detection technology, specifically relating to a method and apparatus for detecting the posture of a mask frame. Background Technology

[0002] In semiconductor coating, photolithography, and display panel manufacturing processes, photomasks can be assembled inside a frame to form framed photomask workpieces for production line transfer and process handling. There are four standard orientations during the transfer of framed photomask workpieces: frame down and narrow edge forward, frame down and wide edge forward, frame up and narrow edge forward, and frame up and wide edge forward. Different orientations correspond to different execution logics of downstream processes such as flipping, position alignment, and parameter calibration. Therefore, accurate identification of the four orientations of the photomask is a necessary pre-processing step in the photomask processing procedure. Conventional wafer orientation detection solutions cannot meet the requirements for distinguishing the four orientations of this workpiece.

[0003] Existing workpiece transfer and detection solutions of the same type mostly use laser rangefinders, ordinary photoelectric switches or fixed threshold laser through-beam sensing structures. Such solutions have the following technical defects.

[0004] 1. The detection range adopts a static fixed division method, and the judgment threshold is a fixed parameter. When there are dimensional tolerances between different batches of framed mask workpieces, or when frameless single-piece mask workpieces or framed mask workpieces are placed off-center, the judgment threshold needs to be manually adjusted repeatedly on-site. The workload of equipment debugging and maintenance is large, and manual adjustment is prone to introducing errors. The consistency of detection between different batches of workpieces is poor.

[0005] 2. Binary logic judgment is based solely on the counting results of the optical path conduction edge and the optical path occlusion edge, only distinguishing whether the optical path is occluded or not; when there are working conditions such as slight deformation of the framed mask workpiece, workpiece eccentric placement, or slight offset of optical path installation, it is easy to produce attitude misjudgment, and the adaptability of working conditions is limited.

[0006] 3. The open-loop judgment logic without feedback is adopted. Dust blocking the light path, electromagnetic interference, and occasional light path blocking signals caused by mechanical vibration cannot be filtered out. If there is a deviation in the attitude judgment result, the abnormal workpiece will flow into the downstream photolithography process.

[0007] 4. The detection data storage and fault tracing functions are not configured. After the attitude misjudgment occurs, it is impossible to quickly retrieve the original signal data of this detection to locate the cause of the abnormality. The equipment fault troubleshooting cycle is long, which affects the continuous operation efficiency of the production line.

[0008] The photomask process imposes specific constraints on posture recognition accuracy, requiring precise differentiation of four posture types to match downstream process execution logic. This differs from conventional wafer inspection, which only distinguishes the presence or absence of a workpiece and its orientation. Posture judgment errors can lead to datum errors in downstream alignment and flipping processes, resulting in mask workpiece loss or batch wafer contamination. Therefore, this scenario places higher demands on the stability of the detection scheme and its anomaly protection capabilities. Based on the aforementioned deficiencies in existing technologies, this invention proposes a photomask frame posture detection method and apparatus. Summary of the Invention

[0009] The purpose of this invention is to overcome the above-mentioned defects of the prior art and provide a method and device for detecting the attitude of a mask iron frame. By cooperating with the dual-sided laser beam assembly and the conveying mechanism, and combining dynamic calibration, multi-branch judgment logic, waveform feature matching, multi-level confidence judgment and online self-calibration of the optical path, a high-precision and high-reliability non-contact detection of the attitude of the mask iron frame can be achieved.

[0010] In a first aspect, the present invention provides a method for detecting the posture of a mask frame, comprising: performing an opening operation on a sealed box carrying a framed mask workpiece or a frameless monolithic mask workpiece; driving a conveying mechanism to transport the framed mask workpiece or the frameless monolithic mask workpiece through a horizontal detection optical path formed by dual-sided laser beam-beaming components; retrieving the geometric dimensional parameters of the corresponding batch of framed mask workpieces or frameless monolithic mask workpieces for dynamic calibration calculation, and dividing the conveying path into multiple independent functional detection sections according to the calibration results; and collecting the optical path conduction edge and optical path obstruction edge output by the dual-sided laser beam-beaming components in each independent functional detection section. The algorithm generates a number of iterations to construct a basic feature vector for attitude determination. Based on this feature vector, it executes a pre-defined multi-branch decision logic to distinguish between framed mask workpieces and frameless single-piece mask workpieces. It extracts quantized waveform parameters of the continuous optical path waveform within the independent functional detection section corresponding to the mask body. It retrieves a pre-stored standard attitude reference waveform, performs numerical matching between the extracted quantized waveform parameters and the standard attitude reference waveform, and calculates the attitude matching confidence index based on the workpiece type determination result. Based on the attitude matching confidence index, it compares the results with pre-stored multi-level confidence judgment rules and outputs the corresponding attitude determination or alarm result.

[0011] Furthermore, the independent functional detection section includes four independent functional detection sections. The boundaries of the four independent functional detection sections are generated by dynamic calibration of the continuous optical path waveform corresponding to the framed mask workpiece or frameless monolithic mask workpiece of the current batch. The dynamic calibration process includes: acquiring the continuous optical path waveform corresponding to the framed mask workpiece or frameless monolithic mask workpiece of the current batch, performing first derivative analysis on the continuous optical path waveform, automatically identifying the optical path occlusion start and end points corresponding to the upper and lower edges of the mask and the edge of the mask body, automatically dividing the boundaries of the independent functional detection section according to the optical path occlusion start and end points, and synchronously updating the standard attitude reference waveform stored in the current batch. The functional division of the four independent functional detection sections is as follows: the first independent functional detection section is used to identify the lower edge of the mask, the second and third independent functional detection sections correspond to the mask body to acquire the continuous optical path waveform, and the fourth independent functional detection section is used to identify the upper edge of the mask.

[0012] Furthermore, the preset multi-branch determination logic is based on a basic feature vector constructed from four independent functional detection segments to distinguish four types of framed mask workpiece postures and frameless single-piece mask workpieces; wherein, the optical path conduction edge is the rising edge when the optical path signal switches from the occlusion state to the conduction state, and the optical path occlusion edge is the falling edge when the optical path signal switches from the conduction state to the occlusion state; the preset multi-branch determination logic includes: if the first independent functional detection segment detects a rising edge, then the framed mask workpiece is determined to be in a frame-down posture; wherein, if one side of the laser beam beam unit in the dual-sided laser beam beam assembly detects a rising edge while the other side does not detect a rising edge, it corresponds to a 0° placement posture; if both sides of the laser beam beam unit in the dual-sided laser beam beam assembly detect a rising edge, it corresponds to a 90° placement posture; if the first independent functional If no rising edge is detected in the detection section, and the second and third independent functional detection sections detect the optical path signal corresponding to the mask body, then the framed mask workpiece is determined to be in an orientation with the frame facing upwards. Specifically, if a single-sided laser beam unit in the dual-sided laser beam assembly detects a falling edge in the fourth independent functional detection section, while the other-sided laser beam unit does not detect a falling edge, this corresponds to a 0° placement orientation. If both sides of the laser beam unit in the dual-sided laser beam assembly detect a falling edge in the fourth independent functional detection section, this corresponds to a 90° placement orientation. If none of the four independent functional detection sections detect the optical path signal corresponding to the frame of the framed mask workpiece, and only the second and third independent functional detection sections detect the optical path signal corresponding to the mask body, then the currently detected workpiece is determined to be a frameless monolithic mask workpiece.

