NetWork
Novel capacitance detection analog interface asic based on time-multiplexed fully differential technology
FANG Shuo;HOU Changbo;WANG Lijing;LIU Yuntao;SHAO Lei;[Objective] Capacitance detection has emerged as a highly accurate technique for micro-electromechanical system(MEMS)sensors, owing to its compatibility with integrated circuits(ICs), low drift current, minimal temperature dependence, and low power consumption. The resolution of capacitance detection interface ICs based on the CMOS process is mainly limited by the DC offset and 1/f noise of the charge amplifier. Although the fully differential structure of the switched capacitor auto-zeroing and correlated double sampling(CDS) improve the resolution, the capacitance detection interface ICs still suffer from nonlinear problems caused by common mode errors, mismatches, and complex operations. The nonlinear problems hinder further improvement of the capacitance detection resolution. This paper presents a novel time-multiplexed fully differential(TMFD) capacitance detection interface application-specific integrated circuit(ASIC) with high linearity developed to improve the resolution, sensitivity, and dynamic range of MEMS sensors.[Methods] The proposed ASIC adopted a CDS architecture to enhance the CMRR. Key components included a C-V converter,programmable gain amplifier(PGA), low-pass filter, clock generator, reference voltage circuit, temperature sensor, and one-time programmable memory. Unlike conventional full-bridge sensing structures, the interface employed a novel TMFD C-V converter based on a half-bridge measurement configuration and two single-ended charge amplifiers, enabling TMFD detection. To address nonlinearity from unbalanced parasitic capacitance, the converter performed precise capacitance compensation in each clock cycle. In addition, it reduced the gain at the virtual ground node, thereby suppressing common-mode voltage effects during charge transfer. The PGA used a fully differential switched capacitor design to suppress nonlinear fluctuations and enhance stability. A multiple feedback low-pass filter was applied to eliminate high-frequency noise and reduce residual nonlinear distortion. A temperature sensor enabled temperature calibration,mitigating nonlinearity induced by temperature fluctuations. [Results] The novel capacitance detection analog interface ASIC designed has a strong nonlinear suppression ability and significantly improves the detection range and accuracy of MEMS sensors. While connected to the MEMS accelerometer, the experimental results of the novel TMFD capacitance detection interface ASIC showed the following merits:1) The detection sensitivity of the interface ASIC reached 1.342 V/g, and the detection range achieved ±3 g. 2) The noise performance and bias stability significantly improved, and the noise floor decreased to 10.5 μg/Hz1/2. 3) The range of the maximal output variation within 4 h was only 0.3 mV. 4) The temperature sensitivity ratio was reduced to 0.007‰/°C. 5) The readout ASIC also operated at a low drive voltage(5 V), consumed low power(7.5 mW), and enlarged the bandwidth(1.5 kHz). The interface ASIC structure presented here is significantly superior to others that have been developed. [Conclusions] By integrating specific half-bridge TMFD detections and precise capacitance compensation, the novel capacitance detection analog interface ASIC has a strong nonlinear suppression ability, and the detection range and accuracy of MEMS sensors are significantly improved. The TMFD capacitance detection technology proposed in this paper further improves the accuracy of capacitance detection and significantly promotes the development and application of high-precision MEMS sensor systems.
