NetWork
Prediction and safety evaluation of tunnel deformation based on monitoring data and RBF neural network
LI Jianhong;JING Wei;LIU Gongning;[Objective] The deformation law and its magnitude during tunnel construction are always an important criteria for the stability of surrounding rock of tunnel. Due to the many factors such as monitoring cost， construction period and complex site conditions， it is impossible to monitor every section on site. Therefore， it is necessary to propose a set of targeted tunnel deformation prediction methods， and to fit and modify the prediction model through on-site statistical data， so as to provide a reliable means for the safety of tunnel construction， which is the main focus and difficulty of current research. [Methods] To solve above mentioned problem， a tunnel deformation prediction and safety evaluation method based on monitoring data and RBF neural network is proposed in this paper. Firstly， taking a long tunnel in Southwest China as an example， the deformation of the tunnel is measured and analyzed on site. Then， based on the RBF neural network model， the measured data of tunnel vault settlement， peripheral displacement and internal deformation of surrounding rock are fitted and predicted， and compared with the prediction results of traditional BP neural network， the tunnel safety evaluation is carried out according to the predicted results， and the corresponding construction suggestions are put forward. [Results] The analysis results show that: 1) the determination coefficients R2 for predicting tunnel deformation (including tunnel vault settlement， peripheral displacement and internal deformation of surrounding rock) based on RBF neural network proposed in this paper are 0.999， 0.998 and 0.978， respectively， which are above 0.997， with high accuracy. However， the determination coefficients R2 of the traditional BP neural network are 0.992， 0.984 and 0.886， respectively. 2) According to the obtained data， it can be clearly explained that the prediction generalization ability of RBF neural network is better than that of traditional BP neural network (especially the internal displacement of surrounding rock)， and it can achieve a good agreement with the field tunnel deformation monitoring data. 3) Combined with the RBF neural network prediction results and the three-level safety evaluation criteria of tunnel engineering， the safety evaluation of 8 sections of the relied tunnel project is carried out. The vault settlement and the average displacement variation rate around the first five sections are 0.1mm/d， which is less than the safety criterion of 0.2mm/d. The surrounding rock is basically stable and can be constructed normally. For K33+228 cross section， its average variation rate is greater than 0.2～1.0mm/d， and it should be strengthened observation. For K33+240 and K33+251 sections， the predicted average rate of vault settlement and peripheral displacement is greater than 1.0mm/d， and the lining should be strengthened. Namely， most sections of the tunnel meet the safety requirements and can ensure normal construction， but the average deformation rate of individual sections is still large， and it is necessary to strengthen support and observation in the future. With the increase of time， the internal displacement of different parts of the surrounding rock decreases， and the displacement changes tend to be stable. [Conclusions] The deformation prediction method proposed in this paper can not only provide quantitative decision-making basis for the current project， but also accumulates important technical experience for subsequent tunnel construction under similar geological conditions.
Test methods for bulk density， moisture content， porosity， specific yield of quaternary salt lake core samples in the Qaidam basin
LIU Liang;YIN Lucheng;JIA Jiantuan;XIAO Yuping;ZHAO Yuxiang;HAN Guang;ZHU Yunjun;LI Haiming;[Objective] The physical properties of salt lake core samples， such as bulk density， moisture content， porosity， and specific yield， are fundamental parameters for exploring potash deposits. These properties are crucial for evaluating brine deposits and calculating salt and brine reserves. Due to their highly heterogeneous structure and susceptibility to external forces， salt-lake core samples often deviate from their original state，cause structural deformation. In addition， their strong hygroscopicity leads to the loss of crystallized water， altering their composition. During the drilling process， brine inevitably drains from the salt core samples， complicating the measurement of brine volume discharged under gravity. This makes conventional methods unsuitable for determining porosity and specific yield. [Methods] After extracting the salt-lake core samples from the drill rods， brine is drained under gravity. Bulk density is determined using Archimedes’ principle， while moisture content is measured at 45 ℃， including the crystallized water from thenardite (sodium sulfate) and borax. After measuring bulk density， the sample is placed in a siphon test bottle filled with brine from the same borehole and depth. A vacuum pump creates negative pressure， allowing brine to infiltrate the sample’s pores for the specific yield test. The volume of brine injected equals the volume that would be discharged under gravity， and the ratio of injected brine volume to the total sample volume defines the specific yield. Porosity is calculated by adding the