Digital twinning-based RGA control system and control method thereof

Abstract

The invention discloses an RGA control system based on digital twinning and a control method thereof, and relates to the technical field of mass spectrometers, the RGA control system comprises a terminal layer, an edge layer and a cloud layer, the terminal layer comprises an enhanced sensor array and a function execution module; the enhanced sensor array is used for collecting working state data; the edge layer is used for sequentially performing disturbance prediction and feed-forward compensation by using a lightweight digital twin model and an adaptive feed-forward algorithm based on the working state data and the current control parameters to obtain an optimization control instruction; the function execution module is used for acquiring an ion current signal by utilizing the optimization control instruction; and the cloud layer is used for finely adjusting key parameters of the high-fidelity digital twinborn model through an online learning algorithm according to the ion current signal to obtain an RGA control result. According to the method, the technical upgrade from traditional passive feedback control to active feed-forward compensation is realized, and meanwhile, the breakthrough improvement of the performance of the RGA instrument is realized from multiple dimensions by relying on the continuous self-learning capability of the model and a hardware-software compensation strategy.

Description

Technical Field

[0001] This invention relates to the field of mass spectrometry technology, and in particular to an RGA control system and control method based on digital twins. Background Technology

[0002] The Residual Gas Analyzer (RGA), as a core device for gas composition analysis in a vacuum environment, converts gas molecules into charged ions through an ionization source. These ions are then separated by a quadrupole electric field and captured by a detector to achieve gas component concentration analysis. Its performance directly determines the accuracy, stability, and reliability of gas detection. In existing technologies, some RGA control systems employ a dual microcontroller unit (MCU) chip architecture. The hardware portion includes radio frequency (RF) / direct current (DC) power supply circuits, ion source and detector power supply circuits, signal acquisition and processing circuits, and vacuum and system monitoring circuits, responsible for ion generation, screening, detection, and environmental status monitoring. The control portion relies on an embedded processor to execute feedback control algorithms, completing core functions such as mass scanning, ion current signal processing, and equipment parameter calibration. This forms the basic architecture that ensures the basic operation of the RGA.

[0003] With the advancement of industrial environments and the diversification of scientific research scenarios, the inherent limitations of traditional RGA control systems are becoming increasingly apparent, making it difficult to meet the demands for high-precision and intelligent applications in complex environments. Firstly, traditional systems employ passive feedback control logic, only able to correct detection deviations after environmental interference or aging of core components (filaments, electron multipliers, electrodes) causes errors. This leads to a significant decrease in system stability under non-ideal operating conditions, making it difficult to maintain laboratory-level detection accuracy. Secondly, the hardware parameters of traditional RGA control systems are mostly fixed configurations. Adapting to different application scenarios, such as the "high-speed mode" required for vacuum leak detection and the "ultra-high resolution" required for scientific analysis, is problematic. The system suffers from several drawbacks. First, it requires manual replacement of hardware modules or complex parameter adjustments, making the operation cumbersome and inefficient. Second, regular calibration and parameter tuning rely on expert users with specialized knowledge, which is difficult for ordinary users to perform independently. Furthermore, the aging process of core components lacks effective real-time monitoring and early warning mechanisms, making it impossible to predict their lifespan in advance. This can easily lead to unplanned downtime, increasing maintenance costs and production / research risks. Third, the various hardware modules of the system mostly operate independently, with data interaction limited to basic control commands and status feedback. There is a lack of real-time data fusion and global collaborative decision-making across modules, making it impossible to dynamically optimize the parameters of each component based on the overall operating status. This results in the system's overall performance being unable to overcome bottlenecks.

[0004] While some existing RGA products have attempted to improve reliability and compatibility through modular design, digital control, and standardized interfaces, they have failed to fundamentally solve core problems such as "passive response to interference," "complex operation and maintenance," and "high maintenance costs." Digital twin technology, as a key technology for achieving real-time mapping, data synchronization, and collaborative optimization between physical entities and virtual models, has demonstrated powerful predictive maintenance, interference prediction, and proactive control capabilities in fields such as intelligent manufacturing and industrial control. However, it has not yet been effectively integrated into RGA control systems, making it impossible to predict changes in the operating status of the physical RGA control system through the virtual model and implement control compensation in advance.

