153results about How to "Deepen understanding" patented technology

The invention discloses a shear capacity analysis method for a reinforced concrete beam with stirrups based on machine learning, and the method comprises the steps: constructing a training set and a test set based on a shear capacity database of the beam with stirrups; establishing an interpretable prediction model by utilizing machine learning, setting hyper-parameters for the interpretable prediction model, and performing training; obtaining a shear bearing capacity prediction value of the samples in the test set by using the trained prediction model; and quantifying the contribution of each key anti-shearing characteristic to the anti-shearing bearing capacity predicted value, and analyzing and judging the reasonability of the anti-shearing bearing capacity predicted value output by the prediction model according to the anti-shearing theory. According to the method, the problems of high prediction difficulty, high discreteness, lack of reliable physical models, large calculation amount of high-precision numerical simulation and the like in machine learning are solved for the first time; the prediction model can accurately reflect the complex nonlinear relation between the shear bearing capacity and the influence factors, and meanwhile, the model reasonability is verified and improved in combination with a prediction result and priori knowledge, so that the man-machine interaction trust in practical application is improved.

The invention discloses a feature classifier-based thinking chain reasoning mathematical problem solving method and system. The method comprises the following steps of: (1) acquiring a mathematical problem set, and selecting forward and reverse thinking chain examples to splice to form a new set; (2) respectively inputting the new set elements into a model for reasoning to construct a word-level reasoning path spanning tree, and selecting an attention weight matrix after pooling difference processing as node features for storage; (3) traversing all the generated reasoning path spanning trees, and screening nodes meeting requirements to construct a feature classifier training set; (4) training a feature classifier by using a support vector machine algorithm; and (5) through the trained feature classifier, participating in path selection in a pre-training language model reasoning process, and obtaining a relatively accurate reasoning process and answers. By means of the method, adjustment and control of finer granularity of the reasoning path of the pre-training language model can be achieved, accurate reasoning solving of the pre-training language model on mathematical problems is facilitated, and the generalization level of the pre-training language model is improved.

The invention provides a large model-based generative visual document zero sample information extraction method. The method comprises six unit architectures: data preprocessing, a multi-modal feature extraction task, a layout enhancement task based on a large language model, a feature fusion task, an adapter pre-training task and an information extraction task based on the large language model. According to the method, feature fusion and processing are carried out, so that a large language model executes an information extraction task; training is carried out through a text region prediction task and a coordinate mapping text recovery task so as to enhance the feature processing capability; according to the method, all aspects of features of the visual document and the understanding ability of the large language model are fully utilized, the accuracy of document information extraction is improved, document information extraction of predefined entity categories is not needed, therefore, the service range of a supply chain semantic dictionary is rapidly expanded, the flexibility and efficiency of information processing are remarkably improved, and the method is suitable for application and popularization.

The invention relates to the field of medical model training, and particularly discloses a medical treatment guide model training method and system, a terminal and a medium, and the method comprises the steps: carrying out the incremental pre-training of a selected large language model through priori knowledge, and obtaining a basic large language model; acquiring a public real doctor-patient dialogue data set; constructing a doctor-patient dialogue template, and generating and constructing a doctor-patient dialogue data set by using a third-party large language model based on the doctor-patient dialogue template; combining the real doctor-patient dialogue data set and the constructed doctor-patient dialogue data set into a data set, and recording the data set as an initial doctor-patient dialogue data set; performing duplicate removal preprocessing on the initial doctor-patient dialogue data set to obtain a doctor-patient dialogue training data set; and using the doctor-patient dialogue training data set to carry out supervision fine tuning training on the basic large language model to obtain a doctor-seeing and hospital-guiding model. According to the invention, the joint training data set is formed based on the public data set and the automatic construction data set, so that the accuracy of the medical treatment guide model is improved.

The invention belongs to the technical field of bridge monitoring, and particularly discloses a dynamic visual bridge overall displacement monitoring and behavior twinborn analysis method and system. Comprising the steps of obtaining bridge member displacement monitoring data; performing time domain feature and frequency domain feature analysis on the displacement monitoring data; based on the internal change rules of the bridge structure displacement under different external conditions, establishing displacement response models under different working conditions so as to identify an abnormal displacement mode; calculating spatial correlation among different components to analyze interaction and influence of the different components under the action of the dynamic load; predicting the change trend of the bridge displacement data based on an ARIMA time sequence model; a causal inference method is utilized to analyze reasons of displacement changes of different components, and key influence factors are identified. By analyzing the spatial distribution characteristics and the time evolution characteristics of the displacement data of the key component, fusion and comprehensive utilization of the displacement monitoring data are realized, and the method is of great significance to health monitoring of a bridge structure.

