Glass fiber raw material two-stage gate plate quantitative robot collaborative xrf and moisture joint measurement system

Abstract

The application discloses a kind of glass fiber raw material two-stage gate quantitative robot cooperation XRF and moisture joint measurement system, it is related to glass fiber production raw material detection technical field, by the grading control material of two-stage gate and volume self-adaptive calibration realizes the high-precision quantitative material taking of glass fiber raw material, industrial robot realizes seamless transfer and attitude optimization between two detection modules of quantitative material, XRF and moisture detection module synchronous acquisition data and eliminate cross interference by coupling algorithm, finally form "quantitative-detection-analysis-feedback" whole-process closed loop.The application improves quantitative accuracy to ±0.1%, detection efficiency is improved by more than 8 times compared with traditional method, composition and moisture detection data synergy error≤0.3%, can directly link glass fiber batching system to realize process dynamic optimization, meet the stringent requirements of aerospace, high-end wind power and other fields to glass fiber raw material "high consistency, traceable", significantly reduce raw material loss rate and labor cost.

Description

Technical Field

[0001] This invention belongs to the technical field of glass fiber production raw material testing, specifically involving a glass fiber raw material dual-stage gate quantitative robot collaborative XRF and moisture detection system. Background Technology

[0002] As a core reinforcing material for wind turbine blades, aerospace composite materials, and lightweight components for new energy vehicles, the consistency of glass fiber's quality directly determines the mechanical properties and service life of the end products. Raw material testing is the first line of defense in glass fiber production quality control, requiring simultaneous control of composition accuracy and moisture content, both of which affect the glass melt melting effect and the performance of the precursor fiber.

[0003] Existing fiberglass raw material testing technologies suffer from several bottlenecks: First, quantitative sampling accuracy is insufficient. Traditional two-stage gate devices rely solely on fixed-volume metering, which cannot adapt to fluctuations in the bulk density of fiberglass raw materials (such as quartz sand and kaolin). Quantitative errors can reach ±2% or more, resulting in unrepresentative test samples. Second, composition and moisture detection are disconnected. XRF is often used to detect composition first, followed by separate moisture detection. This fragmented process, coupled with the fact that moisture content interferes with the accuracy of XRF spectroscopy, leads to increased errors in composition analysis and prevents the formation of coherent data. Third, testing is disconnected from production. Manual sample transfer is inefficient, and test data cannot be fed back to the batching system in real time, resulting in significant lag. This leads to a large influx of substandard raw materials into the production process, with a raw material loss rate as high as 8%-12%. Fourth, automation is low. Relying on manual operation for material collection, transfer, and calibration is not only inefficient (single sample testing takes over 30 minutes) but also carries the risks of subjective human error and safety hazards, making it difficult to meet the large-scale requirements of tens of thousands of tons of fiberglass production lines.

[0004] In the existing technology, some solutions attempt to optimize quantitative feeding or single detection modules. For example, the dual-stage gate quantitative device disclosed in the patent only realizes the basic material control function and lacks volume adaptive calibration and anti-bridging design. Some intelligent detection systems only focus on the detection of a single component or moisture and do not realize the simultaneous detection of two parameters and interference elimination. Summary of the Invention

[0005] The purpose of this invention is to provide a dual-stage gate-type quantitative robot collaborative XRF and moisture measurement system for glass fiber raw materials, so as to solve the problems mentioned in the background art.