volume of brine discharged (specific yield) to the volume retained (water retention) and dividing by the total sample volume. Porosity is further calculated through water retention and K-value tests(The K-value represents the ratio of the volume of brine that the sample cannot discharge under the action of gravity to the volume of water adsorbed by the sample). The specific yield and porosity data are validated by field pumping tests， with good agreement. [Results] The experiment uses brine from the same location and depth as the sample， which better reflects the actual conditions of Quaternary salt lake strata. Experimental temperatures are strictly controlled to prevent the loss of crystallized water from salts other than thenardite and borax. Specific yield is determined using the siphon test bottle， and porosity is calculated more accurately by determining the K-value and the content of adsorbed water. The specific yield data shows less than 10% variation compared to unsteady flow pumping test results. [Conclusions] The proposed method is simple， straightforward， cost-effective， and fast. Using brine as the medium better matches actual formation conditions. The specific yield data aligns well with field pumping test results and can be used for reserve estimation of Quaternary salt lake liquid mineral layers. However， due to reliance on low negative pressure backfilling， this method is unsuitable for more compact salt-lake core samples from the Tertiary period. Since the test is destructive and the samples are heterogeneous， it cannot be repeated. Therefore， equipment calibration， environmental control， personnel training， and measures to preserve sample integrity during field sampling， storage， and transportation are essential. The test requires two people to cooperate in operation and two people to check the recorded data.Dual-operator procedures are also used to ensure data accuracy and reliability
Contour line reconstruction algorithm based on multifeature fusion matching and multicondition-constraint surface construction
YAO Jinpeng;JIAN Chu;ZHU Sixin;ZHOU He;DENG Jiayue;HE Mengyu;JIAN Xingxiang;[Objective] The geological surface reconstruction method based on contour line data has been widely adopted due to its efficiency and accuracy. Specifically， the contour reconstruction technique utilizing feature point matching and mesh partitioning has become a critical research focus in this field because of its excellent performance. However， existing feature point matching methods often exhibit inadequate robustness and accuracy when dealing with complex geological structures and noisy data. Additionally， current triangulation algorithms face significant challenges in generating high-quality triangular meshes， such as in avoiding elongated and narrow triangles and ensuring uniform distribution without self-intersections. This study aims to address these issues by proposing a novel contour line reconstruction algorithm that integrates multifeature fusion matching with multicondition-constraint surface construction， thereby enhancing the accuracy and reliability of geological surface reconstructions. [Methods] To overcome the limitations of multifeature constraints and the local defects of fuzzy matching algorithms， this paper proposes the mentioned algorithm. First， a similarity quantification evaluation method based on spatial quadrilaterals is introduced， defining four feature measures to enhance robustness and compensate for the shortcomings of basic triangular patches in capturing feature point adjacency information and orientation constraints. Principal component analysis is employed for feature fusion and optimal solution computation， with principal components selected based on cumulative contribution rates. The dissimilarity between matching point pairs is defined using Euclidean distance， establishing a robust matching relationship. Second， a secondary matching mechanism incorporating distance weights and inflection point detection is proposed to mitigate the impact of locally unreasonable matches. Finally， to address the inadequacies of existing triangulation algorithms during the tiling process， an adjacency surface roughness function is defined to assess the quality of adjacent triangles. Surface construction is then performed based on this quality assessment， ensuring the smoothness and detail-capture ability of the involved triangular mesh. [Results] Experimental results demonstrate that the proposed algorithm achieves reasonable outcomes in modeling geological exploration Contour Profile data and geophysical inversion profile data. By introducing multiple feature measures and optimization mechanisms， the accuracy and robustness of contour line reconstruction are significantly improved compared to conventional approaches including global optimal constraint matching and local optimal constraint matching. Notably， when handling complex geological structures and noisy data， the new algorithm exhibits higher adaptability and stability. Thereby enhancing the overall quality of the reconstructed models. [Conclusions] This study provides a robust and efficient solution for geological surface reconstruction through theoretical innovation and methodological improvements， significantly enhancing the accuracy and reliability of geological structure models. It offers substantial support for fields such as resource exploration， environmental monitoring， and disaster prevention.