An experiment on EEG emotion recognition based on SGC-Transformer network
SHI Kaibo;YANG Yong;TANG Lin;[Objective] Electroencephalography(EEG) emotion recognition holds wide application potential in mental health diagnosis,human-computer interaction, brain-computer interfaces, and personalized user experiences. However, the nonlinear characteristics, low signal-to-noise ratio, and non-stationarity of EEG signals challenge traditional methods in extracting stable emotional features. To advance science-education integration, we designed an innovative teaching experiment centered on EEG emotion recognition using an SGCTransformer network(SGCTNet). This architecture integrates graph neural networks and Transformers, leveraging graph convolutional networks'(GCN) strength in processing non-Euclidean spatial data and Transformers' capacity for capturing global dependencies.Additionally, to mitigate deep learning's reliance on large-scale datasets, we propose a data integration strategy enhancing inter-channel relationship modeling and generalization capability. [Methods] The proposed SGCTNet is a hybrid deep learning architecture fusing Simplified Graph Convolution(SGC) and Transformer modules for efficient EEG emotion recognition. First, the SGC module extracts topological spatial features between EEG channels by simplifying the GCN structure: removing intermediate nonlinear activation layers reduces model complexity while preserving rich spatial information. Second, the Transformer module employs a self-attention mechanism to comprehensively capture global long-range dependencies among channel nodes based on these topological features, strengthening channel information utilization efficiency. Furthermore, a data integration strategy improves generalization by incorporating EEG data from multiple historical experimental sessions into current training, maximizing existing data utility. Experiments utilized public datasets SEED and SEED-Ⅳ, employing control groups to systematically evaluate SGCTNet's performance across scenarios, validating model effectiveness and data strategy generalizability. [Results] Experimental results demonstrate significant performance improvements with SGCTNet. On SEED-Ⅳ, the model achieved accuracies of 82.45%, 85.23%, and 87.62% across three sessions. On SEED, it attained 94.94%, 94.21%, and 96.87% accuracy, outperforming traditional CNNs, SVMs, Random Forests, and other deep learning models. Further analysis confirmed the data integration strategy substantially enhanced generalization: accuracy increased by 5.22%(Session 2) and 7.83%(Session 3) on SEEDⅣ, and by 3.58%(Session 2) and 3.72%(Session 3) on SEED. [Conclusions] SGCTNet integrates graph structure modeling and self-attention mechanisms, demonstrating strong modeling capability and excellent generalization in EEG emotion recognition. The developed experimental system possesses significant pedagogical value, facilitating the practical application of deep learning in EEG signal processing and supporting talent cultivation in this field.
Comprehensive experimental design of solid-state fermentation on Pu-erh primary tea by Eurotium cristatum
SHAO Jufang;LENG Yunwei;PENG Yaoli;ZHU Hongwei;[Objective] An undergraduate teaching experimental project was designed based on an undergraduate innovation training program. The project aims to further improve the teaching system, enrich teaching content, develop students' independent thinking and problem-solving abilities, and cultivate high-quality top-notch talents. [Methods] Raw Pu-erh tea(Ligustrum robustum) was fermented via solid-state fermentation using Eurotium cristatum. The growth of Eurotium cristatum during fermentation was observed and examined microscopically using an optical microscope. Seven physicochemical indicators—water extract, caffeine, tea polyphenols, amino acids,total sugar, tea pigments(theaflavins, thearubigins, theabrownins), and hydrated pectin—were compared between raw and fermented Pu-erh tea. Volatile aromatic compounds in both tea samples were identified using gas chromatography-mass spectrometry(GC-MS), and their sensory quality was evaluated. [Results] Significant changes occurred in the seven physicochemical indicators after fermentation. Tea polyphenol and thearubigin content significantly decreased, while water extract, caffeine, amino acids, total sugar, hydrated pectin,theaflavin, and theabrownin content increased. Differences in the types and quantities of aromatic substances were observed: 34compounds were identified in raw tea versus 44 in fermented tea, primarily alcohols, ketones, esters, and aldehydes. The fermented tea exhibited superior sensory quality. Key improvements included an oily appearance with visible golden particles(enhanced ornamental value), a deeper yellowish-brown liquor, and a complex aroma featuring mushroom and fermented-aged notes. The taste was smooth,mellow, significantly less astringent, and had a long aftertaste. The fermented tea leaves were darker(yellowish-brown), smoother, and more elastic. Its sensory evaluation score was higher than that of raw tea. [Conclusions] Fermentation with Eurotium cristatum significantly improved the quality of raw Pu-erh tea. This project effectively addressed challenges such as students' limited experimental skills and access to advanced instrumentation by integrating multidisciplinary techniques with valuable equipment. Students mastered fundamental skills—microbial cultivation/observation, tea fermentation, physicochemical analysis, and GC-MS detection—while deepening their understanding of the relationship between physicochemical changes during fermentation and sensory quality. Their research literacy and innovation capabilities were effectively cultivated. With its interdisciplinary nature and ease of implementation, this project is an ideal comprehensive undergraduate experiment for biology, food science, and tea science programs, meeting the demands for innovative talent cultivation in the modern biology industry.