The invention discloses a long-distance high-pressure water delivery pipeline water hammer and leakage monitoring and positioning test device, which comprises a water tank, a water outlet side experiment unit, a water return side main path experiment unit and a water return side branch experiment unit, and is characterized in that the water outlet side experiment unit and the water return side branch experiment unit are communicated with the water return side main path experiment unit; the water outlet side experiment unit, the water return side main path experiment unit and the water return side branch path experiment unit are all communicated with the water tank. The device solves the problem that tiny leakage is difficult to monitor in the high-pressure water delivery process in the existing related test experiment technology, and the problems that the water hammer experiment signal precision acquisition is insufficient and the water hammer experiment causes impact damage to equipment. The invention further discloses a using method of the long-distance high-pressure water conveying pipeline water hammer and leakage monitoring and positioning testing device.

The invention discloses a graph neural network music recommendation method for independent division and aggregation of large and small paths, and belongs to the field of deep learning, and the method comprises the following steps: 1, constructing a heterogeneous graph neural network classification and aggregation model which comprises a path discriminator, a large-path neighbor node selection module and an intra-path aggregation module; step 2, constructing a loss function optimization training heterogeneous graph neural network classification aggregation model; and step 3, acquiring music data of a current user, inputting the trained heterogeneous graph neural network classification aggregation model, and generating a personalized music recommendation list. The invention aims to create more personalized and accurate music recommendation experience for the user and help the user to explore richness and unique charm of the music more deeply, so that a more wonderful music world is presented for music lovers.

The invention discloses a non-linear production and confluence characteristic analysis method for a valley terrain extra-small watershed. The method comprises the steps that S1, valley terrain-hydrological data are acquired, and hydrological grid units are spatially divided; s2, identifying hydrological behavior characteristics of the hydrological grid units to form independent hydrological response units; s3, analyzing a first influence degree formed by rainfall infiltration, a second influence degree formed by surface runoff and a third influence degree formed by dynamic change of soil moisture in the hydrological response unit, and constructing a valley dynamic fluid model; s4, combining the valley dynamic fluid model with the underground water flow model and the surface water flow model to form a hydrological-hydrodynamic simulation framework; and S5, simulating various rainfall scenes by using a hydrological-hydrodynamic simulation framework, performing analysis in combination with valley microtopography factors, calculating flow change rates of different rainfall scenes, and identifying nonlinear features. According to the method, high-precision simulation and prediction of the valley terrain extra-small watershed are realized, and the accuracy and timeliness of productivity and confluence analysis are improved.

The invention discloses a gas-solid two-phase flow numerical simulation method for bidirectional coupling of non-spherical particles and a flow field, and belongs to the technical field of biomass fuel utilization, and the method comprises the following steps: constructing a flow field model in Fluent, and correspondingly setting an initial boundary condition of the flow field; setting initialization conditions of non-spherical particles in the UDF; establishing an inertial coordinate system and a particle coordinate system, and coupling the two coordinate systems; and obtaining a translational motion table / rotational motion characterization equation, determining the position and direction of the particles by solving the two equations, correcting flow field parameters by using the positions and direction of the particles, and iteratively calculating particle motion parameters to realize bidirectional coupling of the non-spherical particles and the flow field. According to the numerical simulation method, through Fluent-UDF numerical simulation, the interaction between the non-spherical particles and the flow field can be accurately simulated, the accuracy of gas-solid two-phase flow numerical simulation is guaranteed, understanding of the complex process of direct combustion coupling of biomass particles of a coal-fired boiler is deepened, and theoretical support is provided for optimization of combustor design and the combustion process.

The invention belongs to the technical field of coal bed gas exploitation fracturing control, and particularly discloses a coal bed gas exploitation staged fracturing intelligent monitoring control system which collects rock samples from a target coal bed, conducts microstructure detection on the collected rock samples, removes the rock samples with natural fractures, and obtains the coal bed gas exploitation staged fracturing intelligent monitoring control system. Then a servo motor is used for applying the same stress in different directions to the rock sample for fracturing simulation, crack propagation monitoring is conducted after fracturing simulation, and therefore the crack propagation direction of the rock sample is associated with the stress applying direction to obtain crack propagation dominant microstructure distribution characteristics corresponding to the rock sample; the understanding of the crack propagation mechanism is deepened to the greatest extent, the definition of the crack propagation direction is increased, the crack propagation effect is optimized when fracturing operation is carried out in the coal seam without natural cracks, and the accuracy and effectiveness of fracturing are improved.