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

[0007] A dual-stage gate-type quantitative robotic collaborative XRF and moisture measurement system for fiberglass raw materials, characterized in that it includes: The dual-stage gate intelligent quantitative module includes a feeding section, a metering section, a discharging section, an upper gate assembly, a lower gate assembly, a volume adaptive calibration unit, and an intelligent vibration anti-bridging unit. The metering section is located between the feeding and discharging sections. The upper and lower gate assemblies are respectively located at the junctions of the feeding section-metering section and the metering section-discharging section, forming a dynamically sealable quantitative metering cavity. This addresses the technical challenge of traditional fixed-volume metering methods failing to adapt to fluctuations in the bulk density of fiberglass raw materials. The volume adaptive calibration unit consists of a retractable cavity wall and a high-precision laser ranging sensor. The system is composed of components that can collect raw material stacking height data in real time and dynamically adjust the metering chamber volume in combination with a preset raw material stacking density model to achieve high-precision material dispensing with a quantitative error of ≤±0.1%, which is significantly lower than the error of more than ±2% of traditional double-stage gates. The intelligent vibration anti-bridging unit is assembled on the outside of the metering section and adopts a high-frequency low-amplitude vibration mode. Its vibration frequency and amplitude can be adaptively matched according to the particle size characteristics of glass fiber raw materials, which can not only completely eliminate the quantitative deviation caused by raw material bridging, but also avoid the raw material stratification problem caused by excessive vibration, thus ensuring sample uniformity. The industrial robot collaboration module includes a six-axis industrial robot, a composite end effector, a deep learning vision positioning unit, and a human-machine safety protection unit, constructing a seamless collaborative transfer mechanism between the quantitative material handling and dual detection modules. The composite end effector integrates a dual structure of vacuum adsorption and mechanical clamping, and can automatically switch operating modes according to the particle size of the glass fiber raw material, adapting to the transfer needs of various types of raw materials. The deep learning vision positioning unit consists of an industrial camera and a dedicated recognition algorithm, which can accurately identify the spatial coordinates and raw material posture of the quantitative module outlet, XRF detection station, and moisture detection station, with a positioning accuracy of ≤±0.2mm, ensuring the consistency of posture when the sample is transferred to the detection station, providing a guarantee for the accuracy of subsequent synchronous detection, and breaking through the limitation of traditional robots that can only achieve simple transfer. The XRF component detection module includes a high-performance XRF spectrometer, a dedicated protective detection chamber, a compatible sample carrier, and a batch-based automatic calibration unit, optimized for glass fiber raw material detection scenarios. The dedicated protective detection chamber features a radiation shielding layer with a lead equivalent of ≥2mm and an argon inert gas protection system, preventing radiation leakage and ensuring operational safety while effectively preventing raw material oxidation during detection, thus avoiding interference with the detection accuracy of key components such as SiO2 and Al2O3. The batch-based automatic calibration unit incorporates a standard sample library covering various glass fiber raw materials, including E-CR and high-silica materials. It automatically calibrates the instrument after testing every 10 batches of raw materials, ensuring stable component detection accuracy at the ppm level, meeting the stringent requirements of high-end glass fiber for component detection. The moisture detection module employs a high-frequency capacitive detection principle, combined with a constant temperature and humidity control chamber and a robot-assisted sample spreading mechanism, to achieve synchronous linkage with XRF component detection. The constant temperature and humidity control chamber can control the detection environment temperature at 25±1℃ and humidity at 50±5%RH, completely eliminating the interference of ambient temperature and humidity on moisture detection. The robot-assisted sample spreading mechanism, through linkage with the movement of an industrial robot, evenly spreads a quantitative amount of raw material onto the detection tray, with a spreading thickness error ≤0.5mm, ensuring the consistency of raw material thickness in the detection area. This allows the moisture detection range to cover 0.01%-20%, the detection accuracy to reach ±0.02%, and the detection time ≤3s, achieving synchronous detection of two parameters without delay. The central control and data processing module, including an industrial controller, a high-speed data acquisition card, a moisture-XRF cross-interference coupling analysis algorithm module, and a full-process data traceability unit, serves as the core hub for the system's collaborative operation. The coupling analysis algorithm module constructs a correlation database between moisture content and XRF characteristic spectral line intensity based on over 100,000 sets of glass fiber raw material samples. It uses a CNN convolutional neural network algorithm to fit a correction function, iteratively correcting the original detection data. This effectively eliminates the cross-interference of moisture content on XRF component detection, ensuring that the corrected data deviation is ≤0.3%, thus solving the technical bottleneck of disconnected component and moisture data and the inability to eliminate interference in traditional detection methods. The full-process data traceability unit uses a unique batch identifier to bind quantitative parameters, dual detection data, equipment operating status, and other information for each batch of raw materials and stores them in the cloud, achieving full-process traceability of the detection process and meeting the auditing requirements of high-end customers for raw material quality records. The closed-loop feedback module establishes a multi-level response control mechanism to achieve real-time linkage between testing data and the production system, breaking through the limitations of traditional testing and production being disconnected. This module compares the testing data processed by the central control module with preset standard values ​​to generate targeted control instructions. These instructions are transmitted in real time to the fiberglass raw material batching system to achieve dynamic optimization of the raw material ratio. On the other hand, they are fed back to the dual-stage gate intelligent quantitative module to complete the adaptive correction of the measurement parameters, forming a closed-loop control of the entire process of "quantification-detection-analysis-feedback-optimization", which significantly reduces the raw material loss rate.

[0008] Preferably, the dual-stage gate intelligent quantitative module achieves high-precision and rapid quantitative measurement through graded action coordination. The workflow is as follows: the lower gate assembly closes to form a bottom seal of the metering chamber → the upper gate assembly opens to allow the raw material to enter the metering chamber → the laser rangefinder sensor detects the raw material accumulation height in real time and calculates the actual material intake based on the chamber volume and accumulation density model → the volume adaptive calibration unit fine-tunes the chamber volume to compensate for errors and ensure quantitative accuracy → the upper gate assembly closes to form a top seal of the metering chamber → the intelligent vibration anti-bridging unit is activated to vibrate and break arches with parameters adapted to the characteristics of the raw material → the lower gate assembly opens to allow the raw material to fall accurately into the end effector. The single quantitative cycle is ≤8s, which is more than 3 times more efficient than the traditional quantitative process.

[0009] Preferably, the industrial robot collaboration module adopts a dual control mode of "deep learning vision guidance + dynamic path planning". The robot end effector integrates a force sensor, which can collect the actual weight data of the raw materials in real time and feed it back to the central control module. It forms a two-way secondary verification with the measurement data of the dual-gate intelligent quantitative module, further controlling the quantitative comprehensive error within ±0.1%, which significantly improves the reliability compared with a single measurement method.

[0010] Preferably, the core logic of the coupling analysis algorithm module is as follows: based on the natural correlation between the moisture content of glass fiber raw materials and the XRF spectral response, a massive sample correlation database is constructed. The inherent correlation of the data is mined through the CNN convolutional neural network algorithm and a dedicated correction function is fitted. The original detection data is iteratively corrected in multiple rounds to specifically eliminate the XRF component detection error caused by moisture interference. Finally, the synergy deviation between component and moisture detection data is ≤0.3%, solving the technical problem that dual-parameter interference cannot be quantitatively eliminated in traditional detection.

[0011] Preferably, the multi-level response mechanism of the closed-loop feedback module is as follows: when the detection data exceeds the preset deviation range of ±0.5%, the batching system parameters are immediately triggered for emergency adjustment and the feeding of the batch of raw materials is suspended to prevent batches of unqualified raw materials from entering production; when the deviation range is ±0.1%-±0.5%, the metering parameters are finely adjusted through the dual-stage gate intelligent quantitative module to achieve dynamic correction without affecting the continuity of production; when the deviation is ≤±0.1%, the current production parameters are maintained to operate stably, and the balance between quality and efficiency is achieved through graded precise control, so that the fiberglass raw material loss rate is reduced from the traditional 8%-12% to below 3%.