Research on Simulation Teaching of Carbon Emission Flow Calculation in Power Systems under Low-Carbon Transition Context
CHEN Feixiong;FU Xiaoying;SUN Wenjing;SHAO Zhenguo;[Objective] Conventional carbon emission flow (CEF) calculation methods for power systems face challenges regarding unfair network loss allocation and difficulty in quantifying renewable energy's carbon reduction contributions. Existing studies often allocate network losses entirely to the load side or neglect the actual carbon reduction effects of renewable energy， leading to inequitable carbon emission responsibility distribution. To address these issues， this study proposes a simulation-based teaching method for CEF calculation grounded in power decomposition. [Methods] To overcome these limitations， the proposed method decomposes the actual power network into a basic power network and a power deviation network. Virtual nodes are introduced into transmission branches to distribute network losses equally between the generation side and the load side. We treat renewable energy units as negative-valued loads to quantify their carbon reduction contributions by calculating the difference in injected carbon flow rates. Finally， an IEEE 57-node system compares carbon flow rate differences under four allocation modes and simulates the impact of real power fluctuations on actual carbon flows. [Results] Experimental results demonstrate that: 1) Under the bidirectional allocation mode， branch carbon flow rates fall between those of generator-side allocation and load-side allocation， validating the rationality of shared responsibility between generation and load sides. 2) When renewable energy unit G2 is treated as a negative load， its carbon reduction contribution increases from 66.49 t/h in the basic power network to 92.74 t/h in the power deviation network. This accurately reflects the impact of renewable power variability on carbon emission reduction within the system. [Conclusions] The proposed power decomposition-based CEF calculation method provides a fair and precise carbon emission measurement tool for power system low-carbon transitions. The teaching framework successfully integrates theoretical knowledge with simulation experiments， enhancing students' understanding of CEF-related theories while cultivating their low-carbon awareness and research literacy.
Application and exploration of FLAC3D numerical simulation technology in coal mining disaster prevention and mitigation teaching
WANG Pu;CHEN Huidan;WEI Zesheng;ZHANG Jun;ZHANG Chuanyang;ZHANG Mei;[Objective] During underground mining， overburden movement， stress evolution， and energy release are the fundamental causes of mine pressure responses such as fault slipping， large deformation， rock bursts， and other hazards， which pose significant threats to mining safety. However， these underground disasters are often difficult to detect in advance， highly complex， and occur suddenly， making traditional disaster prevention and mitigation teaching methods limited in conveying disaster causation and coping mechanisms. To address issues such as poor intuitiveness， weak engineering relevance， and a disconnect between theory and practice in current teaching methods， this study aims to cultivate students’ analytical thinking and engineering decision-making skills in disaster warning and prevention. It introduces Fast Lagrangian Analysis of Continua in 3D(FLAC3D) numerical simulation technology into the teaching framework for disaster prevention and mitigation. Two typical teaching models are designed and implemented to explore the feasibility， practicality， and promotion value of FLAC3D in the teaching system. [Methods] Using FLAC3D， this study constructs a high-precision numerical model to simulate typical disaster conditions， such as reverse fault mining and gob-side roadway support. It integrates parameter regulation， dynamic evolution tracking， and result visualization modules to systematically analyze the entire process from mining disturbance to dynamic response and disaster chain evolution. This forms a “theoretical teaching → modeling practice” model—an effective instructional approach for disaster prevention and mitigation. A closed-loop teaching path is developed: “theoretical teaching → model design → parameter debugging → result interpretation → optimization decision-making → summary.” The model fully considers the influence of fault coal pillar width variation on peak stress and introduces a comparative analysis of roadway surrounding rock stability under multiple support structures， thereby enabling an approximate reconstruction of real working conditions. [Results] Results show that FLAC3D offers excellent dynamic visualization and parametric control capabilities. It clearly presents stress concentrations in mining， fault barrier effects， and rock failure modes and quantitatively analyzes the adaptability and effectiveness of various support schemes in controlling surrounding rock deformation. The deep integration of numerical simulation with teaching objectives has significantly enhanced students’ understanding of disaster mechanisms， triggering paths， and the development of prevention and control strategies. The problem-oriented simulation process has effectively stimulated students’ systematic analysis and engineering judgment. Feedback from teaching practices indicates a marked improvement in students’ awareness of disaster early warning and numerical simulation proficiency. [Conclusions] Embedding FLAC3D numerical simulation technology into the disaster prevention and mitigation teaching system effectively bridges the gap between theory and practice and significantly enhances the intuitiveness， interactivity， and practicality of instruction. This method strengthens students’ quantitative understanding of the spatiotemporal evolution of disasters and offers an important platform for personalized learning paths and complex disaster modeling training. In the future， it can be further integrated with data mining and deep learning capabilities of artificial intelligence to establish a new paradigm in teaching and safety management， combining “numerical simulation + real-time monitoring + intelligent analysis.” This approach will support the high-quality development of coal mine engineering talent and the disaster prevention and control system.
