Physical electricity exploration experiment intelligent guiding system based on cognitive dynamic perception

Abstract

The invention discloses a physical electricity exploration experiment intelligent guiding system based on cognitive dynamic perception. The system is composed of four modules: 1) a self-adaptive exploration guiding module which constructs a self-adaptive dialogue engine based on a large language model, realizes personalized configuration and dynamic deduction of an experiment process, and evaluates the knowledge mastering degree of students in real time through a cognitive diagnosis algorithm; 2) an operation behavior analysis module, which is used for realizing experiment operation action analysis based on a visual large model and a normative recognition algorithm, and generating a multi-dimensional experiment evidence chain; 3) an intelligent experiment platform module which realizes digital design and automatic evaluation of circuit topology through a visual platform and a feature matching algorithm; and the evidence-based evaluation module is used for carrying out data fusion analysis based on the multi-modal large model, generating image-text historical records, carrying out experimental quantitative scoring, carrying out error operation attribution analysis and carrying out scientific literacy development evaluation. The system is helpful for improving the intelligent level of electricity experiment teaching, and provides objective data support for teaching decision making.

Description

Technical Field

[0001] The present invention relates to the fields of educational technology and artificial intelligence, and in particular to an intelligent guidance system for physical and electrical exploratory experiments based on a large language model, a large visual model, and cognitive diagnosis. Background Art

[0002] Inquiry-based experiments play an important role in the teaching of physical electricity, helping to cultivate students' scientific inquiry ability and experimental thinking. However, the current teaching model of physical electricity experiments still has obvious deficiencies in theoretical guidance and intelligent support. Specifically, there are three major systemic defects:

[0003] 1. Discrete cognitive guidance: Traditional classrooms employ a linear process of "teacher demonstrates, student imitates," leading to fragmented development of experimental design thinking. A 2023 study found that only a small percentage of exploratory experiments in junior high school physics can be completed independently, reflecting a significant disconnect between theoretical guidance and practical application.

[0004] 2. Lag in process evaluation: Existing experimental grading methods often rely on a reverse assessment of final results, lacking a multi-dimensional cognitive diagnosis of the operational process. Traditional experimental operation evaluation methods suffer from strong subjectivity and delayed feedback, making it difficult to fully reflect students' actual operational capabilities and thought processes.

[0005] 3. Rigidity of teaching resources: While mainstream virtual simulation platforms (such as PhET and Labster) can visualize experiments, their pre-set experimental processes inhibit independent inquiry. For example, a 2020 study found that while Labster simulations helped students understand relevant concepts and data collection methods, they were limited in preparing students for more complex quantitative analysis or interpreting data sets.

[0006] Among existing physical electrical experiment teaching assistance systems, virtual simulation experiment platforms are relatively common (e.g., CN109697920A). These platforms are mainly based on virtual experiment platforms developed on the Web or Unity, which simulate the experimental process and demonstrate the operation. Although they can intuitively reproduce experimental phenomena, they have difficulty recreating the experimental operation details in real scenarios. They also lack real-time interactive analysis and intelligent feedback on students' operation behaviors, making them difficult to meet the teaching requirements of inquiry-based experiments.

[0007] Experimental teaching systems based on augmented reality (AR) technology can intuitively display physical phenomena such as current flow and node voltage, enhancing the visualization of experimental phenomena. However, such systems typically rely on specialized hardware devices, such as AR glasses or sensor equipment, which are costly and complex to operate, making them difficult to adapt to the needs of popular teaching.

[0008] Experimental teaching equipment based on sensors and electronic modules can collect physical quantities such as current and voltage in real time and visualize the data. However, this type of equipment is primarily used for physical quantity measurement and data analysis, lacking a systematic design to cultivate experimental thinking. This fails to form a complete teaching loop and is not conducive to improving students' comprehensive inquiry capabilities.

[0009] Several multimodal electrical experiment systems have emerged in existing research (e.g., CN118968382A and CN118898533A). While these systems utilize multimodal data and large-scale modeling to enable post-experiment evaluation and summary generation, they still suffer from significant shortcomings: first, they lack intelligent guidance mechanisms during the experiment; second, they are unable to provide real-time dynamic feedback during the experiment. This post-experiment evaluation model struggles to meet the full-process guidance requirements of exploratory experiments.

[0010] The root cause of these technical shortcomings lies in the failure to establish a cross-modal correlation model between "behavior-cognition-literacy." Therefore, developing a physics and electricity inquiry-based experiment guidance system that integrates theoretical guidance, real-time feedback, intelligent scoring, and the cultivation of inquiry skills has important practical significance and broad application prospects for improving students' experimental inquiry capabilities, reducing teachers' workload, and enhancing teaching effectiveness. Summary of the Invention

[0011] This invention aims to establish a teaching paradigm for physical electrical experiments within a multimodal cognitive computing framework, providing a closed-loop experimental guidance system based on artificial intelligence. By building a closed technical loop of "cognitive state perception - dynamic dialogue guidance - multi-source evidence assessment," this system achieves intelligent reconstruction of the entire experimental teaching process. This primarily addresses three technical challenges inherent in traditional teaching models: lags in cognitive diagnosis, a lack of personalized guidance, and incomplete competency assessment.