[0012] Compared with existing technologies, this invention provides a dual-stage gate-type quantitative robotic collaborative XRF and moisture measurement system for glass fiber raw materials, which has the following advantages: Breakthrough improvement in quantitative accuracy: Through volume adaptive calibration and vibration anti-bridging design, the quantitative error of glass fiber raw materials has been reduced from more than ±2% to within ±0.1%, significantly enhancing the representativeness of the test samples and laying the foundation for subsequent test accuracy; Dual improvement in detection synergy and efficiency: Simultaneous detection of XRF components and moisture is achieved, and seamless robot transport reduces the detection time for a single sample from more than 30 minutes to less than 10 minutes, improving efficiency by more than 8 times; the coupling algorithm eliminates moisture interference, making the component detection error ≤0.3%, solving the data distortion problem caused by the disconnect of traditional detection methods; Closed-loop linkage throughout the entire process: Constructing a closed-loop system of "quantification-detection-analysis-feedback", the detection data is fed back to the production system in real time, realizing dynamic optimization of raw material ratio, which can reduce the fiberglass raw material loss rate from 8%-12% to below 3%, significantly reducing production costs; High degree of automation and intelligence: It integrates multiple modules to work together, enabling 24-hour unattended testing. A single production line can reduce the number of testing personnel by more than 80%. At the same time, through full-process data traceability, it meets the supplier audit requirements of aerospace, high-end wind power and other fields, and helps the high-end upgrade of fiberglass products. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] This invention provides, for example Figure 1 shown A dual-stage gate-type quantitative robotic collaborative XRF and moisture measurement system for fiberglass raw materials, characterized in that it includes: The dual-stage gate intelligent quantitative module includes a feeding section, a metering section, a discharging section, an upper gate assembly, a lower gate assembly, a volume adaptive calibration unit, and an intelligent vibration anti-bridging unit. The metering section is located between the feeding and discharging sections. The upper and lower gate assemblies are respectively located at the junctions of the feeding section-metering section and the metering section-discharging section, forming a dynamically sealable quantitative metering cavity. This solves the technical problem that traditional fixed-volume metering cannot adapt to fluctuations in the bulk density of fiberglass raw materials. The volume adaptive calibration unit consists of a retractable cavity wall and a high-precision laser ranging sensor assembly. The system can collect raw material stacking height data in real time and dynamically adjust the metering chamber volume in combination with the preset raw material stacking density model to achieve high-precision material sampling with a quantitative error of ≤±0.1%, which is significantly lower than the error of more than ±2% of the traditional double-stage gate. The intelligent vibration anti-bridging unit is installed on the outside of the metering section and adopts a high-frequency low-amplitude vibration mode. Its vibration frequency and amplitude can be adaptively matched according to the particle size characteristics of the glass fiber raw material. It can not only completely eliminate the quantitative deviation caused by raw material bridging, but also avoid the raw material stratification problem caused by excessive vibration, thus ensuring sample uniformity. The industrial robot collaboration module includes a six-axis industrial robot, a composite end effector, a deep learning vision positioning unit, and a human-machine safety protection unit, constructing a seamless collaborative transfer mechanism between the quantitative material handling and dual detection modules. The composite end effector integrates a dual structure of vacuum adsorption and mechanical clamping, and can automatically switch operating modes according to the particle size of the glass fiber raw material to adapt to the transfer needs of various types of raw materials. The deep learning vision positioning unit consists of an industrial camera and a dedicated recognition algorithm, which can accurately identify the spatial coordinates and raw material posture of the quantitative module outlet, XRF detection station, and moisture detection station, with a positioning accuracy of ≤±0.2mm, ensuring the consistency of posture when the sample is transferred to the detection station, providing a guarantee for the accuracy of subsequent synchronous detection, and breaking through the limitation of traditional robots that can only achieve simple transfer. The XRF component detection module includes a high-performance XRF spectrometer, a dedicated protective detection chamber, a compatible sample carrier, and a batch-based automatic calibration unit, optimized for glass fiber raw material testing scenarios. The dedicated protective detection chamber features a radiation shielding layer with a lead equivalent of ≥2mm and an argon inert gas protection system, which not only eliminates radiation leakage and ensures operational safety but also effectively prevents raw material oxidation during testing, avoiding interference with the detection accuracy of key components such as SiO2 and Al2O3. The batch-based automatic calibration unit has a built-in standard sample library covering multiple types of glass fiber raw materials, including E-CR and high-silica materials. It automatically completes instrument calibration after testing every 10 batches of raw materials, ensuring that the component detection accuracy is stable at the ppm level, meeting the stringent requirements of high-end glass fiber for component detection. The moisture detection module employs a high-frequency capacitive detection principle, combined with a constant temperature and humidity control chamber and a robot-assisted sample spreading mechanism, to achieve synchronous linkage with XRF component detection. The constant temperature and humidity control chamber can control the detection environment temperature at 25±1℃ and humidity at 50±5%RH, completely eliminating the interference of ambient temperature and humidity on moisture detection. The robot-assisted sample spreading mechanism, through linkage with the movement of an industrial robot, evenly spreads a quantitative amount of raw material onto the detection tray, with a spreading thickness error ≤0.5mm, ensuring the consistency of raw material thickness in the detection area. This allows the moisture detection range to cover 0.01%-20%, the detection accuracy to reach ±0.02%, and the detection time ≤3s, achieving synchronous detection of two parameters without delay. The central control and data processing module, including an industrial controller, a high-speed data acquisition card, a moisture-XRF cross-interference coupling analysis algorithm module, and a full-process data traceability unit, serves as the core hub for the system's collaborative operation. The coupling analysis algorithm module constructs a correlation database between moisture content and XRF characteristic spectral line intensity based on over 100,000 sets of glass fiber raw material samples. It uses a CNN convolutional neural network algorithm to fit a correction function, iteratively correcting the original detection data. This effectively eliminates the cross-interference of moisture content on XRF component detection, ensuring that the corrected data deviation is ≤0.3%, thus solving the technical bottleneck of disconnected component and moisture data and the inability to eliminate interference in traditional detection methods. The full-process data traceability unit uses a unique batch identifier to bind quantitative parameters, dual detection data, equipment operating status, and other information for each batch of raw materials and stores them in the cloud, achieving full-process traceability of the detection process and meeting the auditing requirements of high-end customers for raw material quality records. The closed-loop feedback module establishes a multi-level response control mechanism to achieve real-time linkage between testing data and the production system, breaking through the limitations of traditional testing and production being disconnected. This module compares the testing data processed by the central control module with preset standard values ​​to generate targeted control instructions. These instructions are transmitted in real time to the fiberglass raw material batching system to achieve dynamic optimization of the raw material ratio. On the other hand, they are fed back to the dual-stage gate intelligent quantitative module to complete the adaptive correction of the measurement parameters, forming a closed-loop control of the entire process of "quantification-detection-analysis-feedback-optimization", which significantly reduces the raw material loss rate.

