A semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology

Abstract

The application provides a semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology, comprising: an automatic full-link multi-working-condition semiconductor detector particle pulse waveform simulation system for controlling GEANT4, Weightfield2 and PyLTSPICE software, outputting simulated energy data and differential current data; a double-channel pulse waveform automatic acquisition system for collecting and storing accelerator measured energy data and current data by using physical entities and displaying in multiple channels; and a digital twin and physical twin data comparison and analysis system for analyzing waveform matching degree between accelerator measured waveforms and simulated waveforms under the same working condition and the same circuit parameters. The application has the advantages of efficiently and automatically generating massive pulse waveform data under different working conditions.

Description

Technical Field

[0001] This application belongs to the field of particle detection technology, specifically relating to a semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology. Background Technology

[0002] Charged particles in space include galactic cosmic rays, solar cosmic rays, the solar wind, and the Van Allen radiation belts. Galactic cosmic rays consist of high-energy charged particle streams originating from all directions of the Milky Way, primarily composed of protons, alpha particles, and other nuclear components. Solar cosmic rays are high-energy charged particle streams emitted during solar flares, mainly composed of protons. The solar wind is a continuous outward ejection of plasma from the Sun, primarily composed of hydrogen plasma, with minor amounts of other components. The Van Allen radiation belts, also known as the Earth's radiation belts, are located near Earth and are radiation belts captured by the Earth's magnetic field. They are divided into inner and outer radiation belts and are mainly composed of protons and electrons. To study the diverse range of charged particles in space, a precise understanding of their energy spectra, types, and projection angles is necessary for further research into particle origins, acceleration and propagation processes, lifetimes, and dynamics.

[0003] Accelerator experiments are currently the most direct and commonly used method for acquiring pulse waveform data of charged particles incident on semiconductor detectors, and are widely used in particle physics, high-energy physics experiments, and detector performance evaluation. By providing high-energy particle beams through accelerators, researchers can accurately measure the signal response produced by particles in the detector, providing reliable reference data for detector design, optimization, and experimental verification. Despite the significant importance of accelerator experiments, they still have considerable limitations in practical applications:

[0004] 1. Limitations on particle type and energy range:

[0005] The types and energy ranges of particles generated in accelerator experiments are usually limited by the performance of the equipment, which cannot cover all possible physical conditions. Complex experimental requirements may require a lot of time to adjust.

[0006] 2. High costs and resource sharing issues:

[0007] Accelerators have extremely high operating costs, including equipment maintenance, energy consumption, and personnel costs.

[0008] The experimental time for equipment is usually a shared resource, which requires waiting for scheduling and allocation, resulting in delays in research progress;

[0009] Multiple experiments, adjustments, and tests often require a significant amount of time to cover a wide range of parameters, further increasing costs and timelines.

[0010] 3. Detector loss and the stability of experimental results:

[0011] Prolonged exposure to high-energy particle beams can lead to the accumulation of lattice defects in the detector, thereby affecting the accuracy of test results and the long-term performance of the equipment.

[0012] Given the high cost and limited experimental conditions of accelerator experiments, many researchers have begun to utilize simulation methods to study and analyze the physical processes and electronic performance of charged particles incident on semiconductor detectors. Among these, GEANT4, as a highly efficient tool based on the Monte Carlo method, can accurately simulate the trajectory of charged particles within the detector and their energy deposition process by defining the particle source, detector materials, and geometry. Furthermore, combined with physical process models, GEANT4 can meticulously describe the electromagnetic interactions and nuclear reactions between particles and detector materials, providing crucial support for detector design optimization and performance prediction.

[0013] However, GEANT4 simulation results are limited to data on the physical processes of particle energy deposition in the detector, and cannot directly reflect the detector's output signal. Therefore, many researchers have attempted to combine GEANT4 with SPICE software, using SPICE tools (such as PSPICE or LTSPICE) to evaluate the circuit performance of the preamplifier and main amplifier. SPICE tools, through precise transient simulation analysis, can study the circuit's gain characteristics, time response, and noise impact on the input signal, thereby optimizing the signal readout path. Furthermore, this combination can achieve a comprehensive simulation of the particle detector system-level performance to some extent, improving design reliability and efficiency. Nevertheless, the simulation method combining GEANT4 and SPICE software typically only obtains peak information of the pulse signal, lacking accurate simulation of the full-time waveform, and cannot fully reflect the dynamic changes during particle incidence, limiting a deeper understanding of particle behavior.