The invention relates to a digital human personalized emotion interaction method, which belongs to the technical field of emotion calculation, and comprises the following steps: S1, carrying out emotion quantification on interaction input of a user to obtain a pleasure-wake-dominate PAD emotion value in a continuous dimension space; s2, performing emotion state evaluation on the user interaction input emotion to obtain an interaction input emotion evaluation function; s3, calculating a digital human emotion transition probability matrix by using a Bayesian network; s4, calculating an emotional response calculation value of the digital human character corrected by the big five-factor model in the current emotional state; s5, the corrected emotion response passes through an emotion influence factor matrix, and a response with complex emotion features is output; and S6, performing PAD emotional space coordinate calibration on the emotional intensity of the digital person on the six basic emotions to obtain an emotional response of the digital person.

The invention belongs to the field of ionosphere satellite magnetic field seismic anomaly extraction, and particularly relates to a satellite magnetic field seismic anomaly extraction method based on complex non-negative matrix factorization, which comprises the steps of reading Swarm A satellite magnetic field data, and removing a main magnetic field of a ground core and a lithosphere magnetic field of an earth crust to obtain a residual magnetic field; performing short-time Fourier transform on the preprocessed data to obtain a complex time-frequency matrix in a rectangular coordinate representation form of a complex domain; decomposing the complex time-frequency matrix by using complex non-negative matrix decomposition to obtain R feature components; performing inverse short-time Fourier transform on each feature component to obtain a time domain feature component; a mean square root of a time domain seismic component of each orbit is calculated, a threshold value is set by using an over-limit rate method, and abnormal points are extracted from the seismic components of the orbits. The amplitude and phase information contained in the time-frequency domain of the satellite observation data is fully utilized, so that the component related to the earthquake is identified, and the earthquake anomaly extraction is more accurate and reliable.

The invention discloses a reservoir four-relation chart processing method, device and equipment and a medium, and relates to the field of oil well exploration, and the method comprises the steps: building a logging information parameter processing interpretation model, and determining parameters in an oil saturation model; based on the neutron-density intersection chart, calculating the distances from the logging response parameters to different lithologic curves as lithologic indication parameters; drawing a reservoir four-relation chart by taking the porosity of the stratum as an abscissa, the resistivity of the stratum as an ordinate and the lithology indication parameters as color mapping, drawing different water saturation scale marks in the reservoir four-relation chart, and in addition, taking the water saturation or the movable hydrocarbon saturation in the reservoir four-relation chart as a bubble radius; and projecting the logging response parameters of various stratum types and the explaining stratum parameter points into the reservoir four-relation chart, and establishing a logging explaining standard. According to the method, the dynamic change and internal relation among the four properties of the reservoir can be comprehensively and accurately displayed.

The invention discloses a method for monitoring the bonding slip performance and deterioration evolution of an interface in a combined member, and the method comprises the following steps: carrying out the degreasing cleaning pretreatment of a section with relatively high rigidity in a to-be-detected combined member, and obtaining a pretreated section; distributed optical fiber sensors based on Rayleigh scattering are arranged on the surface of the pretreated profile in the direction of measured cohesive force; bonding the profile on which the distributed optical fiber sensors are arranged with a to-be-detected matrix through an adhesive to prepare a combined component; a uniaxial tensile test is conducted on the combined component, distribution and evolution of steel bar strain in the test process are recorded through a distributed optical fiber sensor, distribution and evolution of the bond-slip relation are obtained, the local bond-slip relation can be researched, and the applicability and precision of a bond-slip model are improved; before uniaxial stretching, working conditions are applied to the composite structure to degrade the structure, and the influence of degradation on the bonding performance of the composite structure can be evaluated.

The invention relates to the technical field of three-dimensional scanning, in particular to a 3D printing product personalized design method based on a three-dimensional scanning human body model. The method comprises the following steps: acquiring human body three-dimensional scanning data and human body composition data; performing movable joint analysis on the human body three-dimensional scanning data to generate human body movable joint data; performing distributed three-dimensional model modeling on the human body three-dimensional scanning data based on the human body movable joint data to generate a human body three-dimensional model; initial human body fitting product design is conducted on the human body three-dimensional model based on the 3D printing technology, and initial 3D printing product parameter data are obtained; according to the method, by combining the human joints, the body composition data and the user feedback, the fitness and the personalized design precision of the 3D printing product are optimized, and the precision and the flexibility of the personalized design of the 3D printing product are improved.

