Gallium arsenide solar cell surface crack visual detection system based on deep learning

Abstract

The invention discloses a visual detection system for surface cracks of gallium arsenide solar cells based on deep learning, which comprises the following parts: a four-degree-of-freedom mobile scanning portal frame, a laser emitting head, a near-infrared camera, a laser, a multimode optical fiber, a laser emitting head, a host and a multimode optical fiber, defect pictures with similar characteristics can be summarized according to needs to form a standard defect picture library, and battery piece defects in the production process can be detected in real time, classified and marked in position. The method is used for nondestructive detection of defects of the solar cell panel, and has the advantages of high speed, high efficiency and low cost.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent manufacturing instruments, and mainly relates to a deep learning-based visual detection system for surface cracks in gallium arsenide solar cells. Background Art

[0002] Solar cells are devices that utilize the photoelectric effect of semiconductor materials to directly convert solar energy into electrical energy. As one of the most promising new energy sources, solar energy boasts numerous advantages, including inexhaustible supply, zero pollution, zero emissions, high energy density, safety and reliability, and geographic independence. It is widely used in transportation and communications facilities, commercial lighting, residential solar power generation systems, automotive energy systems, and aerospace energy systems, becoming a crucial component of green energy development both domestically and internationally.

[0003] In the aerospace field, the energy requirements of spacecraft can be divided into two categories: fuel, used to maintain the spacecraft's attitude and orbital adjustments; and electrical energy, used to support normal spacecraft operations. Solar cells and batteries are the primary sources of spacecraft electrical energy, with solar cells occupying a central role. As spacecraft become more multifunctional and diverse, higher requirements are placed on the photovoltaic conversion efficiency of solar cells to ensure sufficient power to support their complex missions. Furthermore, since spacecraft must withstand the complex and changing space environment during their in-orbit service, such as radiation and high temperatures, solar cells must not only have high photovoltaic conversion efficiency but also excellent radiation and high-temperature resistance. Gallium arsenide, a direct bandgap semiconductor material, has a bandgap of approximately 1.43 eV at room temperature, close to the bandgap required for optimal conversion efficiency in solar cells, making it an ideal material for high-performance solar cells. Triple-junction gallium arsenide solar cells are made of three materials with different bandgap widths: the top cell, the middle cell, and the bottom cell. Each layer absorbs light in a different wavelength range, significantly improving the utilization of the solar spectrum and the solar cell's photovoltaic conversion efficiency. Furthermore, GaAs solar cells offer the advantages of light weight, strong radiation resistance, and excellent high-temperature resistance, making them the preferred choice for the aerospace industry. As a core component of solar cells, the quality of GaAs cells is crucial, directly impacting the stability of spacecraft on-orbit operations and the reliability of mission completion. Research has shown that the primary cause of on-orbit cell failure is often internal microcracks, fractures, and surface defects formed during the production process. These defects not only weaken the cell's mechanical strength and photoelectric conversion efficiency but can also lead to a decline in the performance of the entire solar cell system. Therefore, efficiently and reliably detecting surface defects in solar cells has become a critical technical challenge that needs to be addressed in the solar energy application sector.

[0004] Currently, the detection methods for cracks in solar cells mainly rely on electroluminescence (EL) imaging technology. This technology applies a reverse voltage across the two ends of the cell to generate non-equilibrium minority carriers, and when these carriers recombine with majority carriers, near-infrared light is emitted. The fluorescence intensity of the cell is related to the internal carrier concentration, and the fluorescence intensity in the crack defect area is usually lower than that in the normal area. Therefore, the defect location and nature can be determined by observing the intensity of the cell's luminescence. Traditional detection methods include manual visual inspection based on electroluminescence and CCD image detection based on machine vision. For example, an existing patent discloses a defect detection system for solar cell modules based on a convolutional neural network, which also relies on electroluminescence technology to complete the detection.

[0005] However, the existing technologies have significant limitations. Manual visual inspection relies on the subjective judgment of the inspector, with inconsistent standards, prone to misjudgment and missed detection; at the same time, the detection speed is slow, the efficiency is low, the cost is high, and it poses a potential hazard to human health, making it difficult to meet the industrial demand for rapid detection. Although CCD image detection based on machine vision has achieved a certain degree of automation, it requires a DC power supply to provide a bias voltage for electroluminescence, with complex operation, and since it is not a completely non-contact detection, it may cause damage to the cell. In addition, CCD image detection has insufficient resolution for small target cracks, and there is still room for improvement in terms of detection speed, missed detection rate, and misdetection rate.

[0006] To address the above problems, the present invention innovatively designs a vision detection system for surface cracks of gallium arsenide solar cells based on deep learning. The photoluminescence (PL) detection technology utilizes the non-contact characteristic of optical excitation of carrier recombination, and collects the optical signals of surface cracks of the cell through a high-resolution imaging device, without the need for electrode installation or bias voltage, avoiding potential damage to the cell. At the same time, by introducing deep learning algorithms, the degree of automation and accuracy of crack detection are further improved.