[0014] To address this issue, some researchers have begun to combine the Weightfield2 tool to simulate the motion of charge carriers in semiconductor detectors under the influence of an electric field. Weightfield2 is a highly efficient tool that can accurately predict the characteristics of the detector's output signal by rapidly calculating the electric field distribution and the drift and diffusion processes of charge carriers, providing crucial support for detector design and optimization. Compared to simply combining GEANT4 and SPICE, this method can more directly simulate the transient response characteristics of the sensor.

[0015] However, current simulation methods still have the following significant shortcomings:

[0016] 1. Lack of automated data generation capabilities across multiple operating conditions:

[0017] Current simulation processes typically require researchers to manually adjust parameters such as particle type, energy, and incident angle, which cannot efficiently cover a variety of experimental conditions, especially under extreme conditions. The lack of automated simulation systems leads to low data generation efficiency, failing to meet the demand for large-scale, high-quality data for data-driven technologies such as deep learning.

[0018] 2. Difficulty in achieving seamless integration across tools:

[0019] Tools such as GEANT4, Weightfield2, and SPICE each have their own focus, but their combination often relies on complex interface conversions and unified intermediate data formats, resulting in fragmentation of the entire process and reducing the ease of use and efficiency of simulation.

[0020] 3. Lack of experimental noise and actual conditions:

[0021] Current simulation data is typically idealized and does not account for unavoidable noise interference and equipment defects in physical experiments, such as electromagnetic noise and material inhomogeneity. The lack of these factors can cause simulation results to deviate from real-world application scenarios, reducing the model's practical reference value. Summary of the Invention

[0022] The purpose of this application is to overcome the shortcomings of existing technologies, such as the lack of automated data generation capabilities for multiple operating conditions and the difficulty in achieving seamless integration across tools.

[0023] To achieve the above objectives, this application proposes a platform for generating and analyzing particle pulse waveform data of semiconductor detectors based on digital twin technology, the platform comprising:

[0024] An automated, end-to-end, multi-condition semiconductor detector particle pulse waveform simulation system is used to control GEANT4, Weightfield2, and PyLTSPICE software, and output simulated energy data and differential current data.

[0025] A dual-channel pulse waveform automated acquisition system is used to acquire, store, and display measured energy and current data from accelerators using physical entities; and

[0026] The digital twin and physical twin data comparison and analysis system is used to perform waveform matching degree analysis by comparing the measured waveform of the accelerator obtained by the dual-channel pulse waveform automated acquisition system with the simulated waveform under the same operating conditions and circuit parameters obtained by the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system. The analysis process includes comparing the rise slope of energy information, the pulse width of differential current, the mean square error, and wavelet transformation.

[0027] As an improvement to the aforementioned platform, the processing procedure of the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system includes:

[0028] Run the GEANT4 software simulation to output multiple incident energy loss data and plot the curves;

[0029] Run the Weightfield2 software to simulate and output the detector time-domain waveform data based on the incident energy loss data;

[0030] The PyLTSpice software is used to simulate and output the final dual-channel energy and differential current data based on the time-domain waveform data.

[0031] As an improvement to the aforementioned platform, the dual-channel pulse waveform automated acquisition system includes:

[0032] A Si-PN junction ion implantation semiconductor sensor is used to input particles from a calibration source to generate a corresponding pulse current signal.

[0033] A preamplifier is used to amplify the pulse circuit signal and output energy and current signals.

[0034] The main amplifier is used to amplify the energy and current signals output by the preamplifier again.

[0035] A high-speed acquisition board is used to acquire the energy signals and circuit signals output by the main amplifier.

[0036] As an improvement to the aforementioned platform, the data processing procedure of the dual-channel pulse waveform automated acquisition system includes:

[0037] The waveforms of each energy signal and current signal acquired by the high-speed acquisition board are preprocessed to obtain the peak-to-peak value and a peak-to-peak value distribution histogram is plotted.

