Intelligent teaching system based on big language model and welding spot validity detection and grade evaluation
Through the Q&A subsystem based on the large language model and the PCB solder joint detection system, the problems of extensive course content, tight teacher resources and low welding inspection efficiency in the electronic process internship were solved, efficient solder joint detection and evaluation were achieved, and teaching quality and fairness were improved.
Patent Information
- Application Number
- CN202510347966.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-24
- Publication Date
- 2025-07-08
AI Technical Summary
In the existing electronic technology internship courses, the course content is extensive, the teacher resources are tight, the welding detection efficiency is low, and the evaluation is not objective, which affects learning efficiency and fairness.
The Q&A subsystem and PCB solder joint detection subsystem based on large language models are adopted, including professional knowledge base, image acquisition, processing and detection modules, to realize solder joint effectiveness detection and rating evaluation, and provide multimodal data and visual display.
It improves teaching efficiency, reduces the burden on teachers, provides objective welding joint evaluation, supports multi-professional courses, and improves students' learning effectiveness and fairness in evaluation.
Smart Images
Figure CN120278858A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of process practice teaching, and more particularly to an intelligent teaching system based on a large language model and solder joint effectiveness detection and grading evaluation. Background Art
[0002] Currently, electronic process practice courses face various challenges. Firstly, the course content covers a wide range, involving many course categories and knowledge points. When students encounter problems during the practice, it requires high comprehensive quality of the teaching teachers and students. In this case, a more efficient and systematic way is needed to support teaching activities to ensure that students can obtain sufficient guidance and help. Secondly, this course is open to multiple majors and is restricted by the teaching plans of each major, resulting in multiple majors conducting the course simultaneously within the same time period, making the teacher resources tense. A teacher needs to guide students in multiple classes, and this guidance mode is inefficient and cannot meet the needs of a large number of students. Therefore, a method that can reduce the burden on teachers and improve the guidance efficiency needs to be explored. In addition, an important link in the course is circuit soldering and debugging. When students solder PCB circuit boards, they often encounter problems such as ineffective solder joints and non-standard wiring. When the welded circuit cannot achieve the expected function, for the effectiveness detection of solder joints, students can only detect point by point through tools such as multimeters, which is inefficient. The quality of PCB board solder joints is the direct basis for teachers to evaluate students' soldering levels. Traditional assessment and evaluation methods rely on teachers' visual observation and the detection of equipment such as multimeters and oscilloscopes, which not only increases the workload of teachers but also has problems of non-objective and inconsistent evaluation. This not only affects the learning efficiency of students but also the fairness of course assessment. The existence of the above problems has greatly reduced the learning efficiency of students in the course and also greatly reduced the teaching effect. Therefore, how to provide an intelligent teaching system based on a large language model and solder joint effectiveness detection and grading evaluation is an urgent problem for those skilled in the art. Summary of the Invention
[0003] In view of this, the present invention provides an intelligent teaching system based on a large language model and solder joint effectiveness detection and grading evaluation to solve the problems existing in the prior art.
[0004] To achieve the above object, the present invention provides the following technical solutions:
[0005] A smart teaching system based on large language models and solder joint effectiveness detection and grading evaluation, including a Q&A subsystem and a PCB solder joint detection subsystem. Among them, the Q&A subsystem includes a professional knowledge base, a word vector matching module, and a text output module. The professional knowledge base stores multimodal professional materials related to electronic technology practice. The word vector matching module performs word vector similarity retrieval based on the questions input by students. The text output module generates professional answers for electronic technology practice based on prompt templates. The PCB solder joint detection subsystem includes an image acquisition module, an image processing module, a solder joint effectiveness detection module, a grading evaluation module, and a visualization display module. The image acquisition module is used to acquire the PCB solder joint images of students. The image processing module processes the acquired images. The solder joint effectiveness detection module detects the effectiveness of PCB solder joints based on the acquired images. The grading evaluation module grades the effective solder joints. The visualization display module is used to display the solder joint detection and grading evaluation results.
[0006] Optionally, the multimodal professional materials related to electronic technology practice include course textbooks on circuit analysis, digital electronics, analog electronics, and electronic technology practice, teachers' lecture notes, courseware, and practice guides for electronic technology practice. The forms of the materials include text, voice, pictures, and videos.
[0007] Optionally, the solder joint effectiveness detection module detects the effectiveness of PCB solder joints through one of the YOLO, R-CNN, and Fast R-CNN deep learning object detection algorithms.