[0012] The purpose of the present invention is mainly achieved through the following technical solutions:

[0013] The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception of the present invention mainly includes the following four modules: a conversational exploration process guidance module, an experimental operation video intelligent analysis module, a paperless experiment design and data recording module, and an experiment evaluation and evidence-based module.

[0014] 1) Adaptive Inquiry Guidance Module

[0015] (1) Overall process control of the experiment:

[0016] Based on predefined experimental steps, users can customize the experimental process and dynamically generate question dialogues, guided dialogues, and feedback dialogues based on experimental progress and feedback. Specifically, it includes:

[0017] (11) Based on predefined experimental step types, users are allowed to define the overall experimental process;

[0018] (12) Based on the experimental process and feedback, control the type of dialogue generated.

[0019] (2) Dialogue Generation:

[0020] (21) Under the control of (12), three types of dialogue content, question dialogue, guided dialogue and feedback dialogue, are generated based on the large language model, involving the fine-tuning, deployment and calling of the large model;

[0021] (22) Question dialogue refers to the dialogue content that matches the experimental theme and combines relevant experimental knowledge points to ask students questions;

[0022] (23) Guiding dialogue is a prompting dialogue that combines the experimental process and prompts students to design the experiment, fill in the experimental data, draw the experimental images, and perform the actual experimental operation steps;

[0023] (24) Feedback dialogue is the feedback dialogue content generated by the large language model to receive students' answers to questions, experimental design, experimental data filling, experimental image drawing and experimental physical operation results and evaluate the results.

[0024] 2) Operation behavior analysis module

[0025] (3) Experimental behavior analysis:

[0026] Through the large visual model, the experimental operation is intelligently detected, key links such as circuit connection, zero adjustment operation, sliding state, switch state and desktop arrangement are identified, and the correctness of the operation is judged. Specifically, it includes:

[0027] (31) Based on the visual large model, key behaviors and equipment states such as circuit connection, desktop tidying, zero adjustment, sliding rheostat status and switch status during the actual operation of the experiment are detected, and the detection results are used as the actual operation results received by the feedback dialogue (24);

[0028] (32) Determine the correctness of circuit connection, identify the connection status of the actual circuit, and judge whether the connection meets the experimental requirements;

[0029] (33) Determine whether the desktop is ready. After the students have completed the desktop tidying steps, determine whether the experimental equipment has been arranged and placed in the designated area.

[0030] (34) Zeroing behavior detection: detecting whether the ammeter and voltmeter perform zeroing operations during the students’ circuit connection process;

[0031] (35) Sliding state detection, detecting the position of the slider of the rheostat when students connect the circuit, especially paying attention to whether the slider is at the position of maximum resistance;

[0032] (36) Switch status detection: detect whether the switch is in the open state or the closed state during the student's circuit connection process.

[0033] (4) Real-time video recording:

[0034] Automatically record experimental operation videos as experimental graphic history and scoring basis, and store them in the background for later analysis. Specifically including:

[0035] (41) When students are operating objects, the operation videos from multiple perspectives are recorded in real time and saved in the background to provide video evidence generated by historical records for (91)

[0036] 3) Intelligent experiment platform module

[0037] (5) Table and circuit design:

[0038] It supports students to flexibly create experimental data tables, set the number of rows, columns, table headers and cell contents. At the same time, it supports circuit component selection and connection to build a complete circuit design diagram. Specifically including:

[0039] (51) Table design methods and tools that support students to independently set the number of rows, columns, headers, and content of a table;

[0040] (52) Methods and tools that support students in independently selecting circuit components, flexibly placing them, and connecting them using wires to construct complete circuit designs.

[0041] (6) Data filling and image drawing:

[0042] Provides paperless filling and graph drawing tools for experimental data, supporting voltage-current and current-resistance relationship curve drawing. Specifically including:

[0043] (61) A tool that supports students in filling in data in the table designed in (51).

[0044] (62) Support students to draw a visualization method and tool for the relationship between voltage, current, and resistance based on the guided dialogue in (23) and the data filled in (61). Students can independently control the data labels, units, and value ranges of the horizontal and vertical axes, draw based on the data points in (61), and draw the relationship curve based on the drawn points.

[0045] (7) Correctness judgment:

[0046] Check the rationality of submitted forms, circuit diagrams, and graphs, determine the validity of data and compliance with physical laws, and provide feedback. Specifically including:

[0047] (71) Taking the tables, circuits, tabular data, and images submitted by students in (51)(52)(61)(62) as input, analyze the rationality of the submitted combination and provide feedback to (24);

[0048] (72) The evaluation of table design involves whether the number of rows, columns, headers, and cell contents of the table meet the experimental requirements in (23) the guided dialogue.