[0038] Based on the distribution of peak-to-peak value histogram, the interval with the maximum peak-to-peak value count is obtained, and a counting interval is extended outward from this interval on both sides as valid data, while low-quality waveforms obtained due to noise or abnormal acquisition are discarded.

[0039] Find the position of the maximum value of the waveform and extract several points to the left and right of it as output information. If there are not enough points, fill them with the last value.

[0040] As an improvement to the aforementioned platform, the analysis process of the digital twin and physical twin data comparison and analysis system includes:

[0041] Wavelet transforms were performed on the simulated and measured energy and current data respectively, and the number of decomposition levels was determined to generate denoised waveform data.

[0042] Matching analysis is performed on the energy waveform data. The slope of the rising edge of the energy output waveform in the two modes is calculated at different times, and the difference between the slopes is squared to determine the difference. The peak-to-peak difference is calculated to assess the deviation of the absolute energy information.

[0043] Matching analysis is performed on the current waveform data. By performing softmax normalization on the two waveform data and calculating the cross-entropy, the difference in shape can be determined. The difference in peak-to-peak value of the current can be calculated to measure the overall consistency of the current intensity.

[0044] Compared with existing technologies, the advantages of this application are:

[0045] 1. Efficiently and automatically generate massive amounts of pulse waveform data under different working conditions (different temperatures, different incident angles, different energies, different particles, front and rear incident, and different bias voltages);

[0046] 2. The simulation model can be evaluated and optimized using measured data;

[0047] 3. Data generation and analysis platforms can be used to obtain data results that are difficult to simulate from ground-based accelerators;

[0048] 4. Data obtained from the data generation and analysis platform can be applied to newly developed equipment as a supplement to the calibration system. Attached Figure Description

[0049] Figure 1 The diagram shows the structure of a semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology. Detailed Implementation

[0050] The technical solution of this application will be described in detail below with reference to the accompanying drawings.

[0051] With the rapid development of deep learning technology, more and more scholars are beginning to use multilayer perceptrons, convolutional neural networks, or recurrent neural networks for particle identification or position analysis. These methods rely heavily on massive amounts of pulse waveform data, but existing simulation techniques struggle to efficiently generate data with high coverage, multiple scenarios, and matching actual test conditions. This has become one of the bottlenecks restricting the application of deep learning technology, and digital twin technology holds promise for solving this problem. Digital twin technology is an emerging technology that combines physical systems with virtual models, achieving full lifecycle optimization through data interaction and simulation analysis. Its core components include physical entities, digital models, and data interaction interfaces. Its core feature is the high-precision mapping and real-time interaction between virtual and physical entities, allowing the state of complex systems to be intuitively reflected and dynamically optimized in the digital model. Although this technology has been widely applied in industries such as industry, aerospace, and energy, it has not yet been applied in the field of particle detection, especially in pulse waveform particle identification.

[0052] To address the challenges of high cost, time consumption, and low parameter coverage in accelerator experiments during particle detection, as well as the lack of automation in multi-condition simulations, absence of cross-tool data interfaces, and deviations from actual results in particle simulations, this application provides a semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology. This platform includes an automated, end-to-end multi-condition semiconductor detector particle pulse waveform generation system using the digital twin GEANT4+Weightfield2+PyLTSpice, an automated data acquisition system consisting of a physical twin calibration source, semiconductor sensors, preamplifiers and main amplifiers, a fast acquisition board, and host computer software, and a data comparison and analysis system between the digital twin and physical twin. Building upon the existing end-to-end semiconductor detector particle pulse waveform generation system, the digital twin embeds multi-condition automation functions into each simulation step, unifies data interfaces across tools, and utilizes the physical twin to automatically acquire and evaluate measured data under the same conditions. This lays a solid foundation for a comprehensive and in-depth understanding of the characteristics of different charged particle pulse waveforms and the application of deep learning in particle detection.

[0053] The semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology mainly includes three systems: an automated full-link multi-condition semiconductor detector particle pulse waveform simulation system, a dual-channel pulse waveform automated acquisition system, and a digital twin and physical twin data comparison and analysis system.