[0016] The dual-gate intelligent quantitative module achieves high-precision and rapid quantitative measurement through graded action coordination. The workflow is as follows: the lower gate assembly closes to form a bottom seal of the metering chamber → the upper gate assembly opens to allow the raw material to enter the metering chamber → the laser rangefinder sensor detects the raw material accumulation height in real time and calculates the actual material intake based on the chamber volume and accumulation density model → the volume adaptive calibration unit fine-tunes the chamber volume to compensate for errors and ensure quantitative accuracy → the upper gate assembly closes to form a top seal of the metering chamber → the intelligent vibration anti-bridging unit is activated to break the arch with vibration parameters adapted to the characteristics of the raw material → the lower gate assembly opens to allow the raw material to fall accurately into the end effector. The single quantitative cycle is ≤8s, which is more than 3 times more efficient than the traditional quantitative process.

[0017] The industrial robot collaboration module adopts a dual control mode of "deep learning vision guidance + dynamic path planning". The robot end effector integrates a force sensor, which can collect the actual weight data of the raw materials in real time and feed it back to the central control module. It forms a two-way secondary verification with the measurement data of the dual-gate intelligent quantitative module, further controlling the quantitative comprehensive error within ±0.1%, which significantly improves the reliability compared with a single measurement method.

[0018] The core logic of the coupling analysis algorithm module is as follows: Based on the natural correlation between the moisture content of glass fiber raw materials and the XRF spectral response, a massive sample correlation database is constructed. The inherent correlation of the data is mined through the CNN convolutional neural network algorithm and a dedicated correction function is fitted. The original detection data is iteratively corrected in multiple rounds to specifically eliminate the XRF component detection error caused by moisture interference. Ultimately, the correlation deviation between component and moisture detection data is ≤0.3%, solving the technical problem that the dual-parameter interference in traditional detection cannot be quantitatively eliminated.

[0019] The multi-level response mechanism of the closed-loop feedback module is as follows: when the detection data exceeds the preset deviation range of ±0.5%, the batching system parameters are immediately triggered for emergency adjustment and the feeding of the batch of raw materials is suspended to prevent batches of unqualified raw materials from entering production; when the deviation range is ±0.1%-±0.5%, the metering parameters are finely adjusted through the dual-stage gate intelligent quantitative module to achieve dynamic correction without affecting the continuity of production; when the deviation is ≤±0.1%, the current production parameters are maintained to operate stably, and the balance between quality and efficiency is achieved through graded precise control, reducing the fiberglass raw material loss rate from the traditional 8%-12% to below 3%.

[0020] I. Overall System Assembly Scheme This embodiment is designed for a 150,000-ton-class high-end wind power fiberglass production line, used for incoming inspection of mixed raw materials such as quartz sand, kaolin, and limestone. The system is arranged between the raw material pretreatment workshop and the batching workshop, occupying an area of ​​approximately 20 square meters. The assembly details of each module are as follows: 1. Assembly of a dual-stage gate intelligent quantitative module This module is vertically installed below the discharge port of the raw material screening machine. The feeding section is made of stainless steel, with an inner diameter of 300mm and a length of 500mm. The inner wall is polished (roughness Ra≤0.8μm) to reduce raw material adhesion. The metering section is the core functional section, with an outer diameter of 400mm and an initial inner diameter of 300mm. The telescopic cavity wall is made of titanium alloy. Volume adjustment is achieved by a ball screw driven by a servo motor, with an adjustment stroke of ±50mm and an adjustment accuracy of ±0.01mm. Both the upper and lower gate components are pneumatically driven, with cylinder model SMCCDQ2B50-100D, working pressure of 0.6-0.8MPa, and response time ≤0.3s. The gate seal is made of zirconia ceramic with a thickness of 10mm. The contact surface with the chamber adopts a V-shaped sealing structure, and the sealing accuracy reaches zero visible leakage (no pressure drop after holding pressure at 0.8MPa for 5 minutes).