[0008] Optionally, the grading evaluation module classifies the effective solder joints into four grades: excellent, good, qualified, and unqualified, and classifies the unqualified solder joints into one of less soldering, tip pulling, excessive soldering, and bridging.
[0009] Optionally, the visualization display module uses different colors to mark the effective solder joints, ineffective solder joints, and solder joints with different grading levels.
[0010] Optionally, the solder joint effectiveness detection module is deployed on edge devices.
[0011] As can be seen from the above technical solutions, compared with the prior art, the present invention provides a smart teaching system based on large language models and solder joint effectiveness detection and grading evaluation, which has the following beneficial effects: The Q&A subsystem adopted by the present invention should have stronger applicability in vertical fields, which can help students preview before practice and assist in Q&A during the practice process, record the interactive data between the Q&A subsystem and students, provide teaching feedback for teachers, so as to understand the learning situation and needs of students, and support multi-modal data, such as pictures, videos, audio, and text information, etc.; By using computer vision technology and the deep learning object detection algorithm, based on the automatic detection and feature extraction of solder joints, the effectiveness detection of solder joints and the grading evaluation of solder joints are realized. A visual system interface is built to display the solder joint detection results and grading evaluation in an intuitive manner. The system can provide a detailed solder joint detection report, including the number of solder joints of each grade and the number of various types of unqualified solder joints.
[0012] It enables students to improve the solder joint process targeted, and at the same time provides an objective basis for teachers to grade students' soldering processes. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.
[0014] Figure 1 It is a schematic diagram of the smart teaching system of the present invention;
[0015] Figure 2 It is a flow chart of PCB solder joint detection of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.
[0017] An embodiment of the present invention discloses a smart teaching system based on large language models and solder joint effectiveness detection and grading evaluation, as Figure 1As shown in the figure, it includes a question-and-answer subsystem and a PCB solder joint detection subsystem. Among them, the question-and-answer subsystem includes a professional knowledge base, a word vector matching module, and a text output module. The professional knowledge base stores multi-modal professional materials related to electronic technology practice. The word vector matching module performs word vector similarity retrieval based on the questions input by students. The text output module generates professional answers for electronic technology practice based on the prompt word template. The PCB solder joint detection subsystem includes an image acquisition module, an image processing module, a solder joint validity detection module, a grade evaluation module, and a visualization display module. The image acquisition module is used to acquire the PCB solder joint images of students. The image processing module processes the acquired images. The solder joint validity detection module detects the validity of the PCB solder joints based on the acquired images. The grade evaluation module grades the valid solder joints. The visualization display module is used to display the solder joint detection and grade evaluation results.
[0018] In the embodiment of the present invention, the design purpose of the question-and-answer subsystem is to provide students with relevant materials and tutorials for the electronic technology practice course, facilitate students' independent learning, and at the same time answer various questions encountered by students in electronic technology practice in real time, such as the working principle of circuits, the use of common experimental instruments such as electronic devices, oscilloscopes, and signal sources.
[0019] The question-and-answer subsystem is mainly implemented based on the training of large language models in specific vertical fields. Generally, existing large language models, such as DeepSeek, ChatGLM, ChatGPT, etc., are used to achieve question and answer in the vertical field of electronic technology practice through fine-tuning or plugging in relevant domain knowledge bases. By establishing a knowledge base for electronic technology practice and classifying and storing questions and answers, the user interface of the question-and-answer subsystem is simple and clear, which is convenient for students to input questions and view answers. At the same time, it displays recommendations for relevant questions and answers to guide students to carry out independent learning. At the same time, some user usage records are output for teachers to summarize the difficulties of students' learning, and then provide more targeted guidance.
[0020] The question-and-answer subsystem vectorizes and stores the local knowledge base through the RAG technology, and combines with existing large language models (such as ChatGLM) to achieve semantic retrieval and generation: when users ask questions, they first retrieve knowledge base fragments, and then input them into the model to generate accurate answers, realizing an intelligent question-and-answer system with enhanced professional knowledge.
[0021] The design purpose of the PCB solder joint detection subsystem is to detect the validity of the solder joints on the PCB board, detect solder joints with obvious defects, and then distinguish valid solder joints from invalid solder joints. At the same time, it can also evaluate the grades of valid solder joints. Visualization display can help students improve their soldering skills. The generated solder joint detection and grade evaluation report can also assist teachers in grading students' soldering skills.