Mobile empowerment micro-experiment:Technical characteristics,mapping mechanism,teaching mode,and application practice
GUO Feng;FU Song;HE Jun;GAI Longtao;CHEN Peisheng;[Objective] Educational digitalization is a significant national strategy, and developing novel digital applications to empower teaching represents a crucial pathway for higher education reform. As experimental teaching constitutes a vital component of higher education, empowering it through digital technology is a key reform direction. To broaden the scope of digital empowerment and reduce end-user teaching costs, mobile micro-experiments have emerged as an important carrier for digitally empowered experimental teaching.However, research on the technical characteristics, mapping mechanisms, teaching models, and practical applications of mobile-empowered micro-experiments remains insufficient. To address this gap, the research team defined and investigated mobile micro-experiments. [Methods] Through literature review and technical capability analysis, the technical characteristics of mobile micro-experiments were examined. By comparative analysis with physical experiments, their mapping mechanisms were explored.Theoretical analysis was employed to study their teaching models, while empirical analysis evaluated their practical application. [Results]Technical Characteristics: Micro-intelligence(edge/cloud intelligence) supports teacher-guided experimental decisions; Micro-push uses a lightweight engine for personalized knowledge/operation delivery; Micro-operation leverages feasible space, reasonable boundaries, and efficient interaction to trigger consistent learning experiences; Micro-feedback offers targeted, detail-oriented feedback(individual unconscious/collective conscious); Micro-service enables modular content design, supporting adaptable learning paths. Mapping Mechanism(Digitization-Reality Dual Drive): Device constraints establish finite space simulation mapping(virtual?physical), enhancing teaching ubiquity; Online-offline integration enables hybrid interactive collaborative mapping, boosting teaching efficiency;Randomness-determinacy integration creates composite value-added mapping(collaborative mobile?physical devices), increasing teaching value. Teaching Mode: Perception teaching focuses on data analysis(teacher conscious active/student unconscious passive); Agile teaching emphasizes student response, encouraging unconventional thinking; Feedback teaching stresses case analysis and outcome evaluation via models; Extended teaching promotes practice expansion across time, space, and content. Application Practice:Design standards cover principle authenticity, volume adaptability, content clues, experience consistency, interactive usability, and process gameplay. Evaluation encompasses 36 indicators across experimental design, process, and results. [Conclusions] Mobile micro-experiments utilize mobile terminals as tools, mobile applications as carriers, and visual/interactive elements as objects. This approach demonstrably enhances students' practical and innovative abilities. Implementation analysis involving 72 junior undergraduates at H School confirmed that mobile micro-experiments significantly improve experimental efficiency and teaching quality. Students universally express high expectations for mobile-empowered micro-experiments.
Design of coal floatability evaluation experiment based on mathematical statistics
SUN Meijie;ZHONG Jiali;LI Jiangcheng;LYU Ziqi;TU Yanan;ZHOU Lingmei;XIE Weiwei;[Objective] The integration of theoretical knowledge and practical skills in mineral processing engineering education is crucial for cultivating students' scientific research capabilities and innovative thinking. It is necessary to optimize slime flotation conditions to improve coal recovery rates and reduce resource waste in mineral processing industries. To investigate the optimization of slime flotation through systematic experimental design and data analysis, this study focuses on the practical teaching modules of Design and Research Methods for Mineral Processing Experiments and Mineral Processing(2): Interface Sorting Technology. The key problem in selecting flotation process conditions under the influence of complex factors is addressed, as this directly affects the efficiency and quality of the flotation experiment exploration process. By incorporating orthogonal experimental design and statistical analysis into the curriculum, this research not only enhances students' hands-on skills but also bridges the gap between theoretical concepts and real-world applications.[Methods] A four-factor, three-level orthogonal experimental design was employed to evaluate the effects of collector dosage(Factor A),frother dosage(Factor B), pulp concentration(Factor C), and aeration rate(Factor D) on the coal slime flotation performance. The experiment was conducted using a single flotation cell, using coal samples from the Outer Mongolia Data Mine. Key performance indicators, including clean coal yield, concentrate ash content, and the flotation efficiency index, were measured using standardized procedures. Data analysis involved intuitive analysis and variance analysis. The experimental protocol emphasized student engagement in all stages, including design, operation, and analysis, to foster a comprehensive understanding of flotation dynamics and statistical methodologies. [Results] When employing intuitive analysis with the flotation efficiency index as the evaluation metric, the relative significance of individual factors was not considered. By comparing the k values(average flotation efficiency index of each factor level),the optimal levels for each factor were obtained: level 2 of Factor A, level 3 of Factor B, level 3 of Factor C, and level 1 of Factor D.Without considering significance, by comparing the range of k values(the difference between the maximum and minimum k values) and analyzing the primary and secondary effects of various factors, the frother dosage(Factor B) was found to have the most significant impact on the flotation perfection index among the four factors, followed by the aeration rate(Factor