[0013] Furthermore, the quantized waveform parameters include the timing difference between the two-sided sensing signals, the proportion of continuous optical path occlusion, and the rate of change of the signal edge.

[0014] Furthermore, the attitude matching confidence index is obtained by comprehensively evaluating the timing difference matching degree of the bilateral sensor signals, the matching degree of the proportion of continuous occlusion of the optical path, and the matching degree of the signal edge change rate, combined with the corresponding weight coefficients automatically trained and generated during the current batch dynamic calibration process. It is used to characterize the degree of matching between the quantized waveform parameters and the standard attitude reference waveform; the weight coefficients are adaptively adjusted according to the geometric dimension parameters of the corresponding batch.

[0015] Furthermore, the multi-level confidence determination rule performs hierarchical closed-loop processing based on the attitude matching confidence index; the hierarchical closed-loop processing is used to filter dust, electromagnetic interference, and occasional light path obstruction interference caused by mechanism vibration.

[0016] Furthermore, it also includes an online self-calibration step for the optical path: when the equipment is in an unloaded state, it runs to the reference positions at both ends of the lifting stroke set along the conveying path, and collects the continuous optical path waveforms in the unobstructed state and the completely blocked state respectively, calculates the optical path attenuation difference, and dynamically corrects the judgment level thresholds corresponding to optical path conduction and optical path blockage, so as to compensate for the signal deviation caused by laser aging, optical window dust accumulation, and installation position offset.

[0017] Furthermore, it also includes a full-process fault tracing and storage step: during each detection process, the continuous optical path waveform, boundary parameters of independent functional detection sections, mechanism motion speed curves, and real-time laser light intensity data corresponding to this detection are stored synchronously and bound to the task number; historical continuous optical path waveforms are retrieved according to the task number to reproduce the entire scanning process, and the deviation sections between the collected continuous optical path waveform and the standard attitude reference waveform are identified.

[0018] Furthermore, the dual-sided laser beam-emitting assembly consists of two sets of red laser emitting and receiving units arranged symmetrically on the left and right sides; the spot diameter, response time, scanning period, effective sampling frequency, and moving speed corresponding to the conveying path of the dual-sided laser beam-emitting assembly are adapted to each other, and are used to collect the continuous optical path waveform of the frame mask workpiece frame as it passes through the optical path along the conveying path.

[0019] A second aspect of the present invention provides a mask frame posture detection device, comprising: a sealed box opening mechanism for opening a sealed box carrying a framed mask workpiece or a frameless single-piece mask workpiece; a conveying mechanism for transporting the framed mask workpiece or the frameless single-piece mask workpiece along a conveying path; a dual-sided laser beam assembly symmetrically arranged to form a horizontal detection optical path; and a control processing unit configured to: retrieve the geometric dimensional parameters of the corresponding batch of framed mask workpieces or frameless single-piece mask workpieces for dynamic calibration calculation, and divide the conveying path into multiple independent functional detection sections according to the calibration results; and collect dual-sided laser beams in each independent functional detection section. The number of occurrences of the optical path conduction edge and optical path occlusion edge output by the emitting component is used to construct a basic feature vector for attitude determination. Based on the basic feature vector, a preset multi-branch determination logic is executed to distinguish the workpiece type as a framed mask workpiece or a frameless single-piece mask workpiece. Quantized waveform parameters of the continuous optical path waveform within the independent functional detection section corresponding to the mask body are extracted. A pre-stored standard attitude reference waveform is retrieved, and the extracted quantized waveform parameters are numerically matched with the standard attitude reference waveform. Combined with the workpiece type determination result, an attitude matching confidence index is calculated. Based on the attitude matching confidence index, a comparison is performed using pre-stored multi-level confidence determination rules to output the corresponding attitude determination or alarm result.

[0020] Compared with the prior art, the mask plate iron frame posture detection method and device provided by the present invention has the following significant advantages.

[0021] 1. Simplified hardware structure and controllable hardware investment costs. This solution only requires a dual-sided laser beam assembly and a conveying mechanism to complete attitude detection, without the need for a high-resolution industrial camera, a dedicated supplementary lighting source, or an independent image processing unit; the laser beam assembly has good resistance to ambient light interference and is suitable for use in semiconductor cleanrooms.

[0022] 2. The inspection and conveying processes are executed synchronously, matching the continuous operation cycle of the production line. Signal acquisition and attitude calculation are completed simultaneously as the conveying mechanism carries the workpiece along the conveying path, eliminating the need for a separate inspection stop and preventing additional time commitment to the production line.

[0023] 3. Wide workpiece recognition range and automatic switching of detection modes. This invention can recognize four types of framed mask workpiece postures, while also being compatible with frameless single-piece mask workpiece detection; the same hardware system automatically distinguishes workpiece types based on algorithms, eliminating the need for manual switching of detection configurations and adapting to various process detection scenarios.

[0024] 4. Supports dynamic calibration, adapting to differences in the geometric dimensions of different batches of workpieces. The boundaries of the four independent functional detection sections are not statically fixed, but are automatically generated through a dynamic calibration process based on the geometric characteristics of the current batch of workpieces. After changing the workpiece batch, the section boundaries and judgment parameters can be automatically updated, reducing the frequency of manual on-site threshold adjustments and lowering the workload of equipment debugging and maintenance.

[0025] 5. Dual-layer decision logic enhances adaptability to complex working conditions. This invention integrates a two-layer recognition mechanism: preset multi-branch decision logic and waveform feature matching. Unlike binary decision methods that rely solely on edge counting, this invention completes multi-dimensional quantized waveform parameter matching based on continuous optical path waveforms. It also exhibits better fault tolerance for working conditions such as workpiece deformation with frame masks and workpiece eccentric placement, reducing attitude misjudgment.

[0026] 6. Implement multi-level confidence level closed-loop processing to suppress recognition bias caused by occasional interference. By calculating the attitude matching confidence index and combining it with multi-level confidence judgment rules to implement hierarchical closed-loop processing, the system can filter out occasional light path obstruction signals caused by dust, electromagnetic interference, and mechanism vibration, reducing the probability of abnormal judgment results output due to instantaneous interference.