[0005] Therefore, there is an urgent need for an RGA control system and its control method based on digital twins to address the shortcomings of existing technologies. Summary of the Invention

[0006] The purpose of this invention is to propose an RGA control system and its control method based on digital twins, in order to solve the problems of traditional RGA control systems, such as lack of active interference compensation, low level of intelligence, insufficient stability, low scene adaptation efficiency, and high operation and maintenance costs. By integrating digital twin technology with a dual-chip heterogeneous architecture, predictive compensation for environmental interference and device aging is achieved, promoting the system to upgrade from "passive feedback" to "active feedforward", thereby significantly improving the system's measurement accuracy, stability and automation level, effectively reducing manual maintenance costs, and comprehensively improving the system's intelligence level and overall performance.

[0007] On the one hand, to achieve the above objectives, the present invention provides an RGA control system based on digital twins, comprising: a terminal layer, an edge layer, and a cloud layer, wherein the terminal layer includes an enhanced sensor array and a function execution module;

[0008] The enhanced sensor array is used to collect operating status data;

[0009] The edge layer is used to perform disturbance prediction and feedforward compensation sequentially using a lightweight digital twin model and an adaptive feedforward algorithm based on the working status data and current control parameters, in order to obtain optimized control commands.

[0010] The function execution module is used to acquire ion flow signals using the optimized control instructions;

[0011] The cloud layer is used to fine-tune the key parameters of the high-fidelity digital twin model based on the ion flow signal using an online learning algorithm, and obtain the RGA control results.

[0012] Optionally, the function execution module includes a reconfigurable radio frequency drive module and an intelligent signal chain system-on-a-chip. The reconfigurable radio frequency drive module includes a dual microcontroller chip, a direct digital frequency synthesis signal generator, and a digital power amplifier. The dual microcontroller chip includes a real-time control chip and an intelligent computing chip.

[0013] Optionally, the enhanced sensor array includes an ion source, a quadrupole, a detector, a distributed high-precision temperature sensor, a micro-vibration sensor, and an internal pressure gradient sensor.

[0014] Optionally, the edge layer employs a multi-core ARM processor or a heterogeneous system-on-a-chip.

[0015] Optionally, the high-fidelity digital twin model includes a physical sub-model and a degradation and interference sub-model.

[0016] On the other hand, to achieve the above objectives, the present invention provides a control method for an RGA control system based on digital twins, comprising:

[0017] S1. Collect working status data;

[0018] S2. Based on the working status data and current control parameters, use a lightweight digital twin model and an adaptive feedforward algorithm to perform disturbance prediction and feedforward compensation in sequence to obtain optimized control commands;

[0019] S3. Obtain the ion flow signal using the optimized control command;

[0020] S4. Based on the ion flow signal, fine-tune the key parameters of the high-fidelity digital twin model using an online learning algorithm to obtain the RGA control results.

[0021] Optionally, S2, based on the operating state data and current control parameters, disturbance prediction and feedforward compensation are performed sequentially using a lightweight digital twin model and an adaptive feedforward algorithm to obtain optimized control commands, including:

[0022] The operating status data and current control parameters are input into a lightweight digital twin model for disturbance prediction, and the predicted spectrum deviation is obtained.

[0023] Based on the predicted spectrum deviation, an adaptive feedforward algorithm is used to perform reverse calculation to obtain the fine-tuning amount of the control parameters.

[0024] Based on the fine-tuning amount of the control parameters and the current control parameters, feedforward compensation is performed to obtain optimized control commands.

[0025] Optionally, S3, using the optimized control command, acquiring the ion flow signal includes:

[0026] The optimized control command is parsed to obtain the parsed optimized control command;

[0027] Based on the parsed optimized control instructions, an RF digital waveform and a DC digital level are generated, and a digital-to-analog conversion is performed to obtain an analog signal;

[0028] The analog signal is amplified to obtain a weak ion flow signal.