[0021] The volume adaptive calibration unit uses a Keyence IL-300 laser rangefinder, installed at the top center of the metering section, with a detection distance of 50-500mm and an accuracy of ±0.01mm. The vertical distance between the sensor probe and the raw material drop area is maintained at 200mm. It is fixed by a bracket and equipped with a dust cover (IP67 protection rating). The intelligent vibration anti-bridging unit uses an electromagnetic vibrator, model ZFB-5, installed on the lower outer side of the metering section. The vibration frequency adjustment range is 20-50Hz, and the amplitude adjustment range is 0.1-0.5mm. It is connected to the metering section through an elastic connector to prevent vibration from being transmitted to other modules.

[0022] 2. Assembly of Industrial Robot Collaborative Modules An ABBIRB1200 six-axis industrial robot with a 5kg payload and a repeatability of ±0.02mm was selected and installed on a ground base between the dual-stage gate quantitative module and the detection module. The base is made of cast concrete and reinforced with embedded steel plates to ensure that the vibration displacement during robot operation is ≤0.1mm. The composite end effector is independently designed, with the main body made of aluminum alloy and weighing ≤1.2kg. The upper part integrates a vacuum suction cup (model SMCZPT20UN) with a vacuum level of -0.08MPa, suitable for fine powder materials with a particle size ≤0.1mm; the lower part has two symmetrically arranged mechanical grippers with a gripper stroke of 0-80mm and a gripping force adjustment range of 5-50N, suitable for granular materials with a particle size of 0.1-5mm. The two modes are automatically switched by a solenoid valve with a switching time ≤0.2s.

[0023] The deep learning visual positioning unit consists of two Hikvision MV-CA050-10GM industrial cameras, mounted on the robot's shoulder support and the top of the detection area, respectively. The cameras have a resolution of 5 megapixels, a frame rate of 30fps, and are equipped with 8mm fixed-focus lenses. The visual recognition algorithm is trained based on the YOLOv8 model, with training samples covering more than 1,000 images of fiberglass raw materials with different postures and particle sizes, achieving a positioning accuracy of ±0.2mm. The human-machine safety protection unit uses an infrared grating (model SICKC40S-0903CA010) to construct a protection zone, covering the robot's movement trajectory and the detection module. It is also equipped with an emergency stop button and an audible and visual alarm device. When a human enters the protection zone, the robot immediately stops running, and the alarm device is activated (alarm volume ≥85dB).

[0024] 3. Assembly of XRF component detection module The XRF spectrometer used is a Panaco Epsilon4, with a detection range of Na-U, a characteristic spectral line resolution of ≤150eV, and a detection time adjustable from 0.5 to 10 seconds. In this embodiment, it is set to 2 seconds. The dedicated protective detection chamber is a self-welded structure with an outer layer of 3mm thick stainless steel plate and an inner layer of 2mm thick lead plate (lead equivalent ≥2mm). The chamber door adopts an electromagnetic interlock design, and the instrument can only start detection when the door is completely closed, thus preventing radiation leakage (external radiation dose ≤0.1μSv / h).

[0025] The inert gas protection system uses 99.999% pure argon gas, which is adjusted to 0.1-0.15MPa through a pressure reducing valve and then introduced into the detection chamber through a flow meter (accuracy ±0.01L / min). The gas flow rate is set to 2L / min to ensure that the oxygen content in the chamber is ≤0.5% to prevent the oxidation of the raw materials. The batch automatic calibration unit has 10 built-in glass fiber raw material standard samples (covering E-CR, high silica, alkali-free glass fiber, etc.). The standard sample particle size is 0.1-0.3mm and the moisture content is 0.5±0.02%. The calibration program is automatically started every 10 batches of raw materials (5 samples per batch). The calibration time is ≤3min, and the component detection accuracy is stable at the ppm level after calibration.

[0026] 4. Assembly of the synchronous moisture detection module The moisture detection sensor uses the German Möss MS300 high-frequency capacitive sensor, with a working frequency of 20MHz, a detection range of 0.01%-20%, and an accuracy of ±0.02%. The constant temperature and humidity control chamber adopts a closed structure, with a built-in heating tube (power 500W), a cooling plate (power 300W), and an ultrasonic humidifier. The temperature and humidity are controlled at 25±1℃ and the relative humidity at 50±5%RH by a temperature and humidity controller (model Siemens S7-1200), with temperature and humidity fluctuations ≤0.5℃ / ±2%RH.

[0027] The robot-assisted sample spreading mechanism consists of a spreading scraper and a drive motor. The scraper is made of polytetrafluoroethylene (PTFE) (to avoid scratching the test tray), with a thickness of 2mm and a width of 80mm. It is driven by a servo motor to achieve horizontal reciprocating motion at a spreading speed of 50mm / s. It is linked with the industrial robot's movements. When the robot feeds the raw material into the test tray, the scraper automatically starts and spreads the raw material into a uniform thin layer with a thickness of 2mm (error ≤0.5mm). After spreading is completed, the scraper resets, and the sensor starts detection. The detection time is set to 3s.

[0028] 5. Assembly of the central control and data processing module The industrial controller uses a Siemens S7-1500 PLC, equipped with a CPU1516C-3PN / DP, 1MB of memory, and supports high-speed data acquisition and multi-module linkage control. The high-speed data acquisition card uses an Advantech PCI-1716 with a sampling frequency of 100Hz and a resolution of 16 bits. It synchronously acquires quantitative parameters, XRF component data, moisture data, robot status and other signals. Data transmission uses industrial Ethernet with a delay of ≤1s.