[0027] 7. Equipped with online optical path self-calibration function to compensate for optical path hardware signal drift. The equipment can run to the reference positions at both ends of the lifting stroke in no-load state, and collect continuous optical path waveforms under unobstructed and completely blocked states respectively. It calculates the optical path attenuation difference and dynamically corrects the judgment level thresholds corresponding to optical path conduction and optical path blockage, compensating for signal deviations caused by laser aging, window dust accumulation, and installation misalignment, and reducing manual periodic calibration operations.

[0028] 8. Equipped with full-process fault tracing and storage function, shortening the time for anomaly troubleshooting. Each test synchronously stores continuous optical path waveforms, boundary parameters of independent function detection sections, mechanism motion speed curves, and real-time laser light intensity data, and binds them to task numbers; the historical continuous optical path waveforms can be retrieved and the scanning process reproduced according to the task number, and the deviation sections between the waveform and the standard attitude reference waveform are automatically marked, shortening the fault troubleshooting cycle after attitude misjudgment and ensuring continuous operation of the production line.

[0029] 9. Non-contact laser inspection is adopted to protect the cleanliness of the workpiece. During the inspection process, the dual-sided laser beam assembly does not make physical contact with the surface of the framed mask workpiece or the frameless single-piece mask workpiece, which can avoid the risk of surface contamination and physical scratches caused by contact inspection, and meet the requirements of semiconductor clean process. Attached Figure Description

[0030] The accompanying drawings, which are provided to further illustrate this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0031] Figure 1 This is the overall flowchart of the present invention.

[0032] Figure 2 This is a structural block diagram of the mask plate iron frame attitude detection device of the present invention.

[0033] Figure 3 This is a schematic diagram of the mask plate iron frame posture detection device of the present invention.

[0034] Figure 4 This is a schematic diagram of the structure of the dual-sided laser beam assembly of the present invention.

[0035] Figure 5 This is a schematic diagram of the framed mask workpiece with its frame facing down at 0°, according to the present invention.

[0036] Figure 6 This is a schematic diagram of the framed mask workpiece with its frame facing downwards at 90°, according to the present invention.

[0037] Figure 7 This is a schematic diagram of the framed mask workpiece with the frame facing upwards at 0°, according to the present invention.

[0038] Figure 8 This is a schematic diagram of the framed mask workpiece with its frame facing upwards at 90°.

[0039] Reference numerals: 1. Sealed box opening mechanism; 2. Conveying mechanism; 3. Control and processing unit; 4. Double-sided laser beam assembly; 5. Framed mask workpiece. Detailed Implementation

[0040] The embodiments of this application will now be described in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. Furthermore, the following embodiments and features can be combined with each other unless otherwise specified. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0041] Example 1 This embodiment provides a method for detecting the attitude of a mask frame. The overall solution can be deployed inside a standard loadport unpacking device in a semiconductor production line. After the sealed box (Cassette / FOUP) containing framed or frameless monolithic mask workpieces is opened in a semiconductor cleanroom, the drive conveyor 2 transports the framed or frameless monolithic mask workpieces through a fixed horizontal detection optical path formed by paired laser beam-beaming components. The attitude is determined based on the intelligent control logic of the entire chain. Specifically, it includes the following core steps; please refer to the appendix. Figure 1 .

[0042] S1. Perform an opening operation on the sealed box carrying the framed mask workpiece or the frameless single-piece mask workpiece, and drive the conveyor mechanism to transport the framed mask workpiece or the frameless single-piece mask workpiece through the horizontal detection optical path formed by the double-sided laser beam assembly.

[0043] This step is a pre-processing procedure for mask attitude detection, which completes the exposure and transport of framed or frameless single-piece mask workpieces. It includes three continuous actions: opening the cover, vertical transport of the workpiece, and optical path passage. The technical details of each action are as follows.

[0044] The sealing box opening operation specifically includes: the framed mask workpiece or frameless single-piece mask workpiece to be tested is pre-received inside the sealing box cavity. The sealing box is used to isolate dust and stray light pollution in the semiconductor clean environment; the sealing box opening mechanism 1 performs the opening action, removes the sealing box cover, and exposes the workpiece to be tested inside to be fully exposed within the movement stroke of the transport path, providing unobstructed space for laser optical path detection.

[0045] The vertical transport drive for the workpiece specifically includes: the conveying mechanism 2, which serves as the workpiece carrier and transport carrier, receives the framed mask workpiece or the frameless single-piece mask workpiece after the cover is opened, and moves up and down in a uniform linear motion along the preset transport path; the moving speed of the mechanism is matched and set according to the preset detection accuracy parameters, and the movement speed is stable and controllable, ensuring that the waveform acquisition is uniform and complete during the laser sampling process.

[0046] The process involves passing through a horizontal detection optical path, specifically: dual-sided laser beams emitted horizontally by symmetrically arranged dual-sided laser beam-emitting components 4 on both sides of the equipment, forming a horizontal detection optical path at a fixed height; a conveying mechanism 2 drives the framed mask workpiece or frameless single-piece mask workpiece to move vertically as a whole, with the workpiece body completely traversing the horizontal laser optical path. The workpiece frame and the main body of the mask will periodically block and conduct the laser optical path during the movement process. The dual-sided laser beam-emitting components 4 synchronously collect the on / off signals of the entire optical path, generating the original continuous optical path waveform, providing the original signal data source for subsequent calibration, feature extraction, and attitude determination.

[0047] S2. Retrieve the geometric dimension parameters of the corresponding batch of framed mask workpieces or frameless single-piece mask workpieces for dynamic calibration calculation, and divide the conveying path into multiple independent functional detection sections based on the calibration results.

[0048] In this step, the independent functional detection section includes four independent functional detection sections. The boundaries of the four independent functional detection sections are dynamically calibrated by the continuous optical path waveform corresponding to the current batch of framed mask workpieces or frameless monolithic mask workpieces. The dynamic calibration process includes: acquiring the continuous optical path waveform corresponding to the current batch of framed mask workpieces or frameless monolithic mask workpieces, performing first-order derivative analysis on the continuous optical path waveform, automatically identifying the optical path occlusion start and end points corresponding to the upper and lower edges of the mask and the edge of the mask body, automatically dividing the boundaries of the independent functional detection sections according to the optical path occlusion start and end points, and synchronously updating the standard attitude reference waveform stored in the current batch. The functional division of the four independent functional detection sections is as follows: the first independent functional detection section is used to identify the lower edge of the mask, the second and third independent functional detection sections correspond to the mask body to acquire the continuous optical path waveform, and the fourth independent functional detection section is used to identify the upper edge of the mask.

[0049] In this step, the present invention sets four independent functional detection sections. These four independent functional detection sections do not adopt a static fixed division method. Instead, they can be automatically generated through a dynamic calibration process based on the geometric dimension parameters of the framed mask workpiece or frameless single-piece mask workpiece to be tested in the current batch. This adapts to the dimensional tolerances of framed mask workpieces in different batches and the offset of the mask body placement position.