[0029] Mass scanning and signal acquisition operations are performed based on the weak ion flow signal to obtain the ion flow signal.

[0030] Optionally, S4, based on the ion flow signal, fine-tuning the key parameters of the high-fidelity digital twin model using an online learning algorithm to obtain the RGA control results, including:

[0031] The predicted spectrum is obtained using the predicted spectrum deviation;

[0032] Based on the ion flow signal and the predicted spectrum, the residual between the actual spectrum and the predicted spectrum is obtained;

[0033] The residual between the real spectrum and the predicted spectrum is input into the high-fidelity digital twin model, and the model parameters are fine-tuned through an online learning algorithm to obtain the fine-tuned model parameters.

[0034] Based on the fine-tuned model parameters and RGA running data, the RGA control results are obtained.

[0035] Optionally, the high-fidelity digital twin model includes a physical sub-model and a degradation and interference sub-model.

[0036] Compared with the closest existing technology, the present invention has the following advantages:

[0037] This invention addresses the technical problems of traditional RGA instruments, such as lack of active interference compensation, low intelligence, insufficient stability, low scene adaptation efficiency, and high operation and maintenance costs. It uses digital twin technology as its core, integrating technologies such as the Internet of Things, edge computing, and artificial intelligence. By constructing a lightweight digital twin model and a high-fidelity digital twin model containing physical sub-models, degradation sub-models, and interference sub-models, it establishes a "predictive-feedforward" active control mechanism. This is combined with reconfigurable hardware, online learning algorithms, and personalized software compensation strategies. Furthermore, the online learning system, which integrates reconfigurable hardware and multiple algorithms, achieves a technological upgrade from traditional passive feedback control to active feedforward compensation. Simultaneously, relying on the model's continuous self-learning capabilities and hardware-software compensation strategies, it achieves a breakthrough improvement in RGA instrument performance from multiple dimensions, as detailed below:

[0038] (1) The prediction-feedforward mechanism of lightweight model disturbance prediction and adaptive feedforward algorithm compensation is adopted to predict and cancel the disturbance before the environmental interference affects the equipment output, so that RGA can maintain laboratory-level stability in non-ideal environment. At the same time, relying on the dual control of digital twin real-time correction and feedforward compensation, the measurement accuracy is significantly improved.

[0039] (2) Relying on the characteristics of the high-fidelity digital twin model containing physical sub-model, degradation and interference sub-model, the instrument continuously learns online and automatically completes quality calibration, peak shape correction and sensitivity compensation, achieving “zero” manual calibration. The fully automatic calibration can be completed in just 2 minutes without the need for manual operation by professional personnel, which greatly improves the ease of operation of the instrument.

[0040] (3) By using a high-fidelity digital twin model to accurately simulate and predict the operating status of the core components of the equipment, the status of components such as filament life and electrode contamination can be perceived in advance, and maintenance warnings can be issued proactively to achieve predictive maintenance, thereby extending the maintenance interval by 5 times and effectively avoiding unplanned downtime. At the same time, by using software personalized modeling and precise compensation strategies, the consistency defects of hardware precision manufacturing can be made up for, the high-end manufacturing requirements of core components can be reduced, and the service life of core components can be extended through precise status control and maintenance warnings, thereby reducing the total cost of ownership of the instrument.

[0041] (4) Combining the reconfigurable hardware architecture and the ability to quickly switch between digital twin models, the system can be automatically optimized to high-speed mode or ultra-high resolution mode, adapting to multiple scenario requirements without changing the hardware, while greatly improving the response speed and significantly improving scenario adaptation efficiency and detection efficiency.

[0042] In summary, this invention solves the core technical problems of traditional RGA, transforming RGA from a "black box" instrument requiring expert debugging into an intelligent agent with self-awareness, continuous learning, and optimization capabilities. This comprehensively improves the instrument's measurement accuracy, operational stability, response speed, and intelligence level, while significantly reducing operational difficulty, scenario adaptation costs, and full lifecycle maintenance costs. It represents a breakthrough upgrade at the technical level and signifies the development direction of the next generation of intelligent analytical instruments. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the structure of an RGA control system based on digital twin according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart illustrating a control method for an RGA control system based on digital twins, according to an embodiment of the present invention.