D). The pulp concentration(Factor C) had a more significant impact, and the collector dosage(Factor A) had the lowest impact. The optimal experimental plan was determined as B3D1C3A2. Subsequent variance analysis revealed that factors A, B, C, and D all exerted statistically significant effects on the flotation efficiency index with minimal experimental error. The relative influence of each factor on the flotation efficiency index was quantified by comparing their variance magnitudes: Factor B > Factor D > Factor C > Factor A, consistent with the trend identified through range analysis. This alignment between variance-based rankings and range analysis outcomes further validated the selection of B3D1C3A2 as the optimal condition combination. [Conclusions] This study successfully integrated orthogonal experimental design into mineral processing education, thereby achieving dual objectives: identifying optimal coal slime flotation parameters and cultivating students' scientific acumen. Orthogonal experimental design is a statistical methodology effectively applicable to flotation experiment design, and it leverages its inherent advantages of “balanced distribution” and “uniform comparability” to achieve the identification of optimal solutions while comprehensively assessing the relative significance and statistical impact of individual factors on experimental outcomes. This approach not only reduces the number of required trials and associated workloads, which can thereby shorten experimental timelines, but it also minimizes caused by random variations in conventional methods through the rigorous statistical analysis of holistic datasets. In course instruction and undergraduate innovation training programs, students should conduct extensive literature reviews and participate in regular group discussions to exchange ideas. This framework would cultivate multifaceted competencies, including literature retrieval abilities,self-directed learning, integration of theoretical knowledge with practical operations, and hands-on problem-solving. Furthermore, the application of statistical methods for processing and interpreting experimental results exemplifies the pedagogical benefits of interdisciplinary integration.
Experimental design and practice of sunshades in road tunnels to mitigate the “black hole effect”
WANG Qingzhou;YANG Lingling;WANG Mingqian;MA Shibin;[Objective]The "black hole effect" at tunnel entrances, caused by abrupt luminance transitions between exterior and interior environments, often induces momentary driver blindness and significantly increases traffic accident risk. While sunshade structures can mitigate this sudden luminance drop, current design practices lack systematic, quantitative guidance on key parameters such as length and transmittance. This deficiency hinders the establishment of smooth luminance transitions and full resolution of the black hole effect.Moreover, optimizing sunshade parameters typically relies on costly, high-risk real-vehicle eye-tracking experiments unsuitable for educational purposes. To address these challenges, this study developed a safe, economical, and highly operable eye-tracking platform based on a scaled tunnel model. This platform systematically evaluates the influence of sunshade length and transmittance gradients on drivers' visual adaptation, providing theoretical support for design optimization and serving as a practical tool for traffic engineering education. [Methods]A 1:50 scale tunnel visual behavior eye-tracking platform was constructed using illuminance data from full-scale real-vehicle tunnel experiments. The tunnel structure, built with foam panels, employed directional lighting techniques with controllable luminaires to reproduce typical tunnel entrance luminance transitions. To overcome the limitation of fixed sunshade parameters in real tunnels, blue shading films and a segmented multilayer arrangement method simulated different length and transmittance gradient combinations. Fifteen test scenarios were designed, covering sunshade lengths from 45 m to 105 m and various transmittance gradient configurations. A high-resolution imaging device mounted on a remote-controlled model vehicle simulated the driver's perspective during tunnel passage. Subjects wore Tobii Pro Glasses 3 eye trackers to observe reconstructed driving scenes. Pupil diameter data were collected across three spatial segments—approach zone, sunshade zone, and tunnel entrance zone—to investigate the influence of sunshade design parameters on visual adaptation. [Results]Results revealed that sunshades significantly mitigated the black hole effect by reducing pupil dilation during tunnel entry. Without a sunshade, the average pupil area change rate reached 51.4%. With sunshades, this rate decreased to 20–45%. Longer sunshades(90–105 m) provided stronger buffering, reducing pupil change rates to 20–31%. Crucially, optimized transmittance gradients achieved comparable mitigation at shorter lengths. For instance, a 75 m sunshade using a high–medium–low gradient combination reduced the pupil change rate to 34%, matching the performance of a 90 m sunshade with uniform transmittance.These results confirm that well-designed transmittance gradients can substitute for additional structural length, offering an efficient and economical solution for tunnel entrance safety. [Conclusions]By establishing a scaled-model-based eye-tracking platform, this study quantified the visual buffering effects of tunnel entrance sunshades, overcoming limitations of real-vehicle experiments such as uncontrollable parameters and safety risks. The findings demonstrate that appropriate combinations of sunshade length and transmittance gradient effectively reduce pupil fluctuation and alleviate the black hole effect. Furthermore, under structural constraints, refined transmittance gradient design achieves outcomes comparable to longer sunshades, validating the “light over length” optimization strategy.The proposed platform and methodology not only enhance tunnel visual safety but also provide a replicable, scalable solution for experimental teaching in transportation engineering.