Comprehensive Experimental Design of AuPt Bimetallic Catalytic Oxidation of Ethylene Glycol to Glycolic Acid
YAN Hao;YANG Shixuan;FENG Xiang;LIU Yibin;CHEN Xiaobo;ZHAO Hui;YANG Chaohe;[Objective] With societal development and increasing industry demands， basic experimental teaching no longer meets enterprise requirements for students. Comprehensive experimental teaching reform has consequently become a focal point in higher education. Comprehensive experiments play a vital role in enhancing students' innovative abilities and scientific literacy. Supported by national initiatives promoting integrated science-education models， universities have widely implemented comprehensive experimental programs driven by scientific research outcomes. Concurrently， China's ethylene glycol industry faces structural overcapacity challenges. Addressing this， developing a green production process for glycolic acid via one-step ethylene glycol oxidation offers dual benefits: enabling high-value utilization of surplus capacity and meeting green chemistry principles of atom economy， positioning it as a key research focus. Introducing undergraduate teaching content—including literature research， preparation of AuPt bimetallic catalysts， material structure/morphology characterization， catalytic performance evaluation， and experimental data analysis—holds significant value for advancing students' innovation capabilities and scientific literacy. Building on this foundation， a comprehensive experimental design titled 'AuPt Bimetallic Catalytic Oxidation of Ethylene Glycol to Glycolic Acid' was developed， integrating frontier research topics and scientific achievements. [Methods] This comprehensive experiment involved preparing AuxPty/SBA-15 bimetallic catalysts via the sol-gel method. Different Au/Pt ratios yielded varied catalyst compositions. The catalysts' structure and properties were characterized using X-ray diffraction (XRD)， transmission electron microscopy (TEM)， physical adsorption-desorption analysis， and ultraviolet-visible (UV-Vis) absorption spectroscopy. Catalytic performance was evaluated through reaction testing. [Results] Results demonstrated significantly enhanced catalytic performance of AuxPty/SBA-15 compared to monometallic Au/SBA-15 and Pt/SBA-15. Catalytic activity exhibited a volcanic curve trend relative to Au/Pt ratio， with optimal performance observed at a 1:2 Au/Pt ratio. Characterization revealed that the superior performance stems from the formation of an AuPt alloy structure， which promotes active site dispersion and increases exposure of active centers. [Conclusions] By effectively translating cutting-edge scientific research into undergraduate experimental teaching， a comprehensive project on 'AuPt Bimetallic Catalytic Oxidation of Ethylene Glycol to Glycolic Acid' was established. This project features comprehensive objectives， innovative design， and operational feasibility， aiming to enhance students' innovative capacity and scientific literacy. It broadens students' horizons， cultivates innovative thinking and problem-solving skills， and supports the cultivation of high-quality applied talents.
Preparation and application of graphene-based sensor in design of comprehensive experiment for determination of follicle-stimulating hormone
FAN Yan;GUO Yaohua;WANG Lixia;WANG Lei;[Objective] Electrochemical sensing technology is a highly sensitive and accurate analytical method covering interdisciplinary research fields including sensing technology， electronic technology， electrochemical technology， and nanomaterials. It has drawn wide attention in the fields of medical testing， environmental detection， and food safety. An innovative comprehensive experiment was introduced into the sensing technology course， helping students understand frontier research developments， fostering interdisciplinary scientific thinking， and inspiring scientific exploration ability， thereby improving comprehensive disciplinary proficiency. [Methods] Graphene has excellent photoelectric and catalytic properties. Electrochemical sensors fabricated with graphene-based nanocomposite materials have wide applications in medical detection， environmental testing， new energy batteries， and the aerospace field. Reduced graphene oxide (rGO) is obtained through the redox reaction of graphene oxide. RGO and multiwalled carbon nanotubes (MWCNTs) are carbon-based nanomaterials with favorable electronic properties and an ideal structure for loading electroactive substances. RGO and MWCNTs were compounded to form rGO/MWCNTs nanomaterials using a solution-blending method. Gold nanoparticles (AuNPs) were immobilized on the surface of rGO/MWCNTs nanomaterials through interaction with amine groups. Electroactive substance methylene blue (MB) was attached to rGO/MWCNTs nanomaterials via π-π stacking and subsequently converted into polymethylene blue (PMB) through electrodeposition. The resulting rGO/MWCNTs/PMB/AuNPs nanocomposites were modified onto the surface of the dual-channel working electrode of an electrochemical sensor， acting as the electrochemical active layer. This layer has a large specific surface area， providing sufficient binding sites for the adsorption of follicle-stimulating hormone (FSH) antibody to form an immune sensing recognition layer for the recognition and detection of FSH antigen. Electrochemical immunoassay detection of FSH was based on the principle that the insulated immunocomplex formed by the specific binding of FSH antigen and antibody enhances steric hindrance， affecting