[0046] Figure 3 This is a flowchart illustrating the architecture of the RGA control system proposed in an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0048] The terminology used in the embodiments section of this invention is for the purpose of explaining specific embodiments of the invention only, and is not intended to limit the invention.

[0049] like Figure 1 As shown, this embodiment of the invention provides an RGA control system based on digital twins, which adopts a three-layer heterogeneous hardware architecture of "cloud-edge-device", combines digital twin model and adaptive feedforward algorithm to realize full-dimensional state perception and active compensation control, including: terminal layer, edge layer and cloud layer, each layer works together to realize intelligent control of RGA, wherein the terminal layer (physical RGA unit) serves as the physical execution carrier of the system, and is enhanced on the basis of traditional RGA hardware, including enhanced sensor array and functional execution module;

[0050] The enhanced sensor array is used to collect operating status data;

[0051] The enhanced sensor array is the core of the RGA control system's data acquisition, bearing the crucial responsibility of capturing the equipment's operating status and environmental information from all dimensions. Building upon traditional RGA sensing components, it features functional expansion and precision upgrades. In addition to basic sensing units such as ion sources, quadrupoles, and detectors, it incorporates distributed high-precision temperature sensors (monitoring key circuit nodes and quadrupole cavity temperature), micro-vibration sensors (monitoring the impact of mechanical vibration on the quadrupole), and internal pressure gradient sensors (accurately reflecting ion transport efficiency). This enables real-time acquisition of multi-dimensional operating status data, including the temperature of key circuit nodes, the quadrupole cavity ambient temperature, mechanical vibration intensity, ion transport pressure gradient, filament current, emission current, and core component operating voltage. This comprehensive, accurate, and real-time data support provides essential data for subsequent edge-layer disturbance prediction and feedforward compensation, ensuring the scientific and targeted nature of control decisions.

[0052] The edge layer is used to perform disturbance prediction and feedforward compensation sequentially using a lightweight digital twin model and an adaptive feedforward algorithm based on the working status data and current control parameters, in order to obtain optimized control commands.

[0053] The edge layer (embedded intelligent gateway), as the core hub connecting terminal hardware and the cloud platform, is a crucial link in realizing local real-time intelligent control. The edge layer employs a high-performance multi-core ARM processor (such as the Cortex-A series processor) or a heterogeneous system-on-chip (SoC), running a lightweight digital twin engine to synchronize terminal layer status data in real time, execute adaptive feedforward control algorithms, generate optimized control commands and send them to the terminal layer, and simultaneously perform local data preprocessing and caching, as well as secure communication. Specifically, it performs noise reduction, compression, and feature extraction on the massive spectral data collected by the terminal layer, reducing the amount of data interaction with the cloud, and uses encrypted protocols to exchange encrypted data with the terminal and cloud layers, ensuring data transmission security. The heterogeneous SoC is a heterogeneous computing platform integrating an ARM processor and a Field-Programmable Gate Array (FPGA) architecture on the same chip. The ARM processor is responsible for running the operating system, executing control algorithms, and managing data, while the FPGA handles hardware acceleration and real-time signal processing. Their collaborative work can handle both complex logic processing and high real-time computing requirements. The core working logic of the edge layer revolves around "data processing - predictive analysis - instruction generation", specifically:

[0054] Run the lightweight digital twin engine: synchronize the working status data (temperature, vibration, voltage, current, etc.) collected by the terminal layer enhanced sensor array in real time, and synchronously obtain the current RGA control parameters, such as RF / DC voltage and scan mode settings; then call the built-in lightweight digital twin model to perform fast model prediction. This model can simulate the spectral deviations that may be caused by environmental disturbances (such as temperature transients and vibration disturbances) and device aging (such as filament decay and electrode contamination) by accurately mapping the characteristics of physical devices.

[0055] Execute the adaptive feedforward control algorithm: Based on the adaptive feedforward algorithm, the optimal fine-tuning amount of the control parameters is derived in reverse from the predicted output of the digital twin, and precise optimized control instructions are generated for the terminal layer hardware (such as RF module and ion source) to achieve advance compensation and precise control, and provide real-time guidance for the stable operation of the terminal layer.