The invention discloses a rusted high-strength bolt connection pretightening force intelligent self-optimization prediction method and system, and belongs to the technical field of bolt pretightening force detection.The method specifically comprises the steps that a finite element database of bolt connection pretightening force loss is established, and a machine learning prediction model based on a neural network is trained; when the recognition accuracy of the machine learning prediction model reaches a preset value, pre-tightening force loss of the target bolt is monitored in real time, and real-time geometric feature information of connection of the target bolt is obtained; when the similarity between the real-time geometric feature information of the target bolt connection and a finite element model in the finite element database is greater than or equal to a preset value, calling the pre-tightening force loss of the existing finite element database; when the similarity is lower than a preset value, the trained machine learning prediction model is utilized to output actual bolt connection pre-tightening force loss in real time; the bolt looseness detection device realizes automatic and marker-free bolt looseness detection, is high in bolt image acquisition angle and distance tolerance, and can meet bolt looseness detection at different positions.

The invention discloses a high-temperature and high-pressure visual elemental sulfur reaction device and an elemental sulfur analysis method. The reaction device comprises a reaction kettle mechanism, a heating mechanism, a pressure control mechanism and a visual monitoring mechanism; the analysis method comprises the following steps: (1) transferring a sample under simulated temperature and pressure conditions; (2) checking the gas tightness of the natural gas elemental sulfur equipment; (3) measuring elemental sulfur and the like. The device can be used for elemental sulfur reaction under high-temperature and high-pressure conditions, meanwhile, the visual window can observe the whole reaction evolution process in real time and online through a high-definition camera monitoring system, videos are recorded, and a playback viewing function is achieved; the problem of detecting the content of natural gas dissolved elemental sulfur in oil and gas reservoir fluid is solved.

The invention discloses a multi-field coupling thrust calculation method for a carrier rocket interstage pneumatic separation system. The method comprises the following steps: 1, acquiring control parameters of the pneumatic separation system under an operation condition; 2, determining a real gas state equation in the gas cylinder; 3, deducing to obtain dynamic changes of gas pressure, mass and temperature in the gas cylinder; 4, valve and pipeline dynamic modeling is carried out; 5, establishing a high-speed high-pressure cylinder mathematical model; 6, judging whether the inter-stage separation performance of the first and second substages meets the requirement or not; if the requirement is met, the thrust time history response of the pneumatic separation system is output; and if not, returning to the step 1. The method not only can accurately obtain the thrust change characteristics of the carrier rocket interstage pneumatic separation system under the multi-field combined action, but also can obtain the dynamic response characteristics of the separation process of the first substage and the upper stage, and has reference significance for design optimization of the interstage pneumatic separation system.

The invention relates to the field of carbon neutralization, and discloses a four-dimensional soil moisture drought event tracking and warming signal identification method, which comprises the following steps: data acquisition; four-dimensional soil moisture drought event identification; four-dimensional soil moisture drought event characteristic quantification; artificial warming signal detection of time evolution of four-dimensional soil moisture drought characteristics in historical periods; performing four-dimensional soil moisture drought future evolution estimation; four-dimensional soil moisture drought event driving factors; the method has the advantages that future space-time evolution characteristics of four-dimensional soil moisture drought are explored, driving factors and physical mechanisms of four-dimensional soil moisture drought are analyzed, and a powerful scientific basis is provided for coping with global climate changes, guaranteeing water resource safety and formulating disaster prevention and reduction strategies.

The invention provides an intelligent data auditing method and system based on large language model fine tuning. The method comprises the steps that A1, a cue word template is constructed for an auditing task; a2, acquiring original business data, and forming a labeling data set of each auditing task based on the cue word template; step A3, randomly disrupting and mixing the annotation data set of each auditing task to form a fine adjustment data set; step A4, performing fine tuning on the pre-trained large language model by using the fine tuning data set to obtain a fine-tuned large language model; step A5, assembling the business data to be audited according to the prompt word templates of the audit tasks to form prompt words of the audit tasks; and A6, reasoning the assembled cue word by using the fine-tuned large language model, and outputting an auditing result. The understanding of the large language model on industry knowledge is deepened, and the auditing accuracy is improved.