electron transfer within the electrochemically active layer and thereby lowering the detection current of PMB. Morphology imaging and energy dispersive spectroscopy (EDS) of the rGO/MWCNTs/PMB/AuNPs nanocomposites were performed using a scanning electron microscope (SEM) JSM-7900F. Electrochemical measurements， including square wave voltammetry (SWV) and cyclic voltammetry (CV)， were adopted to measure the performance of the FSH sensor. [Results] rGO， MWCNTs， MB， and AuNPs were used to synthesize the rGO/MWCNTs/PMB/AuNPs nanocomposites via solution blending and electrodeposition. SEM images show a typical wrinkled structure of rGO and tubular structure of MWCNTs with nanoparticles adhering to their surface. EDS analysis confirmed the presence of AuNPs and PMB on the rGO/MWCNTs nanocomposites. The synthesized rGO/MWCNTs/PMB/AuNPs nanocomposites were used to modify a PET-based dual-channel screen-printed electrode， forming an electrochemical sensor for the highly sensitive determination of FSH by direct immunoassay. Under optimized conditions， SWV current responses decreased with increasing FSH antigen concentration. A good linear relationship between SWV peak responses and log concentrations was observed in the range of 1–350 mIU/mL. The limit of detection (LOD) was determined as 0.01 mIU/mL (S/N = 3). Interference assay results showed that the dual-channel immunosensor has excellent specificity for FSH. The immunoassay results demonstrate that the proposed strategy is applicable for the determination of lyophilized FSH samples with acceptable accuracy. [Conclusions] The experiment design translates faculty scientific research into a comprehensive experiment for the sensing technology course. Through systematized experiment design and operation， students gain an understanding of nanocomposite fabrication and characterization methods， master the testing principles and methods of electrochemical detection instruments， and cultivate creativity， scientific thinking， and comprehensive practical skills through cross-disciplinary experimental training.
Virtual simulation experiment for detecting welding residual stress using the finite element method
CHEN Dazhi;LIU Yan;CAO Yujie;ZHANG Pengju;CHEN Hui;CHEN Jingqing;[Objective] A prominent issue in the welding of large engineering structures is the accumulation of residual welding stress， which remarkably compromises the structural strength and service life of welded components. This study focuses on laser–Metal Active Gas Arc Welding (MAG) composite welding， using 16MnR steel as the research subject， and performs virtual simulation experiments for residual stress detection based on the finite element method. The study examines the evolution of temperature and stress under the welding thermal cycle and analyzes the distribution of residual stress in the welded workpiece. The simulation results are validated through measurements of weld pool morphology and post-welding residual stress. [Methods] This study focuses on the laser–MAG composite welding process and employs the finite element simulation software ABAQUS for virtual simulation experiments. The main steps include: 1) creating a geometric model of a 16MnR steel flat plate based on the actual dimensions of the welded parts; 2) setting thermal and mechanical parameters of the base and welding materials， including thermal conductivity， heat capacity， density， Poisson’s ratio， thermal expansion rate， elastic modulus， and yield strength; 3) applying convective heat transfer and radiation boundary conditions to all outer surfaces of the workpiece; 4) using a rotating body heat source model to characterize the energy distribution of the laser heat source and coupling it with a double-ellipsoid heat source model to represent the arc heat source; 5) applying a sequential coupling method to import the transient temperature field into the stress evolution calculation during welding; 6) adopting constraint conditions consistent with the actual welding process in the stress field calculation; and 7) verifying the simulation predictions with measured weld pool morphology and residual stress after welding. [Results] The results of the virtual simulation experiments for residual stress detection in laser–MAG composite welding based on the finite element method are as follows: 1) a finite element model of thermomechanical coupling in laser–arc composite welding was established; 2) the area with high transverse residual tensile stress is concentrated in the heat-affected zone on both sides of the weld， with a maximum value of 318.59 MPa; and 3) the large longitudinal residual tensile stress is mainly distributed in the weld seam and near-seam area， with a peak value of 503.89 MPa at the weld toe. [Conclusions] By using a rotating body heat source to model the laser heat source and a double-ellipsoid volume heat source for the arc heat source， an effective numerical simulation model for the temperature field in laser–MAG composite welding of 16 MnR steel can be established. The morphology of the temperature field obtained through finite element calculations agrees well with the actual weld shape. Based on thermoplasticity theory， the welding stress field of 16MnR steel was calculated using the derived temperature field. The calculated stress results were in good agreement with the measured values， accurately reproducing the actual evolution of residual stress during the welding process. These results provide strong guidance for the laser–MAG composite welding process.