[0050] The dynamic calibration process can be started in the following scenarios: (1) When the batch of workpieces to be tested is changed, and the workpiece is a new model of framed mask workpiece or frameless single-piece mask workpiece, the operator can manually start the calibration through the human-machine interface; (2) When the control processing unit 3 detects that the number of consecutive attitude misjudgments exceeds the preset threshold, the equipment automatically triggers recalibration; (3) The equipment performs preventive dynamic calibration according to the preset cycle, which can be set to once a week.

[0051] The dynamic calibration execution process includes: (1) Reference data acquisition steps: The conveying mechanism 2 carries the current batch of framed mask workpieces or frameless single-piece mask workpieces, and completes the lifting action at a constant speed along the conveying path. The dual-sided laser beam assembly 4 acquires the complete continuous optical path waveform throughout the entire process; (2) Waveform feature recognition steps: The control processing unit 3 performs first-order derivative calculation analysis on the acquired continuous optical path waveform to identify the optical path blocking start and end points corresponding to the upper and lower edges of the framed mask workpiece, and the optical path signal start and end points corresponding to the hollow area of ​​the mask body; (3) Independent function detection section boundary division steps: Based on the geometric dimension parameters of the framed mask workpiece, including the inflection point of the edge thickness change and the boundary parameters of the wafer hollow area, the complete waveform period is divided into four independent function detection sections; among them, the first independent function detection The section corresponds to the workpiece posture start feature acquisition area, which is used to identify the lower edge of the mask. The second independent function detection section and the third independent function detection section are the core judgment areas of the workpiece posture, which correspond to the main body of the mask to acquire continuous optical path waveforms. The fourth independent function detection section corresponds to the workpiece posture end feature acquisition area, which is used to identify the upper edge of the mask. (4) Batch parameter update and storage steps: After the dynamic calibration is completed, the control processing unit 3 updates the boundary parameters of the independent function detection section corresponding to the current batch, the standard posture reference waveforms corresponding to various postures, the judgment thresholds of each dimension and the weight coefficients based on the waveforms acquired this time. The above parameters are associated with the corresponding batch identifier and stored in the equipment recipe storage library. When the same batch of workpieces is subsequently tested, the control processing unit 3 directly calls the calibration parameters of the batch to complete the calculation, without having to repeat the calibration process.

[0052] It should be noted that when the production line switches between batches of framed mask workpieces with different size tolerances, the equipment can automatically update the division parameters of the four independent functional detection sections through the calibration process, reducing the frequency of manual on-site debugging and judgment threshold operations, and reducing the workload of equipment debugging and daily operation and maintenance. This adaptive calibration structure is the core technical feature of this invention that distinguishes it from existing detection schemes that use fixed detection sections and fixed judgment thresholds.

[0053] Taking the calibration of a new batch as an example: the operator issues a calibration start command on the equipment's human-machine interface; the conveyor mechanism 2 carries the current batch of framed mask workpieces or frameless single-piece mask workpieces and moves up and down at a constant speed along the conveyor path; the dual-sided laser beam assembly 4 collects complete and continuous optical path waveforms; the control processing unit 3 performs first-order derivative analysis on the waveforms, locates the start and end points of optical path occlusion corresponding to the upper and lower edges of the framed mask workpiece and the wafer cutout area, divides the identified points into four independent functional detection segments, and records the timing parameters corresponding to the boundaries of each segment; then, based on the waveforms collected this time, it updates the standard attitude reference waveforms corresponding to the four attitudes, various judgment thresholds, and dynamic weight coefficients, and binds all parameters to the current batch number and stores them in the formula storage library; during subsequent testing of this batch of workpieces, the equipment directly retrieves the stored calibration parameters to perform calculations.

[0054] S3. Collect the number of times the optical path conduction edge and optical path occlusion edge occur in each independent functional detection section, and construct the basic feature vector for attitude determination.

[0055] In this step, the dual-sided laser beam assembly 4 consists of two sets of red laser emitting and receiving units arranged symmetrically on the left and right sides. For example, the red laser emitting and receiving unit arranged on the right side of the device is referred to as the first red laser emitting and receiving unit, and the red laser emitting and receiving unit arranged on the left side of the device is referred to as the second red laser emitting and receiving unit.

[0056] The optical path conduction edge corresponds to the rising edge when the optical path signal switches from the blocked state to the conduction state, and the optical path blockage edge corresponds to the falling edge when the optical path signal switches from the conduction state to the blockage state.

[0057] Within each independent functional detection segment, control processing unit 3 counts the number of rising and falling edges corresponding to the first and second red laser emitting and receiving units, and combines them to generate an eight-dimensional basic feature vector. This vector is composed of four independent functional detection segments, two sets of red laser emitting and receiving units, and two types of edge signals, serving as the basic feature vector for executing the preset multi-branch decision logic. In this embodiment, the statistical results of control processing unit 3 are combined to form an eight-dimensional basic feature vector, which consists of three dimensions: four independent functional detection segments, two sets of red laser emitting and receiving units, and two types of optical path signals (optical path conduction edge and optical path occlusion edge). This eight-dimensional basic feature vector is the feature basis for workpiece posture determination.

[0058] Different types and postures of workpieces have differentiated edge statistical characteristics in each independent functional detection section. The signal distribution of each type of workpiece is as follows.

[0059] 1. Frameless monolithic mask workpiece: In the first and fourth independent functional detection sections, the number of rising and falling edges detected by the first and second red laser emitting and receiving units is zero; in the second independent functional detection section, the number of rising edges of both units is not less than 1, and in the third independent functional detection section, the number of falling edges of both units is not less than 1. The optical path signal is generated by the edge of the mask body passing through the optical path. A rising edge is generated when the mask body passes through the second independent functional detection section, and a falling edge is generated when it passes through the third independent functional detection section. The triggering timing of the optical path signals of the two sets of units tends to be synchronized.

[0060] 2. Framed mask workpiece 5, frame facing down (0° placement, narrow side facing forward, see attached document). Figure 5In the first independent functional detection section, the first red laser emitting and receiving unit detects one rising edge, while the second red laser emitting and receiving unit does not detect a rising edge; there are no additional edge signals in other sections; the second and third independent functional detection sections can collect signals of double-sided optical path occlusion. During the workpiece movement, the lower edge of the framed mask workpiece enters the first independent functional detection section first, blocking the optical path of the first red laser emitting and receiving unit and triggering a single rising edge; when the workpiece passes through the second and third independent functional detection sections, it simultaneously blocks the optical paths of two sets of units, and then completely leaves the fourth independent functional detection section.