[0049] (73) The circuit design evaluation involves whether the component types, number of components, and wiring of the components in the circuit logic diagram in (52) meet the experimental requirements in the guided dialogue in (23).

[0050] (74) Evaluation of table data: Based on the voltage, current and resistance values ​​read by students through the electric meter and fixed resistor during actual operation, determine whether the filled data is within a reasonable range and meets the electrical laws.

[0051] (75) Curve image evaluation: evaluate the coordinate axis units, labels, value ranges, the offset of the plotted points relative to the data positions in (61) in the curve image drawn by the student in (62), and the rationality of the deviation between the student's relationship curve and the correct fitted curve.

[0052] 4) Evidence-based evaluation module

[0053] (8) Experimental scoring and score attribution:

[0054] Based on the analysis of experimental operation and paperless data recording, the experimental process is scored. Combined with the experimental literacy assessment, a comprehensive evaluation of experimental operation and experimental literacy is generated. Specifically including:

[0055] (81) Based on the intelligent analysis of the experimental operation video, the feedback of the paperless experimental design and data recording, and the answers to the questions generated by the students (22), the key steps of the experimental process are scored, and combined with the experimental graphic history records generated by (91), the scoring situation and the scoring evidence record are linked.

[0056] (82) Experimental scoring is based on the core competencies of the experiment. Experts formulate evaluation indicators and score multiple steps, including conjecture and design of experiments, circuit connection, exploration of the relationship between current and voltage, exploration of the relationship between current and resistance, instrument inspection and collation, summary and induction, and communication and discussion.

[0057] (83) Score attribution links the score points of each link in (82) and the experimental graphic history in (91), providing graphic basis for the score

[0058] (9) Experimental graphic and text history generation:

[0059] Automatically generate an experimental report containing experimental operation videos, experimental data, curve graphs and scoring results for students to query and teachers to evaluate. Specifically including:

[0060] (91) Based on the interactive records of the entire experimental process, such as dialogues, videos, tables, circuit diagrams, and curve images, a multimodal large model is used to extract and generate experimental graphic and text historical records for score attribution, record query, and summary generation.

[0061] (10) Experimental literacy analysis and summary:

[0062] Based on the scoring results and experimental records, an experimental summary report is generated, including a review of the experimental steps, operational evaluation, improvement suggestions, and directions for improving literacy. Specifically, it includes:

[0063] (101) Using the historical records of (91) and combining them with the core competencies of physical electrical inquiry experiments, an analysis and summary text is generated that includes evaluations of each step of the experiment, evaluations of experimental competencies, an overall summary of the experiment, and suggestions for student operations.

[0064] Furthermore, the specific operation process of this system includes the following steps:

[0065] S1, login experiment;

[0066] S2, process control loads the customized overall experimental process;

[0067] S3. Process control combines the large language model, the overall process of the loaded experiment, and the result feedback during the experiment to control the dialogue generation;

[0068] S4. If the generated dialogue is a question dialogue, the student answers the generated question and gives the answer. The system records the student's answer and judges its correctness. The answer result is used as feedback to return to the process control dialogue generation stage;

[0069] S5. If the dialogue is a feedback dialogue, the student obtains the feedback text of the student submission result generated by the large language model. The student clicks Next and returns to the process control dialogue generation stage;

[0070] S6. If the dialogue is a guided dialogue without an experiment end prompt, the student performs the relevant experimental design or physical operation according to the guided dialogue content and submits the operation results. The system records the results and evaluates them, and uses the evaluation results as feedback to return to the process control dialogue generation stage;

[0071] S7. If the dialogue is a guided dialogue with a prompt indicating the end of the experiment, the system automatically generates a document with the experiment score, graphic and text history records, and experimental literacy analysis and summary. The student obtains the document and the experiment ends.

[0072] The beneficial effects of the present invention are:

[0073] 1. Save labor costs and weaken the traditional teacher role: By using large model technology to implement tasks such as experimental guidance, problem generation, and feedback dialogue, the system can complete experimental guidance and evaluation without the participation of teachers, greatly saving labor costs.

[0074] 2. Instant feedback mechanism to enhance student knowledge acquisition: Through intelligent dialogue generation and real-time video analysis, the system provides immediate feedback on key knowledge points after students perform operations or answers, enabling them to correct errors while experimenting, effectively preventing the accumulation of operational deviations. Through real-time dialogue and guidance, students can always understand experimental requirements and key operational points during the experiment, enhancing experimental effectiveness and knowledge acquisition.

[0075] 3. A more rational scoring process improves evaluation accuracy and fairness: This paper constructs a multimodal data scoring and attribution system that integrates a large language model and a large visual model. It can integrate multiple information such as experimental operation videos, paperless data, and student responses to score experimental operations in real time. Compared with traditional manual scoring, the system is more objective and fair, and can automatically generate scoring basis and operation analysis reports, providing teachers and students with a clear evaluation process and improvement suggestions.