Partial differential equation-based denoising and deblurring models and experiments
WANG Haobai;ZHANG Zhongbo;[Objective] Image denoising and deblurring are important research directions in image processing. The effectiveness and computational speed of denoising and deblurring algorithms are two key metrics for evaluating their performance. [Methods] Existing methods for image denoising and deblurring can be classified into model-driven， data-driven， and joint data/model-driven approaches. Among these， model-driven methods mostly solve the problem using second-order or fourth-order partial differential equations (PDEs). Although these methods perform well in certain aspects， they each have their limitations. This study improves the You–Kaveh (YK) model to enhance its convergence. Specifically， we first use the YK model to construct a fourth-order PDE for image denoising and deblurring and then introduce a third-order term to achieve reverse diffusion during the solution process. Subsequently， the resulting PDE is fully discretized in both time and space， and numerical solutions are obtained using finite differences. Finally， we conducted numerical experiments on multiple classical and remote sensing images to validate the proposed model and compared it with the Total Variation model and the YK model. [Results] Results showed that the proposed model notably improves the convergence and computational speed of the algorithm without compromising image restoration quality. [Conclusions] Furthermore， the proposed model exhibits superiority in both image restoration quality and computational efficiency， showing great potential for wide application.
Experimental design for single-phase grounding fault section selection in distribution networks based on transient zero-sequence power characteristics
GAO Wei;ZHENG Zijie;LI Yufeng;LIN Jianxin;GUO Moufa;[Objective] A single-phase grounding fault is the most frequent type of fault in distribution networks， and identifying the fault accurately and locating the fault section for the safe operation of power networks is crucial. However， because of small fault current， weak fault characteristics， and poor dispersion， the accuracy of fault location in engineering practice is low. In order to enhance the quality of experimental teaching， students’ cognition of single-phase grounding fault in distribution networks， and their practical analytical skills， this paper presents an experimental design scheme for the resonant grounding fault segment in distribution networks. [Methods] First， the design method of the single-phase grounding fault experiment platform is described， including the components of the 10 kV true type experiment system and the hardware and software designs of the single-phase fault grounding current identification terminal to build a real experimental environment for the research. Second， the design scheme of the distribution network grounding fault section is proposed. After an in-depth analysis of the polarity differences between the transient zero-sequence signals in the healthy section and the faulty section， it is found that when transient zero-sequence power is directly calculated， misjudgment will occur because of the influence of arc suppression coil compensation and noise interference. However， some frequency band information of the zero-sequence power contains characteristic information that can effectively suppress the interference. Based on this， the transient zero-sequence power is decomposed by a wavelet packet algorithm， in which the frequency band with strong energy entropy can stably reflect the state characteristics of the line. Considering the randomness and uncertainty of the fault arc and the presence of field noise， the high-entropy characteristic frequency bands of each fault sample are not the same， and the number of characteristic frequency bands available for selection is large. Therefore， this paper further proposes a feature frequency band optimization method based on a genetic algorithm and wavelet packet energy entropy. In the offline stage， the genetic algorithm is used to screen the massive historical data and optimize the selection of feature frequency bands to improve the identification accuracy of the fault section. In the single calculation stage， the five frequency bands with the highest wavelet packet energy entropy are selected as the characteristic frequency bands， and then， the fault characteristic frequency bands with strong characterization ability are determined by intersection operation. This method can effectively suppress the noise interference and enhance the identification performance of the fault section. [Results] Experimental results show that the proposed method is high adaptive， requires no threshold in ground fault current identification， and achieves an accuracy of 95.24%. Compared with other methods in the literature， the proposed method can maintain higher recognition performance under various working conditions. Finally， hierarchical experiment schemes are designed according to different curriculum requirements for students to gradually master the whole process from fault mechanism analysis to advanced intelligent identification methods. [Conclusions] The experimental platform and single-phase grounding fault identification method not only improve the students’ hands-on and engineering application capabilities but also lay the foundation for further research and application of single-phase grounding fault detection technology in distribution networks and provide an innovative practical teaching model for the cultivation of talent in the field of power engineering.