[0029] The coupling analysis algorithm module is developed based on the MATLAB R2023b platform and adopts the CNN convolutional neural network algorithm. The network structure includes an input layer, three convolutional layers, two pooling layers, a fully connected layer, and an output layer. The input data are moisture detection values ​​and raw XRF spectral data, and the output data is the corrected component content values. The algorithm model is trained and optimized through 100,000+ sets of glass fiber raw material samples (covering different moisture contents, particle sizes, and component ratios), with 1000 iterations and a convergence error ≤0.001. The full-process data traceability unit uses an SQL Server database and binds data through a unique batch code (consisting of production date + raw material type + test serial number, a total of 18 digits). The stored content includes quantitative error, test data, equipment operating parameters, calibration records, etc., with a storage time of ≥3 years, and supports remote query and export.

[0030] 6. Closed-loop feedback module assembly This module is linked with the fiberglass raw material batching system (Siemens PCS7) via the Profinet bus, with a feedback command transmission delay of ≤1s; at the same time, it is connected to the servo motor and cylinder controller of the two-stage gate metering module through the analog output module (Siemens SM153-1) to realize adaptive correction of metering parameters, with a command output accuracy of ±0.01mm.

[0031] II. System Operation Process and Parameter Settings This embodiment tests a mixture of quartz sand and kaolin (mixing ratio 7:3) with the following preset standards: SiO2 content 65±0.3%, Al2O3 content 18±0.2%, moisture content 0.5±0.05%, single quantitative sampling amount 500g, and the system operation process and key parameter settings are as follows: 1. Preprocessing and initialization stage After impurities are removed by a vibrating screen (5mm mesh size), the fiberglass raw material is fed into the feeding section of the dual-stage gate intelligent quantitative module. At the same time, the system initialization program is started: the PLC controls the self-test of each module, the robot returns to the origin, the XRF spectrometer completes preheating (preheating time 30min) and performs the first calibration, the moisture detection module starts constant temperature and humidity control, and after the temperature and humidity of the chamber reach the set value (25℃, 50%RH) and stabilize for 10min, the closed-loop feedback module reads the preset standard value of the batching system and stores it.

[0032] 2. High-precision quantitative stage First, the lower gate assembly closes under the drive of a cylinder, sealing the bottom of the metering chamber. The PLC records the initial volume of the metering chamber as 500mL. Second, the upper gate assembly opens, and the raw material enters the metering chamber under gravity. Simultaneously, the laser rangefinder sensor activates, collecting data on the material's stacking height every 10ms and transmitting it to the PLC in real time. Third, the PLC, combined with a preset raw material bulk density model (quartz sand bulk density 1.6g / cm³),... 3 Kaolin bulk density is 1.4 g / cm³. 3 The bulk density after mixing is 1.52 g / cm³. 3 The process involves six steps: First, the actual material intake is calculated. When the intake is close to 500g (error ≤ ±0.5%), the PLC controls the upper gate assembly to decelerate and close. Second, the volume adaptive calibration unit is activated, and the servo motor drives the retractable cavity wall to fine-tune the volume based on laser ranging data, ensuring the actual material intake is accurately 500g with a quantitative error within ±0.1%. Third, the upper gate assembly is fully closed, forming a seal at the top of the metering cavity. The intelligent vibration anti-bridging unit is activated, setting the vibration frequency to 30Hz, amplitude to 0.3mm, and vibration time to 2s to eliminate material bridging. Fourth, the lower gate assembly opens, and the material falls into the robot's composite end effector. The end force sensor collects the material weight (500±0.05g) in real time, forming a secondary verification with the laser ranging measurement data. After passing the verification, the process proceeds to the next stage, with a single quantitative cycle of ≤8s.

[0033] 3. Intelligent transfer and synchronous detection stage The deep learning vision positioning unit is activated, and the industrial camera acquires the spatial coordinates of the quantitative module outlet, the XRF detection station, and the moisture detection station. The algorithm identifies the posture of the raw material inside the end effector and plans the optimal transfer path (path length 1.5m, transfer time 2s). The robot uses a mechanical clamping mode with a composite end effector (raw material particle size 0.2-3mm) to transfer the raw material to the detection area. In the first step, the robot feeds the raw material into the constant temperature and humidity chamber of the moisture detection module. The spreading mechanism is activated, spreading the raw material evenly into a 2mm thick layer. After spreading, the scraper resets, the moisture sensor starts detection, and after 3 seconds, the moisture data is collected and transmitted to the PLC. In the second step, the robot transfers the same batch of raw material to the XRF detection chamber. The chamber door automatically closes, the argon gas protection system is activated, and after 3 seconds of ventilation, the XRF spectrometer starts detection. After 2 seconds, the raw composition data (spectral signals of key components such as SiO2 and Al2O3) is collected. After detection, the chamber door opens, the robot transfers the raw material to the recovery channel, and returns to the origin to await the next transfer.

[0034] 4. Data Processing and Closed-Loop Feedback Stage The first step involves the high-speed data acquisition card of the central control module simultaneously receiving moisture detection data and raw XRF data. The coupled analysis algorithm module then initiates multiple iterations based on a CNN convolutional neural network model to eliminate the interference of moisture content on XRF component detection, outputting accurate detection data after correction. The second step involves the full-process data traceability unit generating a unique batch code (e.g., 20260204-QK-001, representing the first batch of quartz sand-kaolin mixed raw materials on February 4, 2026), binding quantitative parameters, detection data, equipment status, and other information, and storing it in the database. The third step involves the closed-loop feedback module correcting the data. The data is then compared with the preset standard value. In this embodiment, the detection data are: SiO2 content 64.92%, Al2O3 content 17.95%, and moisture content 0.52%. The deviations are all within the allowable range (±0.1%-±0.5%). The PLC generates a fine-tuning instruction and feeds it back to the dual-stage gate metering module to fine-tune the metering chamber volume by 0.05%. At the same time, it sends a "parameter stable" signal to the batching system. If the detection data exceeds the allowable range, it is handled according to a multi-level response mechanism: if the deviation is >±0.5%, the feeding of this batch of raw materials is immediately suspended and the batching ratio is adjusted; if the deviation is ≤±0.1%, the current parameters are maintained.