[0076] 4. Paperless experiment design improves user experience and teaching efficiency: This system provides a visual user interface that supports paperless experiments such as form filling, circuit design, and curve drawing. The system guides students through experimental steps and data recording directly on the electronic interface, reducing the burden of writing paper lab reports and optimizing experimental data management and analysis processes. Paperless operation also reduces laboratory consumables, making it environmentally friendly and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0078] Figure 1 This is a schematic diagram of the framework of the intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception;

[0079] Figure 2 This is a schematic diagram of the operation process of this system;

[0080] Figure 3 It is the main interactive interface of this system;

[0081] Figure 4 Draw an interface for the current-voltage-resistance graph;

[0082] Figure 5 Design the interface for the table;

[0083] Figure 6 Design interfaces for circuit diagrams;

[0084] Figure 7 Provide experimental scoring and score attribution in the experimental report;

[0085] Figure 8 This is the literacy analysis and summary part of the experimental report. DETAILED DESCRIPTION

[0086] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0087] The operation process of the present invention is as follows Figure 2 As shown, students use Figure 3 The dialog interface shown is used to interact with the program. The operation process includes the following steps:

[0088] S1, login experiment;

[0089] S2, process control loads the customized overall experimental process;

[0090] S3. Process control combines the large language model, the overall process of the loaded experiment, and the result feedback during the experiment to control the dialogue generation;

[0091] S4. If the generated dialogue is a question dialogue, the student answers the generated question and gives the answer. The system records the student's answer and judges its correctness. The answer result is used as feedback to return to the process control dialogue generation stage;

[0092] S5. If the dialogue is a feedback dialogue, the student obtains the feedback text of the student submission result generated by the large language model. The student clicks Next and returns to the process control dialogue generation stage;

[0093] S6. If the dialogue is a guided dialogue without an experiment end prompt, the student performs the relevant experimental design or physical operation according to the guided dialogue content and submits the operation results. The system records the results and evaluates them, and uses the evaluation results as feedback to return to the process control dialogue generation stage;

[0094] S7. If the dialogue is a guided dialogue with a prompt indicating the end of the experiment, the system automatically generates a document with the experiment score, graphic and text history records, and experimental literacy analysis and summary. The student obtains the document and the experiment ends.

[0095] In the present invention, a framework of intelligent guidance system for physical electrical exploration experiment based on cognitive dynamic perception is as follows Figure 1 As shown, it includes the following modules:

[0096] 1) Adaptive Inquiry Guidance Module: Builds an adaptive dialogue engine based on a large language model to enable personalized configuration and dynamic deduction of experimental processes, and uses cognitive diagnostic algorithms to assess students' knowledge mastery in real time. Specifically, it includes:

[0097] (1) Overall experimental process control;

[0098] (2) Dialogue generation;

[0099] 2) Operational Behavior Analysis Module: This module uses a large visual model architecture to capture experimental operation motions, combines it with an operational normative recognition algorithm to perform real-time video analysis, and simultaneously generates a multi-dimensional experimental evidence chain with a timestamp. Specifically, it includes:

[0100] (3) Experimental manipulation behavior analysis;

[0101] (4) Real-time recording of experimental videos from multiple perspectives;

[0102] 3) Intelligent Experimental Platform Module: This module implements digital design of circuit topology and data curve plotting through a visual programming platform, and automatically evaluates experimental parameters based on a feature matching algorithm. Specifically, it includes:

[0103] (5) Paperless forms and circuit design;

[0104] (6) Paperless data filling and image drawing;

[0105] (7) Judgment of the correctness of tables, images, and circuit designs;

[0106] 4) Evidence-based evaluation module: Based on a multimodal large model, data fusion analysis is performed to generate historical records including graphics and text, supporting quantitative scoring of experimental results, multi-dimensional attribution analysis of erroneous operations, and scientific literacy development assessment; specifically, it includes:

[0107] (8) Experimental scoring and score attribution;

[0108] (9) Generation of experimental graphic and text historical records;

[0109] (10) Analysis and summary of experimental literacy.

[0110] 1. Overall process control of the experiment:

[0111] 1) The overall process control of the experiment uses meta-operations to define basic behaviors, and the configuration file provides a custom configuration path based on a table-driven process control method.

[0112] 2) First, a set of experimental operation meta-operations are predefined. The meta-operations of the experiment are the most basic experimental steps closely related to the physical electricity inquiry experiment, and are an abstraction of the basic actions and state changes during the experiment. In this embodiment, the following meta-operation types are defined: start the experiment, students answer questions, circuit diagram design, connect the actual circuit, zero the meter, open the switch, close the switch, change the resistance of the sliding rheostat, design the voltage-current table, design the current-resistance table, fill in the voltage-current data, fill in the current-resistance data, draw the voltage-current image, draw the current-resistance image, organize the desktop, generate the experiment summary document, and end the experiment. Each meta-operation also has a completion status event corresponding to the operation, such as meta-operation correct and meta-operation error. Based on the received completion status event and the current meta-operation, the next meta-operation can be automatically selected.

[0113] 3) The configuration file uses a structured JSON format to describe the experimental meta-operation steps, prompt words for guiding prompt generation, meta-operation feedback prompt words, and state transition rules, and structuredly stores the operation sequence, prompt template, and state transition matrix.