[0061] 3. Framed mask workpiece 5, frame facing down (90° orientation, wide side facing forward, see attached document). Figure 6 Within the first independent functional detection section, both the first red laser emitting and receiving unit and the second red laser emitting and receiving unit detect at least one rising edge. In this orientation, the lower border of the framed mask workpiece has a larger lateral dimension. When the workpiece passes through the first independent functional detection section, the two sets of units generate rising edge signals sequentially or synchronously.

[0062] 4. Framed mask workpiece 5, frame facing upwards (0° placement, narrow side facing forward, see attached document). Figure 7 In the first independent functional detection section, neither the first nor the second red laser emitting and receiving unit detected an edge signal; the second and third independent functional detection sections were able to acquire the optical path signal corresponding to the mask body; in the fourth independent functional detection section, the first red laser emitting and receiving unit detected one falling edge, while the second red laser emitting and receiving unit did not detect a falling edge. During the workpiece transport process, the mask body passes through the optical path first, and the upper frame of the framed mask workpiece passes through the fourth independent functional detection section last, blocking the optical path of the first red laser emitting and receiving unit and generating a single falling edge.

[0063] 5. Framed mask workpiece 5, frame facing upwards (90° orientation, wide side facing forward, see attached document). Figure 8 In the first independent functional detection section, the first and second red laser emitting and receiving units did not detect edge signals; the second and third independent functional detection sections were able to acquire the optical path signals corresponding to the mask body; in the fourth independent functional detection section, each of the two sets of units detected one falling edge. In this orientation, the horizontal dimension of the upper frame of the framed mask workpiece is larger. When the workpiece passes through the fourth independent functional detection section, the optical paths of both sets of units are blocked, generating a falling edge signal each.

[0064] S4. Execute preset multi-branch decision logic based on basic feature vectors to distinguish the workpiece type as a framed mask workpiece or a frameless single-piece mask workpiece.

[0065] In this step, the control processing unit 3 runs a preset multi-branch decision logic based on the constructed basic feature vectors, simultaneously completing the identification and differentiation of workpiece type and workpiece posture. This preset multi-branch decision logic uses the basic feature vectors generated by four independent functional detection sections as the basis for judgment, and can distinguish between four types of framed mask workpiece postures and frameless single-piece mask workpieces. All judgment conditions are executed according to the optical path signal feature matrix.

[0066] After completing the construction of the basic feature vector, the control processing unit 3 prioritizes identifying whether an optical path conduction edge is generated in the first independent functional detection segment, and performs judgment in three categories.

[0067] 1. When the first independent functional detection section detects the optical path conduction edge, it is determined that the workpiece to be tested is a framed mask workpiece with the frame facing down; when the first red laser emitting and receiving unit detects the optical path conduction edge and the second red laser emitting and receiving unit does not detect the optical path conduction edge, it corresponds to a 0° placement posture; when both the first red laser emitting and receiving unit and the second red laser emitting and receiving unit detect the optical path conduction edge, it corresponds to a 90° placement posture.

[0068] 2. When the first independent functional detection section does not detect the optical path conduction edge, and the second and third independent functional detection sections can acquire the optical path signal corresponding to the mask body, the workpiece to be tested is determined to be a framed mask workpiece with the frame facing upwards; the first red laser emitting and receiving unit detects the optical path blocking edge in the fourth independent functional detection section, while the second red laser emitting and receiving unit does not detect the optical path blocking edge, corresponding to a 0° placement posture; both the first and second red laser emitting and receiving units detect the optical path blocking edge in the fourth independent functional detection section, corresponding to a 90° placement posture.

[0069] 3. When the four independent functional detection sections fail to collect the optical path signal corresponding to the frame of the mask workpiece, and the second and third independent functional detection sections have the optical path signal corresponding to the mask body, the current workpiece to be tested is determined to be a frameless single-piece mask workpiece.

[0070] The workpiece type differentiation rules are as follows: If there is an optical path conduction edge or optical path occlusion edge corresponding to the frame of a masked workpiece in the first or fourth independent functional detection section, the workpiece is determined to be a framed masked workpiece, and the workpiece posture is identified based on the optical path signal; if there is no optical path signal corresponding to the frame of a masked workpiece in any of the four independent functional detection sections, and there is an optical path signal at the edge of the mask body in the second and third independent functional detection sections, the workpiece is determined to be a frameless single-piece masked workpiece. The entire device uses the same hardware structure and automatically completes the identification and switching between the two types of workpieces based on the built-in algorithm.

[0071] S5. Extract the quantized waveform parameters of the continuous optical path waveform within the independent functional detection section corresponding to the mask body.

[0072] The quantized waveform parameters include the timing difference of the bilateral sensing signals, the proportion of continuous optical path occlusion, and the signal edge change rate. The timing difference of the bilateral sensing signals is a spatial distance parameter calculated from the rising edge trigger time difference between the first and second red laser emitting and receiving units, which can be used to distinguish between narrow-side-forward and wide-side-forward workpiece conditions. The proportion of continuous optical path occlusion is the ratio of the cumulative duration of the optical path conduction state of a single red laser emitting and receiving unit within its corresponding independent functional detection segment to the total duration of the segment, which can help distinguish between framed mask workpieces with the frame facing down and those with the frame facing up. The signal edge change rate is the signal change rate at the rising and falling edge positions, which can be used to distinguish between pure iron frame occlusion and combined occlusion of the iron frame and the mask body; the waveform edge steepness of the two types of conditions shows a identifiable difference.

[0073] This invention optimizes the traditional binary decision logic that relies solely on edge counting into a pattern recognition logic based on continuous waveforms by extracting quantized waveform parameters. This can adapt to unconventional working conditions such as deformation of framed mask workpieces and placement eccentricity, thereby improving the adaptability and stability of detection conditions.

[0074] S6. Retrieve the pre-stored standard attitude reference waveform, perform numerical matching between the extracted quantized waveform parameters and the standard attitude reference waveform, and calculate the attitude matching confidence index based on the workpiece type determination result.

[0075] The control processing unit 3 retrieves the pre-stored standard attitude reference waveform, performs numerical matching calculations between the real-time extracted quantized waveform parameters and the standard attitude reference waveform, and calculates the attitude matching confidence index based on the initial judgment result of the workpiece type. The attitude matching confidence index is comprehensively evaluated by combining the matching degree of the timing difference between the two-sided sensor signals, the matching degree of the proportion of continuous optical path occlusion, and the matching degree of the signal edge change rate, along with weighting coefficients generated adaptively during the dynamic calibration process of the current batch of workpieces. It is used to characterize the degree of matching between the real-time waveform parameters and the standard attitude reference waveform. The weighting coefficients can be adaptively adjusted according to the geometric dimensions of different workpiece batches. The quantitative evaluation of attitude matching reliability is completed through a weighted fusion method of multi-dimensional matching degree and corresponding weighting coefficients.