[0054] The simulation process of the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system includes three key steps: First, the GEANT4 simulation condition space is set using script control, such as energy, direction, particle type, step size, etc., and the multi-incident event simulation is started by reading the corresponding conditions. The output energy loss information and the corresponding condition space are recorded in txt format and sent to the next step. The script checks the log information file of GEANT4 simulation completion in real time. When it is completed, the Weightfield2 software is opened to read the energy loss txt file and the corresponding input conditions (energy, direction, particle type, step size), and the influence parameter space of the carrier transport process (temperature, front and rear incident, bias voltage, magnetic field, etc.) is set. The multi-incident event is simulated using Batch mode, and the final output result txt file includes hole-induced time-domain current waveforms, electron-induced time-domain current waveforms, and total induced time-domain current waveforms. Finally, the script monitors the Weightfield2 simulation log status file. When it is completed, PyLTSPICE is started to read the time-domain current waveforms of multiple incidents and, combined with the circuit parameters of the preamplifier, main amplifier, and shaping module, finally outputs energy data and differential current data.

[0055] The dual-channel pulse waveform automated acquisition system utilizes a low-noise energy output and differential current dual-channel preamplifier, a low-noise main amplifier, a high-speed acquisition board, and a multi-channel fast-view and fully automated data storage host computer to achieve high-speed pulse waveform acquisition from the accelerator and form a particle waveform database through preprocessing.

[0056] The digital twin and physical twin data comparison and analysis system uses the accelerator measured waveform obtained by the dual-channel pulse waveform automatic acquisition module and the simulated waveform under the same operating conditions and circuit parameters obtained by the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system to perform waveform matching degree analysis. The analysis methods include comparing the rise slope of energy information, the pulse width of differential current, the mean square error, and wavelet changes.

[0057] The system components of the semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology are as follows:

[0058] 1. Automated end-to-end multi-condition semiconductor detector particle pulse waveform simulation system

[0059] The system uses a top-level script to control the overall process. The software controlled includes GEANT4, Weightfield2, and PyLTSPICE, and finally outputs the energy and differential current waveforms.

[0060] 1) GEANT4 generates the condition space .mac file

[0061] GEANT4 generates .mac files with different energies, orientations, particle types, and step sizes. Energy options include 1 MeV, 3 MeV, and 5 MeV; orientation options include 0°, 15°, 30°, -15°, and -30°; and particle types include H, He, B, C, O, Al, Si, P, Fe, and Cu. The step size is 1 μm for H ions and 0.1 μm for other heavy ions due to their shorter range. The specific parameter space needs to be adjusted based on the actual testing conditions.

[0062] 2) Run the GEANT4 simulation, output multiple incident energy loss data, and plot the curves.

[0063] After constructing the incident particle, sensor model, event trigger function, event end function, step start function, step end function, etc., and running the .mac file under the current conditions, you can obtain the energy loss step length data txt file, range count curve, energy loss step length curve, simulation conditions txt file, and simulation log txt file after multiple incident events.

[0064] 3) Run the Weightfield2 software and configure it to input the same operating parameters as GEANT4.

[0065] The GEANT4 simulation log file is monitored in real time using a top-level script file. Once the simulation is detected as complete, the Weightfield2 program is launched, using the simulation conditions (txt) input in the previous step as simulation parameters (including energy, direction, step size, and type). Simultaneously, the Weightfield2 condition space itself needs to be generated, including temperature, bias voltage, and magnetic field. Three temperature conditions are selected: 273K, 300K, and 327K. The bias voltage is generally set according to the sensor characteristics; for example, the recommended operating voltage for an Autech ion implantation PN junction detector is 100V, with 60V and 80V set as condition spaces. The magnetic field is generally not considered.

[0066] 4) Run the Weightfield2 simulation and output the detector time-domain waveform data.

[0067] The top-level script sequentially initiates the detector construction, built-in potential field and weight field calculations, and current calculations for Weightfield2. It then starts batch processing mode, simulating the input energy loss step size data sequentially and outputting a txt file containing all current time-domain waveforms under the current operating condition (using the default sampling rate of 40GS / s). After the simulation of a certain operating condition is completed, the program starts the current simulation for the next operating condition (e.g., with temperature changes) until completion. The final current time-domain waveforms include hole-induced time-domain current waveforms, electron-induced time-domain current waveforms, and total induced time-domain current waveforms.