[0114] 4) During actual operation, a state machine-based process control is used, triggering state transitions based on operation completion events. This supports abnormal operation detection and adaptive process adjustments. This model covers the entire closed-loop management process from experiment initiation (S0) to summary generation (S16).

[0115] 2. Dialogue Generation:

[0116] 1) Dialogue generation relies on a large language model. First, the dialogue type and content are designed. Then, through data collection and preprocessing, a corpus of experimental-related data is constructed, containing physical electrical knowledge, experimental instructions, and dialogue examples. Model fine-tuning utilizes specialized training for experimental teaching scenarios, enabling the model to generate guided, feedback-based, and exploratory dialogues. During deployment, the model communicates with the experiment control module via an API interface, and is controlled by the process control module to generate different types of dialogues.

[0117] 2) This embodiment divides the dialogue content into three types: question dialogue, guided dialogue, and feedback dialogue. The question dialogue is used to generate electrical experiment questions. The question content involves daily electrical phenomena introduced by the scenario, key points to consider in the experimental design process, precautions during the actual operation process, and reflections and summaries after the experiment. The guided dialogue is a prompt text for other meta-operations besides answering questions, reminding students which step to start and precautions during the operation. After the students have answered the questions and completed the relevant design behaviors and actual operations, the large model generates feedback on the question answers and experimental operation status based on the completion status.

[0118] 3) The large language model fine-tuning dataset for this embodiment is taken from 1013 middle school physics experiment questions. Each question data consists of the question stem, answer, and answer analysis. First, the questions are classified into scenario introduction questions, experimental design questions, physical operation questions, and thinking and summary questions, corresponding to the four types of question dialogues. Then, based on the question type and answer, the question stem is used to generate a large language model prompt text, which is used as the input for question generation to generate similar questions. At the same time, the generated feedback dialogue results are optimized based on the answer analysis.

[0119] 4) This example fine-tunes the 14B parameter DeepSeek-R1 model distilled by Qwen. During the fine-tuning process, the dedicated dataset constructed for the physical electrical exploratory experiment described above was used. During the model fine-tuning phase, the LoRA (Low-Rank Adaptation) algorithm was used for parameter optimization. Synchronization with the process control module was achieved through a RESTful API.

[0120] 3. Experimental Operation Behavior Analysis:

[0121] 1) The behavior analysis module uses a large visual model to detect objects and identify behaviors in the experimental scene. This includes image recognition and behavioral discrimination of experimental equipment, circuit connections, rheostat adjustments, and switch states. Through feature extraction and behavioral sequence analysis, the module determines experimental meta-operations and feeds the analysis results back to the process control module.

[0122] 2) Device detection and state recognition: Detection models are used to detect the location and category of components in the experimental scenario, providing further data support for behavior recognition.

[0123] 3) Training a large visual model requires a large amount of object detection data from physical electrical experiment scenarios. To this end, this example collected a large number of electrical experiment images from various physical electrical experiment scenarios and manually annotated them. The annotated objects in the images include electrical components: voltmeters, ammeters, rheostats, resistors, switches, power supplies, screwdrivers, as well as small component terminals and wire ends.

[0124] 4) This embodiment is based on the object detection model training. During the training process, a transfer learning strategy is adopted to apply the pre-trained weights to the experimental scene, further optimizing the model's performance on physical electrical experiment images to ensure its application effect in actual teaching environments.

[0125] 5) For detection of the switch state and the sliding rheostat state, this embodiment utilizes a large visual model to perform real-time recognition of the open or closed state of the switch and the position of the sliding rheostat slider.

[0126] 6) Determining zeroing behavior focuses on monitoring the process of connecting the circuit from start to finish. During this process, the positional changes of the hand, screwdriver, meter, and meter pointer are tracked in real time. Specifically, the visual macro model detects the distance between the screwdriver tip and the meter's zeroing screw. Furthermore, the movement of the pointer on the dial, combined with the hand's manipulation characteristics, allows for a comprehensive assessment of whether the student has correctly zeroed the meter.

[0127] 7) To determine the position of the slider of a sliding rheostat, you first need to build a coordinate system that matches the characteristics of the sliding rheostat. Use the connection line of the terminal of the sliding rheostat rail as the reference axis of the coordinate system, and establish an orthogonal axis with the direction of the resistor wire terminal. Next, project the coordinates of the resistor wire terminal of the connecting wire into the coordinate system to confirm the direction and position of the maximum resistance. At the same time, the position of the slider is also mapped to the same coordinate system, and combined with the position relationship of the slider relative to the direction of the maximum resistance, the current resistance state of the sliding rheostat is finally determined. The calculation formula is: Where R is the calculated resistance value, R max is the maximum resistance of the sliding rheostat, P slider is the projection of the position of the slide under the reference axis in the constructed sliding coordinate system, P max P is the projection of the position of the resistance wire terminal without a wire connected to it under the reference axis in the sliding coordinate system, min It is the projection of the position of the resistance wire terminal of the connecting wire under the reference axis in the sliding coordinate system.