[0076] During dynamic calibration, a mapping relationship is established based on the geometric dimensions of the mask body and the width of the iron frame of the framed mask workpiece: the larger the width of the iron frame of the framed mask workpiece, the higher the weight corresponding to the matching degree of the continuous occlusion ratio of the optical path; the higher the operating speed of the vertical lifting conveyor mechanism 2, the higher the weight corresponding to the matching degree of the signal edge change rate; when the optical path installation layout deviation of the dual-sided laser beam assembly 4 is large, the weight corresponding to the matching degree of the timing difference of the dual-sided sensing signals is increased; the sum of the values ​​of all weights is normalized to 1.

[0077] S7. Based on the attitude matching confidence index, compare the results using pre-stored multi-level confidence judgment rules and output the corresponding attitude judgment or alarm results.

[0078] In this step, the attitude matching confidence index is compared with the pre-stored multi-level confidence judgment rules; when the attitude matching confidence index is higher than the preset high threshold, the workpiece attitude judgment result is output; when the attitude matching confidence index is between the preset high threshold and the preset low threshold, the movement speed is reduced and the continuous optical path waveform is re-acquired to update the quantized waveform parameters; when the attitude matching confidence index is lower than the preset low threshold, the equipment fault alarm result is output.

[0079] The control processing unit 3 compares the calculated attitude matching confidence index with the pre-configured multi-level confidence judgment rules, and executes the corresponding hierarchical closed-loop processing procedure according to the index range to filter out occasional light path detection interference caused by dust, electromagnetic interference, and mechanism vibration, thereby improving the stability of attitude judgment.

[0080] When the attitude matching confidence index is higher than the preset high threshold (such as 90%), the attitude recognition result is determined to be valid, the workpiece attitude determination result is directly output, and the workpiece enters the subsequent semiconductor process.

[0081] When the attitude matching confidence index is between the preset low threshold and the preset high threshold (e.g., 70% to 90%), the equipment automatically reduces the moving speed of the workpiece conveyor (e.g., reduces the lifting speed to 1 / 2 of the original speed), re-acquires the continuous optical path waveform and updates the quantization waveform parameters, and completes the secondary attitude determination by increasing the sampling density, thereby reducing the impact of instantaneous interference on the recognition results.

[0082] When the attitude matching confidence index falls below a preset low threshold (e.g., 70%), the equipment outputs an alarm command and suspends the detection process, awaiting manual verification and confirmation. This tiered handling method can avoid attitude judgment deviations caused by occasional working conditions such as floating particles, transient electromagnetic interference, and slight mechanical vibrations, reducing workpiece process abnormality transport problems caused by attitude recognition anomalies.

[0083] This invention also features a full lifecycle operation and maintenance support mechanism, which optimizes the long-term operational stability of the equipment and reduces the frequency of manual maintenance by combining online self-calibration of the optical path with full-process fault tracing and storage.

[0084] The specific implementation steps of online self-calibration of the optical path include: when the equipment is in an unloaded state, it runs to the reference positions at both ends of the lifting stroke set along the conveying path, and continuously collects the optical path waveforms under unobstructed and fully blocked states, respectively, and obtains the full-amplitude voltage and the blocked voltage of the optical path, and calculates the actual attenuation difference of the optical path. At the same time, based on the ratio between the real-time attenuation difference and the factory standard attenuation difference, the preset judgment level threshold is linearly corrected to achieve adaptive calibration of the optical path judgment benchmark.

[0085] The specific calibration logic is as follows: Define the full-scale optical path voltage under unobstructed conditions as Vfull, and the optical path voltage under completely obstructed conditions as Vdark; the formula for calculating the real-time optical path attenuation difference is: The equipment internally stores the standard attenuation difference ΔVstd obtained from the factory calibration, and uses this standard value as a benchmark to complete the threshold ratio correction; the threshold correction formula is: In the formula: Vth_old is the system's original preset threshold for optical path conduction and optical path occlusion determination; Vth_new is the threshold for determination used after calibration. When laser aging, dust accumulation in the optical window, or installation position misalignment causes overall signal attenuation shift in the optical path, the real-time attenuation difference will deviate from the factory standard deviation. The system synchronously fine-tunes the determination threshold through the above proportional correction method to offset the detection deviation caused by long-term optical path drift and maintain stable accuracy in optical path conduction and occlusion determination.

[0086] This online self-calibration process for the optical path can be executed automatically during equipment idle periods, or it can be triggered manually or automatically when continuous signal anomalies are detected. The calibration process relies on the reference positions at both ends of the inherent stroke of the lifting mechanism to complete the acquisition of reference waveforms, eliminating the need for additional calibration structures and independent calibration procedures. This ensures long-term maintenance of optical path signal detection accuracy and suppresses signal drift caused by long-term equipment operation.

[0087] The specific implementation steps of full-process fault tracing and storage include: during each test, the equipment synchronously stores the continuous optical path waveform, boundary parameters of independent functional test sections, mechanism motion speed curve, and real-time laser light intensity data corresponding to this test, and establishes a binding association between all stored data and task number.

[0088] This invention can also retrieve historical continuous optical path waveforms according to the task number, reproduce the entire scanning and detection process, and identify the deviation section between the collected continuous optical path waveform and the standard attitude reference waveform, so as to realize the traceability of the detection process and the location of abnormal features, and provide data support for equipment anomaly investigation and operating condition analysis.

[0089] This invention also includes matching design of relevant parameters for the dual-sided laser beam-beam assembly 4 to ensure the integrity and effectiveness of optical path waveform acquisition. The dual-sided laser beam-beam assembly 4 comprises symmetrically arranged red laser emitting and receiving units, featuring strong optical path directivity and small spot size, making it suitable for detecting minute structural features. The device uses a fixed-band semiconductor red laser, matched to the detector's response characteristics, enabling stable acquisition of optical path change signals; the laser's operating level meets the safety requirements for cleanroom use, making it suitable for semiconductor manufacturing environments.

[0090] The spot diameter, response time, controller scanning cycle, effective sampling frequency, and corresponding movement speed of the conveying path are matched to the dual-sided laser beam assembly 4 to acquire continuous optical path waveforms as the framed mask workpiece edge traverses the optical path along the conveying path. Stable sampling resolution is achieved through parameter matching, enabling high-density data sampling of different thickness features, such as framed mask workpiece edges and frameless single-piece mask workpieces. This ensures sufficient waveform sampling points, guaranteeing that the continuous optical path waveform objectively reflects the actual structural characteristics of the workpiece and meets the needs for distinguishing various workpiece postures. Simultaneously, a reasonable timing margin is configured to avoid sampling interference caused by signal transition jitter, improving the reliability of continuous optical path waveform acquisition.

[0091] Regarding the design of theoretical matching for the judgment threshold, this invention, based on the actual structural dimensions of the workpiece, determines the occlusion width range corresponding to each posture through theoretical calculation, thereby constructing a judgment threshold system adapted to the specifications of semiconductor mask workpieces. The occlusion width is calculated by converting the lifting speed of the mechanism corresponding to the conveying path with the duration of optical path occlusion. Different postures and different workpiece types have different distribution ranges of optical path occlusion width, which can be used as a basis for posture recognition and workpiece type differentiation.