A study on automated evaluation of experimental reports based on large language models ——Taking “Applications of artificial intelligence” course as a case
WANG Meng;LIU Xiaoyan;LUO Haichi;XU Ge;[Objective] In experiment teaching, experiment reports serve as important instruments for assessing students' learning outcomes and complex problem-solving skills. However, relying on professional teachers to evaluate large volumes of experiment reports faces challenges such as low efficiency, strong subjectivity, delayed feedback, and generic feedback templates, making it difficult to meet demands for scalable and personalized evaluation. With breakthrough advancements in natural language processing(NLP), large language models(LLMs) like ChatGPT, equipped with robust natural language understanding and generation capabilities, offer technological opportunities to address inherent issues in manual evaluation. Nevertheless, the systematic evaluation of experiment report content requires further research and validation. [Methods] To address this, this study employs the course "Applications of Artificial Intelligence" offered by a university in eastern China as a case study. This course requires students to model and solve problems by applying learned AI algorithms to experimental tasks, and to write academically rigorous paper-style reports. First, a dual-dimensional evaluation index system for AI experiment reports was constructed based on literature analysis and expert consultation, focusing on linguistic expression and content quality. Then, an LLM-based automated evaluation method for experiment report texts was proposed, leveraging few-shot Chain-of-Thought prompting. This approach involves:(1) preprocessing, segmenting, and extracting secondary indicator values for the linguistic expression dimension;(2) interacting with the model via API interfaces and few-shot Chain-of-Thought prompts to generate secondary indicator scores and evaluation rationales for the content quality dimension;(3) integrating outputs to produce socially empathetic evaluation comments. Finally, three representative LLMs—OpenAI's gpt-4o-preview, Baidu's Wenxin 4.0 Turbo, and Zhipu AI's ChatGLM4—were selected to validate the method's effectiveness based on scoring consistency and feedback quality. [Results]Experimental results demonstrate that regarding scoring consistency, the three models achieved Pearson correlation coefficients(PCC) of 95.65%, 83.52%, and 74.29% with human ratings, respectively, exhibiting highly consistent scoring trends. Among them, the gpt-4o-preview model yielded the smallest mean squared error(MSE) and maintained high reliability on the stricter concordance correlation coefficient(CCC) metric. In terms of feedback quality, comments generated by all three models, as evaluated through professional teacher scoring analysis, achieved high quality levels across four dimensions: readability, relevance, accuracy, and personalization. [Conclusions] The study revealed that LLMs underperformed in extracting statistical features, necessitating integration with specialized statistical tools to optimize multi-dimensional feature extraction. Few-shot Chain-of-Thought prompting enhances the interpretability and transparency of evaluation outcomes by providing limited evaluation examples and explicit assessment steps. This guides and externalizes the model's logical reasoning process to mitigate hallucination phenomena. While open-source LLMs like Wenxin 4.0 Turbo exhibit notable performance gaps compared to closed-source counterparts, future efforts should focus on optimizing open-source LLMs and developing specialized educational LLMs to improve evaluation accuracy and feedback depth. By exploring the feasibility of LLMs in automated experimental report evaluation, this study not only provides a novel technical approach for automated text assessment but also offers empirical evidence for applying LLMs in educational evaluation, thereby advancing the intelligent and personalized development of experiment teaching.