[0056] The function execution module is used to acquire ion flow signals using the optimized control instructions;

[0057] The functional execution module is the physical execution core of the RGA control system. It includes a reconfigurable RF drive module and a smart signal chain SoC. It is used to receive optimized control commands from the edge layer, execute operations such as RF signal output, ion source driving, and signal acquisition, and finally obtain ion flow signals to provide real and reliable raw data for model optimization and result analysis in the cloud layer.

[0058] Reconfigurable RF Driver Module: Adopting a "dual MCU + Direct Digital Synthesis (DDS) signal generator + Digital Power Amplifier (DPA)" architecture, this module comprises dual microcontroller chips, a DDS signal generator, and a digital power amplifier. The dual MCUs and DDS signal generator chips directly generate pre-distortion-corrected RF digital waveforms and DC digital levels, which are then converted into analog signals via a high-speed digital-to-analog converter (DAC). The DPA replaces the traditional analog linear amplifier; its gain and response can be directly controlled via digital signals, supporting online switching of RF frequency (adjusting quality range) and waveform (sine wave, square wave optimized mode) to adapt to different application scenarios. The dual MCU chips include a real-time control chip and an intelligent computing chip, specifically:

[0059] The real-time control chip, acting as the "real-time nervous system," runs the FreeRTOS operating system and is responsible for handling high-priority, time-sensitive tasks such as RF scanning control (1ms cycle), ion source PID (proportional-integral-derivative) control (500µs cycle), signal acquisition direct memory access (DMA) tasks, safety monitoring tasks (10ms cycle), and data preprocessing tasks (5ms cycle). Its storage area is dedicated to real-time control parameters and signal data caching. The intelligent computing chip, acting as the "intelligent brain," runs relatively lower-priority tasks in FreeRTOS, including digital twin engine tasks, communication management tasks, advanced algorithm tasks, system monitoring tasks, and user interface tasks. It also uses a dedicated storage area to store the core digital twin model and algorithm data. The two work together through a high-speed data exchange mechanism. The real-time control chip also directly connects to the analog front-end and real-time control peripherals through a hardware interface, thereby achieving an efficient data flow closed loop of intelligent decision-making, real-time execution, and data feedback.

[0060] Intelligent signal chain SoC: Integrates ultra-low noise preamplifier, programmable filter and high-speed analog-to-digital converter (ADC). The filter parameters and gain can be dynamically adjusted in real time according to the current scan quality and expected signal size, improving the ability to detect weak signals.

[0061] The cloud layer is used to fine-tune the key parameters of the high-fidelity digital twin model based on the ion flow signal using an online learning algorithm, and obtain the RGA control results;

[0062] The cloud layer (digital twin and big data platform) serves as the global optimization and intelligent decision-making center of the RGA control system. It leverages a high-fidelity digital twin model and a big data platform to achieve in-depth data processing and model iteration. The high-fidelity digital twin model is a virtual RGA model based on the fusion of physical principles and machine learning, including physical sub-models and degradation and interference sub-models. By aggregating data from similar devices to form an optimization strategy knowledge base, it achieves global collaborative optimization and fault early warning. The core functions of the cloud layer include:

[0063] High-fidelity digital twin model: It integrates a physical sub-model that accurately simulates ion optics and quadrupole field stability, and a degradation and interference sub-model that simulates the effects of filament aging, electrode contamination, circuit temperature drift, vibration interference, etc. on the output spectrum through learning from historical data.

[0064] Model self-learning and optimization: The cloud layer receives the ion flow signal uploaded by the function execution module, compares it with the predicted spectrum, calculates the residual between the two, and fine-tunes the key parameters (such as aging factor and interference coefficient) of the high-fidelity digital twin model through online learning algorithms such as gradient descent, Bayesian update or recursive least squares method, so that the model continuously fits the actual operating state of the physical device and achieves self-evolution.