[0022] Furthermore, multimodal professional materials related to electronic process internship include course materials for circuit analysis, digital electronic technology, analog electronic technology and electronic process internship, teachers' electronic process internship handouts, courseware, and internship guides, and the materials are in the form of text, voice, pictures, and videos.
[0023] Furthermore, the solder joint validity detection module detects the validity of the PCB solder joints by using one of the YOLO, R-CNN and Fast R-CNN deep learning target detection algorithms.
[0024] Furthermore, the grade evaluation module classifies effective solder joints into four grades: excellent, good, qualified, and unqualified, and classifies unqualified solder joints into one of insufficient solder joints, pulled tips, excessive solder joints, and continuous solder joints.
[0025] In the embodiment of the present invention, the grade evaluation module is trained by a large number of data sets with solder joint grade classification. The specific evaluation criteria are the evaluation criteria for excellent solder joints, good solder joints, qualified solder joints and unqualified solder joints set by teachers with reference to the evaluation criteria for solder joint quality in the electronic process internship course. The classification of unqualified solder joints is also set by teachers with reference to the relevant standards of the electronic process internship course.
[0026] Furthermore, the visual display module uses different colors to mark effective solder joints and invalid solder joints as well as solder joints of different rating levels.
[0027] In the embodiment of the present invention, the statistics of the number of solder joints of different levels can be implemented by using a multi-target tracking algorithm such as Deepsort to count the number of solder joints of different levels.
[0028] Furthermore, the solder joint validity detection module is deployed on the edge device.
[0029] In an embodiment of the present invention, the solder joint validity detection module is deployed on edge devices with low performance requirements, such as embedded chips, mobile phones, tablets, etc. It is low-cost, universal and portable, and convenient for teachers or students to use in laboratories, dormitories, classrooms, etc.; when the solder joint validity detection module is deployed on a computer, it can be connected to a mobile phone or computer camera to input PCB board solder joint image data, supporting multi-source information collection methods.
[0030] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0031] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A smart teaching system based on large language models and solder joint effectiveness detection and grading evaluation, characterized in that, It includes a question-and-answer subsystem and a PCB solder joint detection subsystem. Among them, the question-and-answer subsystem includes a professional knowledge base, a word vector matching module, and a text output module. The professional knowledge base stores multimodal professional materials related to electronic technology practice. The word vector matching module performs word vector similarity retrieval based on the questions input by students. The text output module generates professional answers for electronic technology practice based on the prompt word template. The PCB solder joint detection subsystem includes an image acquisition module, an image processing module, a solder joint validity detection module, a grade evaluation module, and a visualization display module. The image acquisition module is used to acquire the PCB solder joint images of students. The image processing module processes the acquired images. The solder joint validity detection module detects the validity of PCB solder joints based on the acquired images. The grade evaluation module grades the valid solder joints. The visualization display module is used to display the solder joint detection and grade evaluation results.
2. The intelligent teaching system based on the large language model and the solder joint effectiveness detection and grade evaluation according to claim 1, characterized in that, The multimodal professional materials related to electronic technology practice include course textbooks on circuit analysis, digital electronic technology, analog electronic technology, and electronic technology practice, teachers' lecture notes, courseware, and practice guides for electronic technology practice. The forms of the materials include text, voice, pictures, and videos.
3. The intelligent teaching system based on the large language model and the solder joint effectiveness detection and grade evaluation according to claim 1, characterized in that, The solder joint validity detection module detects the validity of PCB solder joints through one of the deep learning object detection algorithms of YOLO, R-CNN, and Fast R-CNN.
4. The intelligent teaching system based on the large language model and the solder joint effectiveness detection and grading evaluation according to claim 1, wherein The grade evaluation module classifies the valid solder joints into four grades: excellent, good, qualified, and unqualified, and classifies the unqualified solder joints into one of less soldering, tip pulling, excessive soldering, and bridging.
5. The intelligent teaching system based on the large language model and the solder joint effectiveness detection and grading evaluation according to claim 1, wherein The visualization display module uses different colors to mark the valid and invalid solder joints and the solder joints of different rating grades.
6. The intelligent teaching system based on the large language model and the solder joint effectiveness detection and grade evaluation according to claim 1, characterized in that, The solder joint validity detection module is deployed on edge devices.