[0092] Based on the above-mentioned size distribution pattern, the present invention also configures multiple fixed judgment thresholds and dynamic calibration thresholds. The various thresholds work together to adapt to the size specifications of conventional workpieces and the size deviations of different production batches, and complete the distinction and judgment of framed mask workpieces in 0° and 90° placement postures as well as frameless single-piece mask workpieces.

[0093] Regarding environmental interference resistance and anomaly protection design, this invention is equipped with an anti-interference structure and anomaly handling mechanism adapted to the working conditions of semiconductor cleanrooms. The dual-sided laser beam assembly 4 is arranged inside the sealed mechanism, forming a physical light-shielding barrier with the surrounding structure to avoid interference from stray light from the environment on the laser detection optical path and ensure the stability of the optical detection environment.

[0094] In this embodiment, the equipment can use a high-rigidity bracket and guide structure to install the dual-sided laser beam assembly 4, reducing the relative displacement of the components caused by equipment operation vibration and ensuring good repeatability of the detection trajectory along the conveying path. Simultaneously, a multi-level anomaly identification logic is set up to identify non-standard operating conditions such as optical path signal loss, continuous signal obstruction, continuous signal conduction, and abnormal workpiece placement. When an abnormal operating condition is detected, the anomaly information is promptly reported and the detection process is paused, awaiting manual verification and processing, to prevent abnormal workpieces from flowing into subsequent production processes.

[0095] Example 2 Please refer to the appendix. Figure 2-4 This embodiment also provides a mask plate iron frame posture detection device for implementing a mask plate iron frame posture detection method as provided in Embodiment 1. The device mainly includes a sealed box opening mechanism 1, a conveying mechanism 2, a double-sided laser beam assembly 4, and a control processing unit 3.

[0096] The hermetically sealed box opening mechanism 1 is used to open the hermetically sealed box carrying the framed mask workpiece 5 or the frameless monolithic mask workpiece. The hermetically sealed box uses a standard mechanical interface wafer transfer box. Existing automated hermetically sealed box opening methods are mainly divided into two categories: one is base fixing and box cover lifting for opening; the other is box cover fixing and base sinking for opening. This embodiment adopts the box cover fixing and base sinking opening scheme: when the equipment is working, the hermetically sealed box cover is kept in a fixed position. By driving the base carrying the workpiece to sink downwards, the base is separated from the box cover, exposing the framed mask workpiece 5 or the frameless monolithic mask workpiece inside, providing working space for subsequent optical path inspection processes.

[0097] The conveying mechanism 2 is used to transport the framed mask workpiece 5 or the frameless single-piece mask workpiece along the conveying path. The conveying mechanism 2 can use linear guides for motion guidance and is powered by a servo motor or stepper motor, enabling controllable and uniform lifting motion. The operating speed can be adaptively adjusted according to actual testing requirements. The lifting stroke of the conveying mechanism 2 covers the entire height range of the framed mask workpiece 5 or the frameless single-piece mask workpiece, ensuring that the workpiece passes completely through the horizontal detection optical path constructed by the dual-sided laser beam assembly 4. The conveying mechanism 2 can be equipped with a high-precision position encoder, which can provide real-time feedback of the mechanism's position information, providing a positional reference for the precise division of multiple independent functional detection sections along the conveying path.

[0098] The dual-sided laser beam-beam assembly 4 is symmetrically arranged on the left and right sides of the conveying path to construct a horizontal detection optical path. The dual-sided laser beam-beam assembly 4 consists of symmetrically arranged red laser emitting and receiving units, with each side integrating both a laser emitting unit and a laser receiving unit. The spot diameter, signal response time, system scanning cycle, effective sampling frequency, and workpiece movement speed corresponding to the conveying path of the dual-sided laser beam-beam assembly 4 are matched and set according to the detection accuracy requirements, and are used to acquire the continuous optical path waveform as the frame of the masked workpiece 5 crosses the detection optical path along the conveying path.

[0099] The control processing unit 3 is the core control and data processing module of the device. It uses a programmable logic controller (PLC) as its hardware carrier and has a built-in algorithm program for attitude detection. The control processing unit 3 is configured to: retrieve the geometric dimensions of the corresponding batch of framed mask workpiece 5 or frameless monolithic mask workpiece for dynamic calibration calculation; divide the conveying path into multiple independent functional detection sections based on the calibration results; collect the occurrence counts of the optical path conduction edge and optical path occlusion edge output by the dual-sided laser beam-beam assembly 4 within each independent functional detection section to construct a basic feature vector for attitude determination; execute a preset multi-branch judgment logic based on the basic feature vector to distinguish between framed mask workpiece 5 and frameless monolithic mask workpiece; extract the quantized waveform parameters of the continuous optical path waveform within the independent functional detection section corresponding to the mask body; retrieve the pre-stored standard attitude reference waveform, perform numerical matching calculations between the extracted quantized waveform parameters and the standard attitude reference waveform, and calculate the attitude matching confidence index based on the workpiece type determination result; and, based on the attitude matching confidence index, compare the results using pre-stored multi-level confidence judgment rules to output the corresponding attitude determination or alarm result.

[0100] In addition, the control and processing unit 3 also has an online self-calibration function for the optical path and a full-process fault tracing and storage function. The specific implementation of the corresponding functions has been described in Embodiment 1 and will not be repeated here.

[0101] The above description, in conjunction with specific embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the posture of a mask frame, characterized in that, include: The sealed box carrying the framed mask workpiece or the frameless single-piece mask workpiece is opened, and the conveying mechanism is driven to transport the framed mask workpiece or the frameless single-piece mask workpiece through the horizontal detection optical path formed by the double-sided laser beam assembly. The geometric dimension parameters of the corresponding batch of framed mask workpieces or frameless single-piece mask workpieces are retrieved for dynamic calibration calculation, and multiple independent functional detection sections are divided along the conveying path based on the calibration results; In each independent functional detection section, the number of occurrences of the optical path conduction edge and optical path occlusion edge output by the dual-sided laser beam assembly are collected to construct the basic feature vector for attitude determination. Based on the basic feature vector, a preset multi-branch decision logic is executed to distinguish the workpiece type as a framed mask workpiece or a frameless single-piece mask workpiece. Extract the quantized waveform parameters of the continuous optical path waveform within the independent functional detection section corresponding to the main body of the mask; Retrieve the pre-stored standard attitude reference waveform, perform numerical matching between the extracted quantized waveform parameters and the standard attitude reference waveform, and calculate the attitude matching confidence index based on the workpiece type determination result. Based on the attitude matching confidence index, the system compares the results using pre-stored multi-level confidence judgment rules and outputs the corresponding attitude judgment or alarm results.