[0007] While improving the detection efficiency and reducing the cost, this technology overcomes the limitations of traditional detection technologies, especially showing significant advantages in small target defect recognition and low sample size scenarios. Its application not only provides technical support for the quality control of solar cells, but also injects new impetus into the development of China's green energy field. Summary of the Invention

[0008] The present invention belongs to the field of vision detection in the photovoltaic industry. Its purpose is to solve the technical problem of rapid and high-quality detection of surface cracks in gallium arsenide solar cells, replace manual visual inspection, and obtain high detection accuracy and speed in the case of fewer samples. The present invention designs and discloses a vision detection system for surface cracks of gallium arsenide solar cells based on deep learning.

[0009] According to some embodiments of the present invention, a deep learning-based visual inspection system for surface cracks in gallium arsenide solar cells is provided. A four-degree-of-freedom mobile scanning gantry has multi-degree-of-freedom motion capabilities and utilizes a marble-based gantry structure. The gantry integrates y- and x-axis linear motor drive modules, a z-axis ball screw lift module, and a z-axis rotation platform. The four-degree-of-freedom mobile scanning gantry is connected to a host computer via a gantry motor drive control cable. The system can achieve three-dimensional translation with a positioning accuracy of ±5μm and rotational scanning with a ±0.1° accuracy. According to some embodiments of the present invention, a deep learning-based visual inspection system for surface cracks in gallium arsenide solar cells is provided, wherein a 532 nm wavelength laser is configured and coupled to a beam expanding and homogenizing output head via a multimode optical fiber, outputting a TEM00 mode laser beam with a spot uniformity exceeding 95%. According to some embodiments of the present invention, a deep learning-based visual inspection system for surface cracks in gallium arsenide solar cells is provided. A near-infrared camera with a resolution of 1280×1024 is used, and its spectral response covers the photoluminescence band of gallium arsenide solar cells. A SWIR lens is used as the optical lens. The camera sensor dimensions are (4.8μm×1280)×(4.8μm×1024)=6.14mm×4.92mm. The working distance is tentatively set at 300mm, and the field of view is 80mm. The lens focal length f is 23.025mm. Therefore, a lens with a focal length of 25mm is selected. The lens working distance is between 300mm and ∞. The optical lens is compatible with the camera interface and used in combination.

[0010] According to some embodiments of the present invention, a deep learning-based visual detection system for surface cracks in gallium arsenide solar cells is provided, wherein the host is a Dongtian Core industrial computer, and the host communicates with a camera system via Ethernet communication, thereby enabling the collection and storage of image information.

[0011] Compared with traditional manual visual inspection technology, the present invention has the following advantages over the existing technology: The GaAs solar cell surface crack visual inspection system of the present invention is different from the existing manual visual inspection method. It has the advantages of high speed, high efficiency and low cost, saves labor costs, avoids visual fatigue of inspectors, and is beneficial to the occupational health of inspectors.

[0012] The GaAs solar cell surface crack visual inspection system of the present invention can realize the recognition of various types of defect images on the surface of the cell, and can summarize defect images with similar characteristics to form a standard defect library as needed; it can detect cell defects in the production process in real time and classify and mark their locations.

[0013] The visual inspection system for surface cracks of gallium arsenide solar cell wafers of the present invention is applicable to the identification and detection of surface crack defects of gallium arsenide solar cell wafers, and is also applicable to the detection of surface crack defects of silicon crystal solar cell wafers. The system is small in volume and light in weight, can be used in small-space places, and can be used portably.

[0014] The visual inspection system for surface cracks of gallium arsenide solar cell wafers of the present invention can effectively reduce the missed detection rate and error rate of cell wafer defects, and ensure the consistency and stability of the surface quality of the produced cell wafers. It replaces manual observation and detection, and greatly improves the detection speed and detection efficiency.

[0015] The visual inspection system for surface cracks of gallium arsenide solar cell wafers of the present invention adopts photoluminescence imaging detection technology, which has the advantages of non-contact, fast detection speed, intuitive detection results, etc. It can realize qualitative detection of crack defects of solar cell wafers, has a simple operation method, high efficiency, high equipment safety, strong feasibility, is beneficial to the health of detection personnel and environmental protection, etc. Description of the Drawings

[0016] Figure 1 is a schematic diagram of the principle of a visual inspection system for surface cracks of gallium arsenide solar cell wafers based on deep learning in the implementation of the present invention.

[0017] Reference numerals in the figure: 1-four-degree-of-freedom moving and scanning gantry, 2-laser emitting head, 3-near-infrared camera, 4-three-junction gallium arsenide solar cell wafer, 5-laser, 6-host computer, 7-camera signal transmission line, 8-gantry motor drive control line, 9-multimode optical fiber.