[0065] Global Collaboration and Health Management: Aggregates operational data from similar RGA devices to form an optimization strategy and fault mode knowledge base. This enables early warning of equipment performance degradation trends (such as cleaning cycle and filament life warnings) and global control parameter push. It can be used for global optimization and early fault warning of individual RGA devices. Finally, it generates RGA control results including optimal control parameters, equipment health status, and maintenance warning information. The optimized model parameters and control strategies are pushed to the edge layer to achieve global collaborative optimization and ensure long-term stable and efficient operation of the system.

[0066] In summary, the RGA control system acquires multi-dimensional operational status data in real time through an enhanced sensor array at the terminal layer. Disturbance prediction and feedforward compensation are achieved via a lightweight digital twin model and adaptive feedforward algorithm at the edge layer, generating optimized control commands. These commands are then executed by the functional execution module at the terminal layer to acquire precise ion flow signals. Finally, online learning algorithms at the cloud layer complete model parameter iteration and global optimization, forming a closed-loop control system encompassing data acquisition, real-time compensation, signal output, and model evolution. This system significantly reduces the impact of physical interference and device aging on measurement accuracy through local real-time compensation at the edge layer, while continuously improving model adaptability through global iterative optimization at the cloud layer. Ultimately, this results in a significant improvement in the measurement accuracy, anti-interference capability, and long-term stability of the RGA equipment, while also providing technical support for predictive maintenance and global collaborative optimization of the equipment.

[0067] Further reference Figure 2 This invention also provides a control method for an RGA control system based on a digital twin. The control method for an RGA control system based on a digital twin provided by this invention is described below. The control method for an RGA control system based on a digital twin described below can be referred to in correspondence with the RGA control system based on a digital twin described above. This control method includes:

[0068] S1. Collect working status data;

[0069] The enhanced sensor array at the terminal layer collects RGA operating status data in real time, including parameters such as temperature of key nodes, mechanical vibration parameters, internal pressure gradient, filament current, emission current, and RF / DC voltage. This operating status data is then uploaded to the edge layer via a high-speed communication interface. This step provides high-fidelity, multi-dimensional raw data support for subsequent disturbance prediction and control compensation, avoiding the limitations of traditional single-sensor data.

[0070] S2. Based on the working status data and current control parameters, use a lightweight digital twin model and an adaptive feedforward algorithm to perform disturbance prediction and feedforward compensation in sequence to obtain optimized control commands;

[0071] After receiving the operating status data and current control parameters, the edge layer uses a lightweight digital twin model for advanced simulation. This model simulates the impact of physical interference and device aging on the spectrum, predicting deviations such as mass peak shift and sensitivity fluctuations in advance. Subsequently, an adaptive feedforward algorithm is used to derive the optimal fine-tuning amount of the control parameters, generating optimized control commands, such as RF frequency fine-tuning values ​​and ion source voltage adjustments, and sending them to the function execution module of the terminal layer. This step enables early detection and proactive compensation of interference, avoiding the lag effect of interference on measurement results and significantly improving the anti-interference capability and measurement stability of the RGA.

[0072] S3. Obtain the ion flow signal using the optimized control command;

[0073] The terminal layer's functional execution module responds to optimized control commands issued by the edge layer, dynamically adjusting hardware configurations such as the RF drive waveform, ion source operating parameters, and detection circuit gain. It then precisely executes quality scanning and signal acquisition operations, including gas ionization, ion screening, and signal conversion, to acquire the ion flow signal, which is then uploaded to the edge and cloud layers. This step, by translating optimized commands into precise physical actions, ensures the accuracy and reliability of the ion flow signal, providing a realistic measurement basis for subsequent model iterations.

[0074] S4. Based on the ion flow signal, fine-tune the key parameters of the high-fidelity digital twin model using an online learning algorithm to obtain the RGA control results;

[0075] The cloud layer uses online learning algorithms (such as gradient descent and Bayesian update) to fine-tune key parameters of the digital twin model, such as aging factor and interference coefficient, based on the residuals between the real and predicted spectra. This ensures the model continuously aligns with the actual operating state of the physical equipment. Furthermore, it integrates global equipment operating data for collaborative optimization, generating RGA control results that include optimal control parameters, equipment health status, and maintenance warning information. The optimized model parameters are then pushed to the edge layer for iterative model evolution. This process achieves self-evolution and global optimization of the model, improving the long-term adaptability of the digital twin model and providing a basis for predictive maintenance and global collaborative optimization of the equipment.