[0035] 5. Cyclic Operation and Batch Calibration Phase After a single sample test is completed, the system automatically enters a cyclic operation mode, repeating the above quantitative, transport, test, and feedback process. Every 10 batches of raw materials are tested, the XRF spectrometer automatically starts batch calibration. During the calibration process, the system pauses testing and resumes operation after calibration is completed to ensure the stability of testing accuracy.

[0036] III. Effect Verification and Data Comparison This system operated continuously for 30 days, 24 hours a day, on a 150,000-ton-level high-end wind power fiberglass production line, testing a total of 1,440 batches of raw materials, with 5 samples per batch. The system's performance was verified by comparing it with traditional testing methods. Specific data are as follows: 1. Comparison of Quantitative Analysis and Detection Accuracy The quantitative error of traditional dual-stage gate valves is ±2.1%-±3.5%, while the quantitative error of this system is stable at ±0.05%-±0.1%, improving accuracy by more than 20 times. In traditional detection methods, XRF component detection is affected by moisture, with an error of ±0.8%-±1.2%. After correction by the coupling algorithm, the synergistic deviation between component and moisture detection in this system is ≤0.3%. The detection accuracy of key components such as SiO2 and Al2O3 reaches the ppm level, and the moisture detection accuracy is ±0.02%, meeting the requirements of high-end glass fiber production.

[0037] 2. Comparison of detection efficiency Traditional testing methods rely on manual material handling, transportation, and testing, with a single sample testing time of 30-40 minutes. This system achieves fully automated operation, with a single sample testing time of ≤10 minutes, improving efficiency by 8.5 times. It can test 288 more samples per day, adapting to the needs of large-scale production lines.

[0038] 3. Comparison of production losses and costs Traditional testing is disconnected from production, resulting in a raw material loss rate of 9.2%. This system, through closed-loop feedback throughout the entire process, optimizes batching parameters in real time, reducing the raw material loss rate to 2.8%. Based on a production capacity of 150,000 tons / year, this can reduce raw material loss by 960 tons annually, lowering production costs by approximately 1.8 million yuan. At the same time, it enables 24-hour unattended testing, reducing the number of testing personnel on a single production line from 5 to 1, and lowering labor costs by 80%.

[0039] 4. Comparison of product quality stability After 30 days of continuous operation, the performance deviation between batches of produced glass fiber precursors was controlled within ±1.8%, which is a significant improvement compared to traditional production (deviation ±3.5%-±5%). According to third-party testing, the tensile strength of the precursor reached 8.5GPa, which meets the mechanical performance requirements of glass fiber for high-end wind turbine blades.

[0040] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dual-stage gate-type quantitative robot collaborative XRF and moisture measurement system for glass fiber raw materials, characterized in that, include: The dual-stage gate intelligent quantitative module includes a feeding section, a metering section, a discharging section, an upper gate assembly, a lower gate assembly, a volume adaptive calibration unit, and an intelligent vibration anti-bridging unit. The metering section is located between the feeding and discharging sections. The upper and lower gate assemblies are respectively located at the junctions of the feeding section-metering section and the metering section-discharging section, forming a dynamically sealable quantitative metering cavity. This addresses the technical challenge of traditional fixed-volume metering methods failing to adapt to fluctuations in the bulk density of fiberglass raw materials. The volume adaptive calibration unit consists of a retractable cavity wall and a high-precision laser ranging sensor. The system is composed of components that can collect raw material stacking height data in real time and dynamically adjust the metering chamber volume in combination with a preset raw material stacking density model to achieve high-precision material dispensing with a quantitative error of ≤±0.1%, which is significantly lower than the error of more than ±2% of traditional double-stage gates. The intelligent vibration anti-bridging unit is assembled on the outside of the metering section and adopts a high-frequency low-amplitude vibration mode. Its vibration frequency and amplitude can be adaptively matched according to the particle size characteristics of glass fiber raw materials, which can not only completely eliminate the quantitative deviation caused by raw material bridging, but also avoid the raw material stratification problem caused by excessive vibration, thus ensuring sample uniformity. The industrial robot collaboration module includes a six-axis industrial robot, a composite end effector, a deep learning vision positioning unit, and a human-machine safety protection unit, constructing a seamless collaborative transfer mechanism between the quantitative material handling and dual detection modules. The composite end effector integrates a dual structure of vacuum adsorption and mechanical clamping, and can automatically switch operating modes according to the particle size of the glass fiber raw material, adapting to the transfer needs of various types of raw materials. The deep learning vision positioning unit consists of an industrial camera and a dedicated recognition algorithm, which can accurately identify the spatial coordinates and raw material posture of the quantitative module outlet, XRF detection station, and moisture detection station, with a positioning accuracy of ≤±0.2mm, ensuring the consistency of posture when the sample is transferred to the detection station, providing a guarantee for the accuracy of subsequent synchronous detection, and breaking through the limitation of traditional robots that can only achieve simple transfer. The XRF component detection module includes a high-performance XRF spectrometer, a dedicated protective detection chamber, a compatible sample carrier, and a batch-based automatic calibration unit, optimized for glass fiber raw material detection scenarios. The dedicated protective detection chamber features a radiation shielding layer with a lead equivalent of ≥2mm and an argon inert gas protection system, preventing radiation leakage and ensuring operational safety while effectively preventing raw material oxidation during detection, thus avoiding interference with the detection accuracy of key components such as SiO2 and Al2O3. The batch-based automatic calibration unit incorporates a standard sample library covering various glass fiber raw materials, including E-CR and high-silica materials. It automatically calibrates the instrument after testing every 10 batches of raw materials, ensuring stable component detection accuracy at the ppm level, meeting the stringent requirements of high-end glass fiber for component detection. The moisture detection module employs a high-frequency capacitive detection principle, combined with a constant temperature and humidity control chamber and a robot-assisted sample spreading mechanism, to achieve synchronous linkage with XRF component detection. The constant temperature and humidity control chamber can control the detection environment temperature at 25±1℃ and humidity at 50±5%RH, completely eliminating the interference of ambient temperature and humidity on moisture detection. The robot-assisted sample spreading mechanism, through linkage with the movement of an industrial robot, evenly spreads a quantitative amount of raw material onto the detection tray, with a spreading thickness error ≤0.5mm, ensuring the consistency of raw material thickness in the detection area. This allows the moisture detection range to cover 0.01%-20%, the detection accuracy to reach ±0.02%, and the detection time ≤3s, achieving synchronous detection of two parameters without delay. The central control and data processing module, including an industrial controller, a high-speed data acquisition card, a moisture-XRF cross-interference coupling analysis algorithm module, and a full-process data traceability unit, serves as the core hub for the system's collaborative operation. The coupling analysis algorithm module constructs a correlation database between moisture content and XRF characteristic spectral line intensity based on over 100,000 sets of glass fiber raw material samples. It uses a CNN convolutional neural network algorithm to fit a correction function, iteratively correcting the original detection data. This effectively eliminates the cross-interference of moisture content on XRF component detection, ensuring that the corrected data deviation is ≤0.3%, thus solving the technical bottleneck of disconnected component and moisture data and the inability to eliminate interference in traditional detection methods. The full-process data traceability unit uses a unique batch identifier to bind quantitative parameters, dual detection data, equipment operating status, and other information for each batch of raw materials and stores them in the cloud, achieving full-process traceability of the detection process and meeting the auditing requirements of high-end customers for raw material quality records. The closed-loop feedback module establishes a multi-level response control mechanism to achieve real-time linkage between testing data and the production system, breaking through the limitations of traditional testing and production being disconnected. This module compares the testing data processed by the central control module with preset standard values ​​to generate targeted control instructions. On the one hand, the instructions are transmitted in real time to the fiberglass raw material batching system to achieve dynamic optimization of the raw material ratio; on the other hand, they are fed back to the dual-stage gate intelligent quantitative module to complete the adaptive correction of the measurement parameters, forming a closed-loop control of the entire process of "quantification-detection-analysis-feedback-optimization", which significantly reduces the raw material loss rate.