An experimental teaching platform for robotic collaborative sheet forming-additive manufacturing
LI Yanle;HAN Fuzhen;LUAN Xiaona;CHEN Heng;YANG Chenglong;LI Jianyong;LI Fangyi;[Objective] Sheet Forming-Additive Manufacturing(SF-AM) is an advanced manufacturing concept proposed to address the challenge of fabricating thin-walled structures with complex geometries. Combining robotic incremental sheet forming with laser additive manufacturing, this innovative approach holds significant potential for aerospace and other advanced manufacturing fields. To advance this concept and enhance intelligent manufacturing competencies for engineering students, we developed an experimental teaching platform featuring robotic collaborative SF-AM. The platform leverages the high degree of freedom motion capabilities of robot systems while integrating multiple auxiliary modules to enable multifunctional manufacturing operations. [Methods] The platform comprises two core modules—the incremental sheet forming module and the laser additive manufacturing module—alongside several auxiliary modules,including an energy field auxiliary adjustment module, a rolling surface strengthening module, and an online quality monitoring module.The incremental sheet forming module utilizes a KUKA KR 500 R 2380 industrial robot and supporting equipment. The laser additive manufacturing module employs a KUKA KR 50 R 2100 industrial robot equipped with a laser cladding head and supporting equipment.Both industrial robots integrate the RobotTeam cooperative control software package, increasing the SF-AM system's motion degrees of freedom and creating a dual-robot cooperative control system for SF-AM. These core modules can operate individually for incremental sheet forming of complex curved components or laser additive manufacturing, or collaboratively for SF-AM short-process composite manufacturing, enabling intelligent manufacturing and remanufacturing of sheet parts. The platform's SF-AM capability was validated through test production of a sample. Engineering students can use the platform to conduct SF-AM experiments, learn advanced manufacturing concepts, and master intelligent industrial robot control methods based on dual-robot collaboration. [Results] Following SF-AM sample test production and macro/microscopic testing verification, the sample demonstrated sufficient macroscopic precision for industrial plate requirements. Microscopic examination confirmed good-quality metallurgical bonding at the interface between the fusion cladding layer and the substrate, verifying the platform meets SF-AM requirements. Furthermore, the experimental teaching program integrating advanced manufacturing concepts with practice not only fulfills requirements for cultivating students' innovation and comprehensive abilities within new engineering education but also deepens their understanding of advanced manufacturing concepts.[Conclusions] This experimental teaching platform overcomes the limitation of simple incremental sheet forming in manufacturing complex-feature thin-walled structures. It promotes the application and development of SF-AM as an advanced manufacturing technology while fostering practical abilities and innovative thinking for students specializing in intelligent manufacturing within new engineering courses, thereby laying a solid foundation for talent development.
Teaching experiment design of mining equipment based on vibration characteristics of ball mill
LUO Xiaoyan;HE Saisai;XU Huazhi;YANG Lirong;[Objective] To facilitate students' rapid and in-depth understanding of ball mill operational characteristics and precisely address the educational need for integrating scientific research within the context of intelligent mining equipment development, this National Natural Science Foundation of China-supported study innovatively designs a teaching experiment for investigating ball mill vibration characteristics. [Methods] This experimental system establishes an integrated practical framework spanning hardware to algorithms,encompassing the entire ball mill operation process. First, a ball mill signal acquisition device is constructed. Sensors are precisely selected and optimally positioned based on the mill's structure and operating environment to ensure stable collection of multi-dimensional signals(e.g., vibration, noise). During feature extraction, time-domain(e.g., mean, variance, kurtosis) and frequency-domain(utilizing Fourier transform and wavelet analysis to extract frequency features) analysis methods are comprehensively employed to extract equipment status information from the signals. Concurrently, deep learning algorithm application is integrated: the publicly available ResNet-34 architecture is selected and adapted based on the ball mill's operational mechanism and experimental requirements to construct a load identification model. The experimental design rigorously accounts for key operational parameters, scientifically planning 16comparative experiments based on two core variables: filling rate(covering typical low, medium, and high intervals to simulate varying grinding medium fill levels) and material-to-ball ratio(setting multiple ratios to reflect actual production material variations). Using professional signal analysis software and deep learning platforms, collected signals are preprocessed, features are extracted, and models are trained. [Results] Test set verification demonstrates a ball mill load identification accuracy of 89.63%, meeting the precision and reliability requirements for teaching experiments. [Conclusions] From a teaching perspective, this experiment overcomes the limitations of purely theoretical instruction. It bridges abstract ball mill working principles(e.g., cylinder rotation, steel ball impact, material grinding interactions) and practical vibration characteristic analysis(encompassing signal collection to intelligent recognition), guiding students to master cutting-edge technologies like signal acquisition(sensor selection, installation, DAQ integration), signal processing(filtering,denoising, time-frequency feature extraction), and deep learning algorithms(model construction, training, optimization, application). This enhances students' engineering practice and innovation capabilities within intelligent mining equipment contexts. From an educational reform perspective, leveraging a scientific research project(NSFC) fosters deep integration of research and experimental teaching. It provides a “research feeding back into teaching” paradigm for mining engineering, mechatronics, and related disciplines, addressing the demand for interdisciplinary talent versed in both mineral processing and intelligent technologies. This supports the adaptation of university mining curricula to industry digitalization and intelligent transformation trends, cultivating high-quality graduates with theoretical depth and practical innovation capabilities for green and smart mines.