[0068] 5) Construct a preamplifier, main amplifier, and shaping circuit using LTSpice.

[0069] The prerequisite for using PyLTSpice for automated simulation is to construct a corresponding circuit diagram based on the actual circuit architecture. The preamplifier mainly includes two outputs: a charge-sensitive preamplifier section and a current-sensitive preamplifier section. The main components include a front-end input pulse waveform current source, an operational amplifier, feedback resistors and capacitors, shaping resistors and capacitors. The main amplifier includes an operational amplifier and amplification ratio resistors. After the circuit is configured, export the .cir file as the input for PyLTSpice.

[0070] 6) Run PyLTSpice to output the final dual-channel energy and differential current data.

[0071] First, the back-end circuit parameter traversal space is set up. Key parameters include the feedback resistor and capacitor of the preamplifier, as well as the shaping resistor and capacitor. A single current pulse waveform file is loaded into the pulse waveform current source, and the corresponding energy and differential current data are output. After all simulation outputs for a certain set of parameters are completed, the simulation is automatically re-run with new parameters, repeating the above steps until the simulation ends. All output data files are named after the current full-process simulation input parameters, including particle type, energy, incident direction, step size, temperature, bias voltage, front and back ends, preamplifier feedback resistor and capacitor, shaping resistor and capacitor, and main amplifier proportional resistor.

[0072] 2. Dual-channel pulse waveform automated acquisition system

[0073] The system utilizes physical entities for automated acquisition and storage of accelerator measurement data, as well as multi-channel fast-viewing.

[0074] Physical entity dual-channel pulse waveform acquisition and storage architecture design:

[0075] The dual-channel pulse waveform automated acquisition system includes a Si-PN junction ion-implanted semiconductor sensor, a low-noise energy output and differential current dual-channel preamplifier, a low-noise main amplifier, a high-speed acquisition board, and a multi-channel fast-view and fully automated data storage host computer. In actual experiments, particles from the calibration source first enter the Si-PN junction ion-implanted semiconductor sensor at the front end, generating a corresponding pulse current signal. The semiconductor sensor used is the CU-013-150-300 from Aotec. The current signal is amplified by the preamplifier to output both the energy and current signals. The preamplifier used is the Model 2003BT, a low-noise dual-channel preamplifier from Canberra. Due to the relatively small current signal, the main amplifier further amplifies it. The main amplifier used is the 9302 from Aotec, with a magnification factor of 20x. The final output signal is acquired by the high-speed acquisition board, and data is stored simultaneously during waveform observation in the accompanying fast-view system. The high-speed acquisition board uses a 1GS / s sampling rate board from Zhongke Caixiang.

[0076] The automated data preprocessing functions of the dual-channel pulse waveform automated acquisition system include:

[0077] Each waveform obtained by rapid sampling is preprocessed to obtain the peak-to-peak value and a peak-to-peak value distribution histogram is plotted. Due to the fluctuation of particle energy and the influence of noise, the peak-to-peak value will also be presented in the form of spectrum. Based on the distribution of the peak-to-peak value histogram, the interval with the maximum peak-to-peak value count is first obtained, and then a counting interval is expanded outward from this to both sides as valid data. Low-quality waveforms obtained due to noise or abnormal sampling are discarded. In order to further reduce the amount of data and improve the quality of the database, the position of the maximum value of the waveform is first found, and C points to the left and D points to the right are extracted as key information. If insufficient, the data is filled according to the last value and a new Excel file is generated, where each row is a waveform data. In addition, in order to more easily feed this data into the subsequent deep learning model architecture for particle identification, a normalized Excel file will also be output.

[0078] 3. Digital twin and physical twin data comparison and analysis system

[0079] The system uses the measured waveform of the accelerator obtained by the dual-channel pulse waveform automated acquisition system and the simulated waveform under the same operating conditions and circuit parameters obtained by the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system to perform waveform matching degree analysis. The analysis methods include comparing the rise slope of energy information, the pulse width of differential current, the mean square error, and wavelet transformation.

[0080] 1) Data import into the data analysis system

[0081] The system pre-stores energy and current waveform data of various operating conditions measured by the accelerator. After completing the digital simulation of a certain operating condition, it directly inputs the energy waveform, current waveform and simulation parameters output by the digital twin into the data analysis system and compares and analyzes them with the measured waveforms.