2. The fiberglass raw material dual-stage gate quantitative robot collaborative XRF and moisture measurement system according to claim 1, characterized in that: The dual-stage gate intelligent quantitative module achieves high-precision and rapid quantitative measurement through graded action coordination. The workflow is as follows: the lower gate assembly closes to form a bottom seal of the metering chamber → the upper gate assembly opens to allow the raw material to enter the metering chamber → the laser rangefinder sensor detects the raw material accumulation height in real time and calculates the actual material intake based on the chamber volume and accumulation density model → the volume adaptive calibration unit fine-tunes the chamber volume to compensate for errors and ensure quantitative accuracy → the upper gate assembly closes to form a top seal of the metering chamber → the intelligent vibration anti-bridging unit is activated to break the arch with vibration parameters adapted to the characteristics of the raw material → the lower gate assembly opens to allow the raw material to fall accurately into the end effector. The single quantitative cycle is ≤8s, which is more than 3 times more efficient than the traditional quantitative process.

3. The fiberglass raw material dual-stage gate quantitative robot collaborative XRF and moisture measurement system according to claim 1, characterized in that: The industrial robot collaboration module adopts a dual control mode of "deep learning vision guidance + dynamic path planning". The robot end effector integrates a force sensor, which can collect the actual weight data of the raw materials in real time and feed it back to the central control module. It forms a two-way secondary verification with the measurement data of the dual-gate intelligent quantitative module, further controlling the quantitative comprehensive error within ±0.1%, which significantly improves the reliability compared with a single measurement method.

4. The fiberglass raw material dual-stage gate quantitative robot collaborative XRF and moisture measurement system according to claim 1, characterized in that: The core logic of the coupling analysis algorithm module is as follows: based on the natural correlation between the moisture content of glass fiber raw materials and the XRF spectral response, a massive sample correlation database is constructed. The inherent correlation of the data is mined through the CNN convolutional neural network algorithm and a dedicated correction function is fitted. The original detection data is iteratively corrected in multiple rounds to specifically eliminate the XRF component detection error caused by moisture interference. Finally, the synergy deviation between component and moisture detection data is ≤0.3%, solving the technical problem that the dual-parameter interference in traditional detection cannot be quantitatively eliminated.

5. The fiberglass raw material dual-stage gate quantitative robot collaborative XRF and moisture measurement system according to claim 1, characterized in that: The multi-level response mechanism of the closed-loop feedback module is as follows: when the detection data exceeds the preset deviation range of ±0.5%, the batching system parameters are immediately triggered for emergency adjustment and the feeding of the batch of raw materials is suspended to prevent batches of unqualified raw materials from entering production; when the deviation range is ±0.1%-±0.5%, the metering parameters are finely adjusted through the dual-stage gate intelligent quantitative module to achieve dynamic correction without affecting the continuity of production; when the deviation is ≤±0.1%, the current production parameters are maintained to operate stably, and the balance between quality and efficiency is achieved through graded precise control, so that the fiberglass raw material loss rate is reduced from the traditional 8%-12% to below 3%.

Images