[0128] 8) To determine the correctness of physical circuit connections, a directed graph of the circuit connections is first constructed based on the components and wiring results identified by the detection model. The detected component terminals (nodes) and wire connections (edges) are converted into an adjacency matrix. Next, the correct connection results predefined in the configuration file are converted into a directed graph of the same structure. A depth-first search (DFS) is performed starting from the power supply to compare the connectivity of the actual circuit with the standard diagram to determine the correctness of the circuit connection and common errors such as short-circuit paths (incorrect edges), open nodes (missing edges), and polarity errors (inconsistent edge directions).

[0129] 4. Real-time recording of experimental videos from multiple perspectives:

[0130] 1) In this embodiment, the multi-view real-time recording module for experimental operation videos uses a video streaming communication protocol (such as RTSP) to store experimental operation images under multi-view cameras. The video stream adopts segmented storage and timestamp management to ensure the complete recording of experimental operations while reducing storage overhead. The background management module supports the retrieval, playback and multi-view behavior association of video data, facilitating experimental literacy assessment and traceability analysis, and providing data support for the subsequent generation of graphic and historical records.

[0131] 5. Paperless forms and circuit design:

[0132] 1) The paperless spreadsheet and circuit design interfaces are designed with full assessment requirements and convenient operation as the fundamental principles. Paperless spreadsheet design includes flexible configuration of headers, data columns, and content, while paperless circuit design adheres to the principles of optimized human-computer interaction, providing intuitive drag-and-drop circuit component selection, wire connection, and circuit layout capabilities. In this embodiment, the interface is implemented using the Qt framework, ensuring a user-friendly and responsive interface.

[0133] 2) The key to the table design program is to allow students to design tables freely to fully test their table design capabilities, while also maintaining ease of operation during the design process. This embodiment uses a grid-based table design method, such as Figure 5 As shown, a basic grid layout of 6 rows and 5 columns is preset. Users can click on any grid to design content. It supports setting header titles and table input content.

[0134] 3) Circuit design interface such as Figure 6 As shown in the figure, based on the reference of existing design tools, it realizes component dragging, automatic routing (A* algorithm), and topology export (JSON Schema verification). The design operation is simplified, focusing on three key steps: selecting and placing circuit components, connecting wires, and submitting the design. Students construct the circuit by dragging components, connecting wires and clicking Submit to complete the circuit design. After submission, the system will generate JSON text describing the circuit connection status. This text is consistent with the circuit description format in the configuration file. By analyzing this JSON text, the system can determine the correctness of the circuit design.

[0135] 6. Paperless data filling and image drawing:

[0136] 1) This module provides a user-friendly interface for data entry and graph drawing. Students can enter experimental data through a table and select coordinate axes, data points, and relationship curves on the drawing interface. In this embodiment, this interface is implemented using the Qt framework.

[0137] 4) The data filling interface continues the layout of the table design interface, automatically removing undefined blank cells to avoid unnecessary input. Users only need to enter data in the cells that need to be filled.

[0138] 5) The image drawing function is designed to test students' basic drawing ability, including coordinate axis drawing, data point drawing and relationship curve drawing. To this end, the image drawing interface of this embodiment follows the above steps, such as Figure 4 Students need to set the axis parameters, including the axis name, unit, value range, and axis placement. The system will plot the data points on the drawing interface based on the data entered in the table. Students can then draw the corresponding relationship curve based on the data points and submit the results.

[0139] 7. Judgment of the correctness of tables, images, and circuit designs:

[0140] 1) The correctness of the table design is verified by matching it with the correct table template. The system checks whether the format, layout, and content of the table meet the experimental requirements, ensuring that its design conforms to the specified structure and data logic.

[0141] 2) Data entered in the spreadsheet will be rigorously verified according to experimental requirements, primarily for data rationality and unit consistency. The system calculates the appropriate voltage-resistance-current range based on Ohm's law, the voltage, current, and resistance data in the spreadsheet, and the instrument parameters. The system also checks that the data units and format conform to physical experiment standards.

[0142] 3) The correctness of the image depends not only on the accurate drawing of the data points, but also on the setting of the coordinate axes and the fitting of the relationship curve. The system first checks the correctness of the coordinate axes to ensure that the names, units, and numerical ranges of the coordinate axes are consistent with the experimental requirements. The correct drawing of the data points is checked, and the drawn points are compared with the data in the table. The reasonableness of the data point positions is determined based on the preset error range. In judging the correctness of the curve, the system generates a fitted theoretical relationship curve based on the data points using the least squares method, and compares it with the actual drawn curve to evaluate the difference.

[0143] 4) The correctness of the circuit design is determined by abstracting it into JSON format. The system represents the circuit components and their connections as structured data and constructs a directed graph of the circuit based on this data. The system then matches the constructed directed graph with the correct circuit diagram in the experimental requirements to verify the integrity and correctness of the circuit design.