[0082] 2) Comparative analysis of simulation data and measured data

[0083] The main purpose of analyzing simulation data and measured data is to analyze the degree of matching between the two waveforms, and the data will be analyzed from several dimensions.

[0084] The first step is to perform wavelet transform on the simulated and measured energy and current data respectively, determine the number of decomposition layers, and generate denoised waveform data to reduce the impact of noise on waveform matching analysis.

[0085] The second step involves performing a matching analysis on the energy waveform data. This is done by calculating the slope of the rising edge of the energy output waveform at different times for the two modes, such as the slope at 10%, 30%, 50%, and 90% of the waveform height. The squared difference between these four data points is then used to determine the discrepancies. Additionally, the peak-to-peak difference is calculated to assess the bias in the absolute energy information.

[0086] The third step is to perform matching analysis on the current waveform data. Using the concept of cross-entropy in the deep learning classification model, the waveform data of the two modes are first normalized by softmax and the cross-entropy is calculated to determine the difference in shape. In addition, the difference in peak-to-peak value of the current is calculated to measure the overall consistency of the current intensity.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of this application do not depart from the spirit and scope of the technical solutions of this application, and should all be covered within the scope of the claims of this application.

Claims

1. A semiconductor detector particle pulse waveform data generation and analysis platform based on digital twin technology, characterized in that, The platform comprises: an automated full-link multi-condition semiconductor detector particle pulse waveform simulation system for controlling GEANT4, Weightfield2 and PyLTSPICE software, outputting simulated energy data and differential current data; a dual-channel pulse waveform automated acquisition system for collecting and storing accelerator measured energy data and current data using physical entities and displaying the data in multiple channels; and a digital twin and physical twin data comparison and analysis system for performing waveform matching degree analysis on accelerator measured waveforms obtained by the dual-channel pulse waveform automated acquisition system and simulated waveforms under the same condition and circuit parameters obtained by the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system, the analysis process including comparing energy information rising edge slope, differential current pulse width size, mean square error and wavelet change; the dual-channel pulse waveform automated acquisition system comprises: a Si-PN junction ion implantation type semiconductor sensor for inputting particles of a calibration source to generate corresponding pulse current signals; a preamplifier for amplifying the pulse current signals to output energy signals and current signals; a main amplifier for amplifying the energy signals and current signals output by the preamplifier again; a high-speed acquisition board card for collecting the energy signals and current signals output by the main amplifier; the data processing process of the dual-channel pulse waveform automated acquisition system comprises: obtaining peak-to-peak values of the waveforms of each energy signal and current signal collected by the high-speed acquisition board card through preprocessing and drawing a peak-to-peak value distribution column chart; obtaining the interval of the maximum count of the peak-to-peak values according to the distribution of the peak-to-peak value column chart, and regarding the interval as effective data by extending one count interval to each side, and eliminating low-quality waveforms obtained due to noise or abnormal collection; finding the maximum value position of the waveforms and extracting a plurality of points to the left and right of the maximum value position as output information, and filling in the insufficient values according to the tail values; the analysis process of the digital twin and physical twin data comparison and analysis system comprises: performing wavelet transform on the simulated and measured energy and current data to generate denoised waveform data after determining the decomposition level; performing matching analysis on the energy waveform data by calculating the different time discrete slopes of the rising edges of the energy output waveforms of the two modes, and calculating the difference square of the discrete slopes for judging the difference; calculating the peak-to-peak value difference for evaluating the deviation of the absolute energy information; performing matching analysis on the current waveform data by performing softmax normalization on the two modes of waveform data and calculating the cross-entropy for judging the difference in shape; calculating the current peak-to-peak value difference for measuring the overall consistency of the current intensity.

2. The digital twin technology based semiconductor detector particle pulse waveform data generation and analysis platform of claim 1, wherein, the processing process of the automated full-link multi-condition semiconductor detector particle pulse waveform simulation system comprises: running the GEANT4 software to simulate and output multiple incident energy loss data and draw a curve; running the Weightfield2 software to simulate and output detector time domain waveform data according to the incident energy loss data; The final bi-channel energy and differential current data are simulated from the time-domain waveform data by running the PyLTSpice software.

Images