The invention relates to the technical field of geology research in oil exploration and development, and provides a shale oil reservoir pore space quantitative characterization method and system, and the method comprises the steps: carrying out the statistics of micron-order equivalent pore structure parameters of a shale oil reservoir; carrying out statistics on nanoscale organic pore structure parameters of the shale oil reservoir; performing standardization treatment on the micron-scale equivalent pore structure parameters and the nano-scale organic pore structure parameters to obtain theoretical volumes of standardized pore structures in different intervals under different resolutions of the reservoir sample; respectively calculating the volume fraction of the theoretical volume in the standardized total pores and the volume fraction of the theoretical volume in the reservoir sample; and establishing a full-scale pore distribution curve according to the volume fractions, and representing the pore volume proportions of different apertures of the shale oil reservoir. According to the method, CT scanning and FIB-SEM are organically combined in the shale oil reservoir space quantitative characterization process, and visual quantitative characterization of a shale sample from micron-scale bedding seams to nano-scale apertures is achieved.

The invention relates to the technical field of biology, in particular to application of aldolase A protein 330 amino acid modified posttranslational protein in tumor resistance and a mutant. The aldolase A protein mutant is obtained by carrying out substitution on an amino acid sequence as shown in SEQ ID No. 1; and lysine at the 330 position is substituted by arginine. Experiments prove that the aldolase A protein mutant can significantly reduce the growth speed of mouse malignant breast cancer cells 4T1 and mouse malignant black tumor cells B16-F10 under the skin of a mouse, and shows excellent antitumor activity. A brand-new tumor treatment method is created, and the technical blank in the field is filled.

The invention discloses an intelligent display system and method for art design teaching, and belongs to the technical field of art teaching, and the method specifically comprises the steps: shooting a canvas image in the drawing process of a teacher, synthesizing the canvas image into a work image, and recording a white balance parameter in the shooting parameters of the canvas image; detecting light color temperature values of environments where teachers and students are located by using a sensor, and respectively marking the light color temperature values as a source color temperature value and a target color temperature value; correcting a white balance parameter of the work image according to the source color temperature value to generate a corrected work image; setting a standard interval based on the source color temperature value, and when the target color temperature value is in the standard interval, directly displaying a corrected work image at a student terminal; when the target color temperature value is not in the standard interval, respectively calculating tristimulus values corresponding to the two color temperature values, constructing a color matrix, shifting image pixels by using the color matrix, and outputting a shifted image; according to the invention, the teaching efficiency in art design education is improved.

The invention relates to the technical field of targeted intervention of chordoma, in particular to application of a medicine to treatment of chordoma of different molecular types. The invention provides application of anlotinib, an oxidative phosphorylation inhibitor and desumab in preparation of products for preventing and / or treating chordoma. The chordoma comprises interstitial epithelial transformed chordoma, mesenchymal derived chordoma and bone microenvironment dominant chordoma. The invention also provides a pharmaceutical composition, which comprises an effective dose of anlotinib, and / or an oxidative phosphorylation inhibitor, and / or desumab. Under the condition that current treatment selection is limited, proliferation, clone formation and tumor growth of tumor cells can be remarkably inhibited by applying the medicine to specific molecular subtype chordoma, and compared with existing treatment medicine, the medicine has a better treatment effect, has good safety and shows potential clinical application value.

The invention discloses a network security risk assessment system for optimizing digital society, and particularly relates to the field of network security, comprising an assessment data acquisition module, a dynamic model training module, a context awareness module, a multi-dimensional analysis module, a decision response module, a risk comprehensive assessment module and a continuous improvement and feedback module. According to the network security risk assessment system for optimizing the digital society, through quantitative analysis of data set features, understanding of an existing situation is deepened, prediction accuracy of future possible occurrence conditions is improved, and a decision path is optimized; by ensuring that the decision-making process is based on static data, a dynamically changing environment is flexibly adapted, and challenges caused by uncertainty are effectively coped with; decisions and response effects transmitted by the risk comprehensive evaluation module are fed back to a training model in the dynamic model training module, the accuracy and response speed of the training model are further optimized, the performance of the context awareness module can be continuously improved, and the context awareness module is more intelligent and high in adaptability.