[0144] 8. Experimental scoring and score attribution:

[0145] 1) Use the expert weighting method to quantitatively score the key steps, and generate a comprehensive scoring report based on the student operation log. The attribution method is based on right and wrong analysis and behavior tracing to identify the weak links in the knowledge mastery in the experiment. Its visual effect is as follows Figure 7 shown.

[0146] 2) Scoring consists of seven primary indicators: hypothesis and experimental design, circuit connection, exploring the relationship between current and voltage, exploring the relationship between current and resistance, instrument inspection and organization, summary and induction, and discussion. These primary indicators are further refined into 14 secondary indicators. Experts evaluate and design the score for each indicator, and students are scored item by item based on their responses, practical operation, paperless experiment design, paperless data entry, and image creation.

[0147] 3) Experimental scoring is based on student responses, practical manipulation, paperless design, data entry, and image creation. This facilitates experimental attribution. Experimental graphic and text historical records serve as attribution material, and a document is created in the form of a webpage. Links are created between the score sheet and the graphic and text historical records. Clicking on the details redirects you to the system-recorded graphic and text evidence.

[0148] 9. Experimental graphic and text history record generation:

[0149] 1) Use a multimodal model to summarize the experimental steps in the experimental operation video, supplemented by dialogue text, paperless design and record keeping, and physical operation completion records, to integrate them into a graphic and textual record of the experiment. This record is detailed down to the specific experimental steps, using operation video screenshots, charts, and scoring criteria to form a complete experimental archive for teachers to review and students to review.

[0150] 10. Experimental literacy analysis and summary:

[0151] 1) The student literacy involved in the evaluation process includes four levels: scientific concepts, scientific thinking, scientific inquiry, and experimental attitude and responsibility. The analysis and summary examples are as follows: Figure 8 shown.

[0152] 2) This embodiment uses a large language model to implement the summary and analysis function. By designing appropriate prompt templates, the large model can quickly generate the required summary analysis text, while further controlling the text output format. To save resources, this function and the large language model used for question generation use a unified model (DeepSeek-R1-Distill-Qwen-14B). After testing and verification, it can generate high-quality summary and analysis text when provided with complete graphic and text history records and reasonable prompt templates.

Claims

1. An intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception, characterized by: It includes adaptive inquiry guidance module, operation behavior analysis module, intelligent experiment platform module and evidence-based evaluation module, among which, Adaptive Exploration Guidance Module: Builds an adaptive dialogue engine based on a large language model to enable personalized configuration and dynamic deduction of experimental processes, and uses cognitive diagnostic algorithms to assess students' knowledge mastery in real time. Specifically, it includes: (1) Overall experimental process control; (2) Dialogue generation; Operation behavior analysis module: This module uses a large visual model architecture to capture experimental operation motions, combines it with an operation standardization recognition algorithm to perform real-time video analysis, and simultaneously generates a multi-dimensional experimental evidence chain with a timestamp. Specifically, it includes: (3) Experimental manipulation behavior analysis; (4) Real-time recording of experimental videos from multiple perspectives; Intelligent Experimental Platform Module: This module implements digital design of circuit topology and data curve plotting through a visual programming platform, and automatically evaluates experimental parameters based on a feature matching algorithm. Specifically, it includes: (5) Paperless forms and circuit design; (6) Paperless data filling and image drawing; (7) Judgment of the correctness of tables, images, and circuit designs; Evidence-based evaluation module: Based on a multimodal large model, data fusion analysis is performed to generate historical records including graphics and text, supporting quantitative scoring of experimental results, multi-dimensional attribution analysis of erroneous operations, and scientific literacy development assessment; specifically, it includes: (8) Experimental scoring and score attribution; (9) Generation of experimental graphic and text historical records; (10) Experimental literacy analysis and summary; The operation process of the system specifically includes the following steps: S1, login experiment; S2, process control loads the customized overall experimental process; S3. Process control combines the large language model, the overall process of the loaded experiment, and the result feedback during the experiment to control the dialogue generation; S4. If the generated dialogue is a question dialogue, the student answers the generated question and gives the answer. The system records the student's answer and judges its correctness. The answer result is used as feedback to return to the process control dialogue generation stage; S5. If the dialogue is a feedback dialogue, the student obtains the feedback text of the student's submitted results generated by the large language model. The student clicks Next and returns to the process control dialogue generation stage; S6. If the dialogue is a guided dialogue without an experiment end prompt, the student performs the relevant experimental design or physical operation according to the guided dialogue content and submits the operation results. The system records the results and evaluates them, and uses the evaluation results as feedback to return to the process control dialogue generation stage; S7. If the dialogue is a guided dialogue with a prompt indicating the end of the experiment, the system automatically generates a document with the experiment score, graphic and text history records, and experimental literacy analysis and summary. The student obtains the document and the experiment ends.

2. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The overall process control of the experiment in step (1) specifically includes: (11) Based on predefined experimental step types, users are allowed to define the overall experimental process; (12) Based on the experimental process and feedback, control the type of dialogue generated.