[0018] Figure 2 is a schematic diagram of the mechanical principle of a four-degree-of-freedom moving and scanning gantry in the implementation of the present invention.

[0019] Reference numerals in the figure: 1-1-y axis, 1-2-x axis, 1-3-z axis, 1-4-rotation platform around the z axis.

[0020] Figure 3 is a schematic diagram of the principle of the imaging unit in the implementation of the present invention.

[0021] Reference numerals in the figure: 2-laser emitting head, 3-near-infrared camera.

[0022] Figure 4 are surface cracks of gallium arsenide solar cell wafers collected in the implementation of the present invention Detailed Embodiments

[0023] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where 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 by referring to the accompanying drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation of the present invention. In the description of the present invention, it should be understood that the orientation descriptions, such as up, down, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In the description of the present invention, "a plurality of" means more than two. If there is a description of first and second, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features. In the description of the present invention, unless otherwise clearly defined, words such as "set", "installed", "connected", etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0024] The present invention will be further described in detail below with reference to the accompanying drawings: According to some embodiments of the present invention, a vision inspection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, wherein the four-degree-of-freedom moving and scanning gantry (1) has the ability of multi-degree-of-freedom movement, adopts a gantry structure with a marble base, integrates a y-axis linear motor drive module (1-1), an x-axis linear motor drive module (1-2), a z-axis ball screw lifting module (1-3), and a rotating platform (1-4) around the z-axis. The four-degree-of-freedom moving and scanning gantry is connected to the host (6) through the gantry motor drive control line (8). It can achieve three-dimensional translation with a positioning accuracy of ±5μm and rotational scanning of ±0.1°; According to some embodiments of the present invention, a vision inspection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, wherein a 532nm wavelength laser (5) is configured, and is coupled to the beam expander and homogenizer output head (2) through a multi-mode optical fiber to output a TEM00 mode laser beam, and the spot uniformity is more than 95%; According to some embodiments of the present invention, a vision detection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, wherein the near-infrared camera (3) has a resolution of 1280×1024, and its spectral response can cover the photoluminescence band of gallium arsenide solar cell wafers. The optical lens uses a SWIR lens. The sensor size of the camera used is (4.8μm×1280)×(4.8μm×1024)=6.14mm×4.92mm. The tentative working distance is 300mm, and the field of view size is 80mm. The focal length f of the lens is 23.025mm. Therefore, a lens with a focal length of 25mm is selected. The working distance of the lens is 300mm~∞. The optical lens is adapted to the interface of the near-infrared camera (3) and used in combination.

[0025] According to some embodiments of the present invention, a vision detection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, wherein the host (6) is an Eastfield Core i7 industrial computer. The host communicates with the camera system through Ethernet communication, and can realize the acquisition and storage of picture information.

Claims

1. A vision detection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, characterized in that: Four-degree-of-freedom moving and scanning gantry (1), the laser emitting head (2) and the near-infrared camera (3) are fixed to the z-axis (1-3) of the four-degree-of-freedom moving and scanning gantry (1) with screws. The triple-junction gallium arsenide solar cell (4) is placed on the rotation platform (1-4) that rotates around the z-axis of the four-degree-of-freedom moving and scanning gantry (1). The laser (5) is connected to the laser emitting head (2) through a multimode optical fiber (9), the mainframe (6), the camera signal transmission line (7), the gantry motor drive control line (8), and the multimode optical fiber (9).

2. A vision inspection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, characterized in that: The four-degree-of-freedom moving and scanning gantry (1) includes the y-axis (1-1), the x-axis (1-2), the z-axis (1-3), the rotational movement of the rotation platform (1-4) around the z-axis, and the bottom base (1-5). Among them, the x-axis (1-2) and the y-axis (1-1) are driven by linear motors, the z-axis (1-3) is driven by a ball screw to move up and down, the bottom base (1-5) is made of marble, and the four-degree-of-freedom moving and scanning gantry (1) is connected to the mainframe (6) through the gantry motor drive control line (8).

3. A visual inspection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, characterized in that: The laser (5) is a 532-nm green light source. The laser is introduced into the emitting head (2) through a multimode optical fiber (9), and the emitting head (2) expands and homogenizes the laser. The laser mode used for excitation detection is the TEM00 mode.

4. A vision detection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, characterized in that: The near-infrared camera (3) has a wavelength band of 300 - 1100 nm and can collect the excitation at a wavelength of 660 nm of the triple-junction gallium arsenide solar cell (4).

5. A vision detection system for surface cracks of gallium arsenide solar cell wafers based on deep learning, characterized in that: The mainframe (6) is an industrial PC. The mainframe (6) is linked to the gantry motor drive control line (8) and the camera signal transmission line (7) and controls the movement of the four-degree-of-freedom moving and scanning gantry (1) and collects the information on the near-infrared camera (3).

Images