The invention discloses a molecular attribute prediction method and system based on multistage knowledge distillation, and belongs to the technical field of molecular representation learning and molecular property prediction. Comprising the steps of obtaining molecular representation symbols, and determining molecular text prompt features by using a large language model based on the molecular representation symbols; obtaining a molecular topological graph, and processing the molecular topological graph and the molecular text prompt features through a trained molecular attribute prediction model to obtain a molecular attribute prediction result; wherein when the encoder in the molecular attribute prediction model is trained, multi-stage knowledge distillation is carried out in combination with multi-stage actions of molecules. The molecular structure is comprehensively analyzed, comprehensive understanding of molecules can be deepened, and the problem that existing molecular structure analysis is limited by task requirements and local information processing is solved.

The invention belongs to the technical field of drug screening, and particularly relates to an interpretable drug molecule property prediction method based on distribution drift consistency. Comprising the following steps: acquiring a drug molecular map structure data set and selecting positive and negative sample sets of to-be-identified drug molecules from the data set; processing the two sample sets by adopting a basic post-interpretation model to obtain a candidate interpretation sub-image set; processing the candidate interpretation sub-atlas by adopting an agent generator to obtain an agent atlas; inputting the agent image set into a trained GNN model for processing to obtain drug molecular property prediction results of the two sample sets; calculating agent graph comparison learning loss and adjusting parameters of an agent generator; calculating, predicting and acquiring the total loss of the interpretation sub-graphs, and adjusting basic post-interpretation model parameters; repeatedly training until a preset maximum number of iterations is reached, and obtaining a trained prediction model; according to the method, the reliability and interpretability of the compound generation model and the classification model are remarkably improved, so that the compound characteristic analysis is more intuitive and transparent.

The invention discloses a method for simulating halogen plasma etching silicon or silicon dioxide through molecular dynamics. The method comprises the steps of modeling through a silicon or silicon dioxide substrate material, performing LAMMPS software initialization setting, performing plasma etching circulation, and performing data output and analysis processing. According to the method, a plasma-surface interaction mechanism is simulated on the atomic scale by using a molecular dynamics method, the property of a surface modification layer is analyzed, the simulated etching conditions are independent and controllable, plasma etching with different energies, incident angles and incident doses can be simulated, and the influence of a single variable on an etching result can be accurately analyzed; the method can accurately and quickly predict the effect of etching silicon or silicon dioxide by the halogen plasma under different incidence conditions, avoids high cost, complexity and uncertainty caused by experiments, and greatly shortens the development process of the process of etching silicon or silicon dioxide by the halogen plasma.

The invention discloses a multi-agent-based space-time activity analysis and prediction method for urban empty nesters. Comprising the steps of collecting extrinsic features of community residents and constructing a space-time activity sand table, identifying urban empty-nest elderly people and staying points thereof, constructing a typical space-time activity chain of the urban empty-nest elderly people, constructing a multi-agent model of the urban empty-nest elderly people, simulating and screening demands of the empty-nest elderly people, and generating a public service facility layout optimization scheme. And the requirements of the empty nesters are visualized. According to the method, high-precision resident extrinsic features and high-resolution spatial data are acquired, and a multi-agent model is adopted to simulate and screen the demands of the empty-nest elderly, so that full-coverage, precise and multi-dimensional identification of the activity demands of the empty-nest elderly under an urban scale is realized; and a reliable basis is provided for perfecting urban community public service facilities and improving the life quality of old people.

The utility model relates to a phase change thermo-acoustic engine experimental device which comprises a thermo-acoustic conversion system and an acoustic-electric conversion system. The thermo-acoustic conversion system comprises a sound production module and a Helmholtz resonant cavity, a ceramic plate stack soaked with water exists in the sound production module, and the stable temperature difference needed for generating the thermo-acoustic effect at the two ends of the ceramic plate stack is reduced. Sound waves generated by the thermoacoustic effect form standing waves in the Helmholtz resonant cavity, and the sound intensity is amplified; the acoustic-electric conversion system comprises piezoelectric ceramics and a triode common-emitter amplification circuit, the piezoelectric ceramics are arranged at the tail end of the resonant cavity and convert acoustic energy into electric energy, and the electric energy is amplified by the triode common-emitter amplification circuit and then drives a nixie tube to display. The experimental device integrates thermodynamics, fluid mechanics, acoustics, material science and electronic technology, has the characteristics of simple structure, low cost, rich physical process and strong exploration, is suitable for experimental teaching, particularly suitable for new engineering teaching requirements, and is beneficial to improving the multidisciplinary cross practice ability of students.