2. The method for detecting the posture of a mask frame according to claim 1, characterized in that, The independent functional detection section includes four independent functional detection sections. The boundaries of the four independent functional detection sections are generated by dynamic calibration of the continuous optical path waveform corresponding to the current batch of framed mask workpieces or frameless single-piece mask workpieces. The dynamic calibration process includes: acquiring the continuous optical path waveform corresponding to the framed mask workpiece or frameless single-piece mask workpiece in the current batch; performing first derivative analysis on the continuous optical path waveform; automatically identifying the optical path occlusion start and end points corresponding to the upper and lower edges of the mask and the edge of the mask body; automatically dividing the boundaries of independent functional detection sections according to the optical path occlusion start and end points; and synchronously updating the standard attitude reference waveform stored in the current batch. The four independent functional detection sections are divided into the following functional sections: the first independent functional detection section is used to identify the lower edge of the mask; the second and third independent functional detection sections correspond to the mask body to collect continuous optical path waveforms; and the fourth independent functional detection section is used to identify the upper edge of the mask.

3. The method for detecting the posture of a mask frame according to claim 2, characterized in that, The preset multi-branch determination logic is based on the basic feature vector constructed from four independent functional detection segments to distinguish four types of framed mask workpiece postures and frameless single-piece mask workpieces. Among them, the optical path conduction edge is the rising edge when the optical path signal switches from the blocked state to the conduction state, and the optical path blockage edge is the falling edge when the optical path signal switches from the conduction state to the blockage state. The preset multi-branch decision logic includes: If the first independent functional detection section detects an upward edge, the framed mask workpiece is determined to be in a frame-down orientation. Specifically, if one side of the laser beam unit in the dual-sided laser beam assembly detects an upward edge while the other side does not, it corresponds to a 0° placement orientation. If both sides of the laser beam unit in the dual-sided laser beam assembly detect an upward edge, it corresponds to a 90° placement orientation. If the first independent functional detection section does not detect a rising edge, and the second and third independent functional detection sections detect the optical path signal corresponding to the mask body, then the framed mask workpiece is determined to be in the position with the frame facing upwards. Among them, if a single-sided laser beam unit in the dual-sided laser beam assembly detects a falling edge in the fourth independent functional detection section, but the other-sided laser beam unit does not detect a falling edge, it corresponds to a 0° placement posture. If both sides of the laser beam unit in the dual-sided laser beam assembly detect a falling edge in the fourth independent functional detection section, it corresponds to a 90° placement posture. If none of the four independent functional detection sections detect the optical path signal corresponding to the frame of the framed mask workpiece, and only the second and third independent functional detection sections detect the optical path signal corresponding to the main body of the mask, then the workpiece being detected is determined to be a frameless single-piece mask workpiece.

4. The method for detecting the posture of a mask frame according to claim 1, characterized in that, The quantized waveform parameters include the timing difference between the two-sided sensor signals, the proportion of continuous optical path occlusion, and the rate of change of the signal edge.

5. The method for detecting the posture of a mask frame according to claim 4, characterized in that, The attitude matching confidence index is obtained by comprehensively evaluating the matching degree of the timing difference between the two-sided sensor signals, the matching degree of the proportion of continuous occlusion of the optical path, and the matching degree of the signal edge change rate, combined with the corresponding weight coefficients automatically trained and generated in the current batch dynamic calibration process. It is used to characterize the degree of matching between the quantized waveform parameters and the standard attitude reference waveform. The weighting coefficients are adaptively adjusted based on the geometric dimensions of the corresponding batch.

6. The method for detecting the posture of a mask frame according to claim 1, characterized in that, The multi-level confidence determination rule performs hierarchical closed-loop processing based on the attitude matching confidence index; the hierarchical closed-loop processing is used to filter dust, electromagnetic interference, and occasional light path obstruction interference caused by mechanism vibration.

7. The method for detecting the posture of a mask frame according to claim 1, characterized in that, It also includes an online self-calibration step for the optical path: When the equipment is in an unloaded state, it runs to the reference positions at both ends of the lifting stroke set along the conveying path, and collects the continuous optical path waveforms in the unobstructed state and the completely blocked state respectively. The optical path attenuation difference is calculated, and the judgment level thresholds corresponding to optical path conduction and optical path blockage are dynamically corrected to compensate for the signal deviation caused by laser aging, optical window dust accumulation, and installation position offset.

8. The method for detecting the posture of a mask frame according to claim 1, characterized in that, It also includes a full-process fault tracing and storage step: During each test, the continuous optical path waveform, boundary parameters of the independent functional test section, mechanism motion speed curve, and real-time laser intensity data corresponding to this test are stored synchronously and bound to the task number. Based on the task number, retrieve historical continuous optical path waveforms to reproduce the entire scanning process and identify the deviation sections between the acquired continuous optical path waveforms and the standard attitude reference waveforms.

9. The method for detecting the posture of a mask frame according to claim 1, characterized in that, The dual-sided laser beam-emitting assembly consists of two sets of red laser emitting and receiving units arranged symmetrically on the left and right sides. The spot diameter, response time, scanning period, effective sampling frequency, and moving speed corresponding to the conveying path of the dual-sided laser beam-emitting assembly are adapted to each other to collect the continuous optical path waveform of the frame mask workpiece as it passes through the optical path along the conveying path.

10. A mask frame posture detection device, characterized in that, include: The sealed box opening mechanism is used to complete the opening operation of the sealed box that carries the framed mask workpiece or the frameless single mask workpiece. A conveying mechanism is used to transport framed or frameless monolithic mask workpieces along a conveying path. Dual-sided laser beam-emitting components are symmetrically arranged to form a horizontal detection optical path; The control processing unit is configured to: retrieve the geometric dimension parameters of the corresponding batch of framed mask workpieces or frameless monolithic mask workpieces for dynamic calibration calculation, and divide the conveying path into multiple independent functional detection sections based on the calibration results; collect the number of occurrences of the optical path conduction edge and optical path occlusion edge of the output of the dual-sided laser beam-beam assembly in each independent functional detection section to construct a basic feature vector for attitude determination; execute a preset multi-branch determination logic based on the basic feature vector to distinguish whether the workpiece type is a framed mask workpiece or a frameless monolithic mask workpiece; and extract the quantized waveform parameters of the continuous optical path waveform in the independent functional detection section corresponding to the mask body. Retrieve the pre-stored standard attitude reference waveform, perform numerical matching between the extracted quantized waveform parameters and the standard attitude reference waveform, and calculate the attitude matching confidence index based on the workpiece type determination result. Based on the attitude matching confidence index, the system compares the results using pre-stored multi-level confidence judgment rules and outputs the corresponding attitude judgment or alarm results.

Images