3. According to the intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1, it is characterized in that: The step (2) of dialogue generation specifically includes: (21) Under the control of (12), three types of dialogue content, question dialogue, guided dialogue and feedback dialogue, are generated based on the large language model, involving the fine-tuning, deployment and calling of the large model; (22) Question dialogue refers to the dialogue content that matches the experimental theme and combines relevant experimental knowledge points to ask students questions; (23) Guiding dialogue is a prompting dialogue that combines the experimental process and prompts students to design the experiment, fill in the experimental data, draw the experimental images, and perform the actual experimental operation steps; (24) Feedback dialogue is the feedback dialogue content generated by the large language model to receive students' answers to questions, experimental design, experimental data filling, experimental image drawing and experimental physical operation results and evaluate the results.

4. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (3) experimental operation behavior analysis specifically includes: (31) Based on the visual large model, key behaviors and equipment states such as circuit connection, desktop tidying, zero adjustment, sliding rheostat status and switch status during the actual operation of the experiment are detected, and the detection results are used as the actual operation results received by the feedback dialogue (24); (32) Determine the correctness of circuit connection, identify the connection status of the actual circuit, and judge whether the connection meets the experimental requirements; (33) Determine whether the desktop is ready. After the students have completed the desktop tidying steps, determine whether the experimental equipment has been arranged and placed in the designated area. (34) Zeroing behavior detection: detecting whether the ammeter and voltmeter perform zeroing operations during the students’ circuit connection process; (35) Sliding state detection, detecting the position of the slider of the rheostat when students connect the circuit, especially paying attention to whether the slider is at the position of maximum resistance; (36) Switch status detection: detect whether the switch is in the open state or the closed state during the student's circuit connection process.

5. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (4) of recording the experimental video in real time from multiple perspectives specifically includes: (41) When students are performing physical operations, the operation videos from multiple perspectives are recorded in real time and saved in the background to provide video evidence generated by historical records for (91).

6. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (5) of paperless form and circuit design specifically includes: (51) Table design methods and tools that support students to independently set the number of rows, columns, headers, and content of a table; (52) Methods and tools that support students in independently selecting circuit components, flexibly placing them, and connecting them using wires to construct complete circuit designs.

7. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (6) of paperless data filling and image drawing specifically includes: (61) A tool that supports students in filling in data in the table designed in (51). (62) Support students to draw a visualization method and tool for the relationship curve between voltage, current and resistance based on the guided dialogue in (23) and the data filled in (61); students can independently control the data labels, units and value ranges of the horizontal and vertical axes, draw based on the data points in (61), and draw the relationship curve based on the drawn points.

8. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (7) of judging the correctness of the table, image, and circuit design specifically includes: (71) Taking the tables, circuits, tabular data, and images submitted by students in (51)(52)(61)(62) as input, analyze the rationality of the submitted combination and provide feedback to (24); (72) Evaluation of table design involves whether the number of rows, columns, headers, and cell contents of the table meet the experimental requirements in (23) guided dialogue; (73) Circuit design evaluation involves whether the component types, component quantities, and component connections of the circuit logic diagram in (52) meet the experimental requirements in the guided dialogue in (23); (74) Evaluation of table data: Based on the voltage, current and resistance values ​​read by students through the electric meter and fixed resistor during actual operation, judge whether the filled data is within a reasonable range and meets the laws of electricity; (75) Curve image evaluation: evaluate the coordinate axis units, labels, value ranges, the offset of the plotted points relative to the data positions in (61) in the curve image drawn by the student in (62), and the rationality of the deviation between the student's relationship curve and the correct fitted curve.

9. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (8) of experimental scoring and score attribution specifically includes: (81) Based on the intelligent analysis of experimental operation videos, feedback from paperless experimental design and data recording, and the answers to questions generated by students (22), the key steps of the experimental process are scored, and the experimental graphic history records generated by (91) are combined to link the score status and score evidence records; (82) Experimental scoring is based on the core competencies of experimentation. Experts formulate evaluation indicators and score multiple steps, including conjecture and design of experiments, circuit connection, exploration of the relationship between current and voltage, exploration of the relationship between current and resistance, instrument inspection and summary, and discussion. (83) The score attribution links the score points of each link in (82) and the experimental graphic and text history records in (91), providing graphic and text basis for the score.

10. The intelligent guidance system for physical electrical exploration experiments based on cognitive dynamic perception according to claim 1 is characterized in that: The step (9) of generating the experimental graphic and text historical records specifically includes: (91) Based on the interactive records of dialogue, video, tables, circuit diagrams, and curve images of the entire experimental process, a multimodal large model is used to extract and generate experimental graphic and text historical records for score attribution, record query, and summary generation; The step (10) of experimental quality analysis and summary specifically includes: (101) Using the historical records of (91) and combining them with the core competencies of physical electrical inquiry experiments, an analysis and summary text is generated that includes evaluations of each step of the experiment, evaluations of experimental competencies, an overall summary of the experiment, and suggestions for student operations.

Images