A Study on the Application of Large Language Models in Language Experiment Teaching
ZHANG Luni;GU Rentao;WANG Qianya;[Objective] With the rapid development of artificial intelligence， especially the significant improvement of the capabilities of Large Language Model (LLM)， it has shown great application potential in the field of education. LLM is a high-computing knowledge processing tool with initial ideas and wisdom. In the field of foreign language education and research， LLM provides language learners with more natural， friendly， customized and situational language learning support through its powerful natural language processing capabilities. In the field of Foreign Language Teaching， some pioneering studies have emerged to explore the application of LLM， but there is a lack of systematic and in-depth research throughout the teaching process. It is urgent to carry out relevant research to explore a closer and more "coordinated" integration model between LLM and Foreign Language Teaching， so as to provide a reference for teaching research and practice. [Methods] In order to reveal the essence and specific reasons for the slowdown in the development of Language Experiment Teaching， this paper uses the service process modeling method-service blueprint method to conduct a detailed hierarchical analysis of Language Experiment Teaching stages， user behaviors， service touchpoints， front-end operations， back-end operations， and support systems， revealing six teaching pain points and bottlenecks distributed in the six stages of Language Experiment Teaching. On the other hand， in order to systematically and deeply analyze the relationship between the underlying LLM technical characteristics and the surface LLM teaching application potential， and to form more flexible and in-depth insights into the application of technology in teaching， this paper takes affordance as the analytical perspective， constructs LLM teaching affordance and limitation analysis based on the technical characteristics of LLM， and constructs three levels of LLM affordances: operation， action and activity. Based on the hierarchical affordance structure of LLM and six major language experiment teaching problems， a series of LLM empowerment development strategies for Language Experiment Teaching are constructed. Finally， limited by the inherent limitations of LLM， such as the model illusion， lack of real understanding and reflection ability， opaque generation mechanism， knowledge limitations and lag， no real memory function， token restrictions， and harmful output， the teaching application of LLM often leads to a series of associated problems: reducing the effectiveness of experimental evaluation， affecting the effectiveness of ability training， and weakening the role of teachers. In order to make up for the lack of systematic and scientific research on the theme of LLM teaching associated problems in academia， the classic problem-solving model in Triz innovation theory， the Substance-Field Model， is used as a research method to deeply explore the associated problem-solving strategy of LLM Foreign Language Experiment Teaching application. [Results] A service blueprint for Language Experiment Teaching was constructed， and the pain points and bottlenecks that hindered the development of the series were sorted out; an LLM hierarchical affordance framework was constructed; six experiment teaching reform strategies based on the affordance framework were constructed: LLM empowers the construction of teaching resource libraries， LLM empowers the construction of personalized learning paths， LLM empowers the creation of language practice opportunities， LLM empowers teachers to reduce their burden and increase efficiency， LLM empowers Language Experiment evaluation， and LLM empowers classroom feedback; at three levels， LLM foreign Language Experiment Teaching application accompanying problem solution strategies were constructed: experiment evaluation effectiveness， ability training effectiveness， and teacher role positioning. [Conclusions] Following the path of "teaching problem disclosure-LLM affordance analysis-teaching strategy formulation-accompanying problem solution"， this paper constructed an LLM affordance framework that links technical characteristics， implementation methods and corresponding effects based on the hierarchical affordance theory， and comprehensively used the service blueprint method and the Substance-Field Model theory to explore the application strategies and accompanying problem solutions of LLM in Language Experiment Teaching， which provided a meaningful reference for the in-depth integration of LLM and Language Experiment Teaching， and provided a reference method for studying the teaching integration design of LLM.