[0076] like Figure 3The diagram shows the architecture flowchart of the RGA control system. The control method uses physical entities, virtual twins, and closed-loop optimization as its core logic, and the process is as follows: When the physical RGA unit (terminal layer) operates in a physically disturbed environment, its enhanced sensor array collects operating status data such as temperature, vibration, pressure, and electrical parameters, and uploads the data to the edge gateway. The edge gateway first uses a lightweight digital twin model, combined with the operating status data and current control parameters, to predict spectral deviations caused by physical disturbances and device aging. Then, it generates optimized control commands through adaptive feedforward control and sends them to the physical RGA unit. The physical RGA unit's function execution module responds to the optimized control commands, adjusts hardware parameters, and completes gas ionization and separation... The system filters and collects raw spectra (ion flow signals) and transmits them back to the edge gateway. The edge gateway then uploads the raw spectra to a cloud-based high-fidelity digital twin and big data platform. The platform compares the real and predicted spectra through model optimization and knowledge mining, calculates the residuals, and fine-tunes the key parameters of the high-fidelity digital twin model using online learning algorithms. This ensures real-time synchronization between the model and the actual state of the physical RGA unit. Simultaneously, it generates RGA control results containing optimal control parameters, equipment health status, and maintenance warning information, and pushes the optimized model parameters back to the edge gateway. This completes a fully closed-loop intelligent control process encompassing data acquisition, disturbance prediction, command compensation, signal feedback, and model iteration. The overall process satisfies the real-time control requirements of the edge terminal while achieving long-term performance optimization through cloud-based model iteration, effectively improving the RGA system's measurement stability, control accuracy, and lifecycle adaptability in complex interference environments.

[0077] As one possible implementation, in the above embodiments, step S2 may specifically include the following steps:

[0078] S2-1. Input the working status data and current control parameters into the lightweight digital twin model to perform disturbance prediction and obtain the prediction spectrum deviation;

[0079] By packaging real-time operating status data of the device collected by sensors together with control parameters issued by the current system (such as RF / DC voltage, scan settings, etc.), the data is synchronized at high speed to a lightweight digital twin model on the edge gateway, which has undergone parameter dimensionality reduction and redundancy detail simplification. The model then immediately "preemptively runs" the next scan point in virtual space. Based on the mathematical mapping relationship of the built-in core functional modules and the quantification of environmental interference and device aging impact factors, the model quickly simulates the effect mechanism of various disturbance factors on the measurement process of the RGA device under the current operating conditions. It predicts the spectral deviations such as mass peak shift and sensitivity changes that will occur under the current environmental interference (such as temperature transients) and device status (such as a slight decrease in filament emission efficiency).

[0080] The lightweight digital twin model is a core algorithm model deployed at the edge layer. Its core design goal is to achieve rapid simulation and disturbance prediction of the operating status of physical RGA devices under the constraints of limited edge hardware computing power and storage resources. This model simplifies and reduces the parameters of the physical device's three-dimensional structure, working principle, and multi-field coupling characteristics, eliminating redundant physical details that have no significant impact on control decisions. It retains only the mathematical mapping relationships of core functional modules such as ion source discharge, quadrupole ion screening, and detector signal conversion. Simultaneously, it integrates quantitative influence factors of environmental interference (temperature, vibration) and device aging (filament decay, electrode contamination), enabling rapid reading of operating status data and current control parameters uploaded by the enhanced sensor array. It can perform advanced predictions of measurement anomalies such as spectral deviation and sensitivity fluctuations within milliseconds, providing accurate disturbance analysis basis for the adaptive feedforward algorithm. The model's technical advantage lies in balancing computational efficiency and prediction accuracy, meeting the real-time control requirements of the edge layer while avoiding prediction distortion caused by over-simplification, making it a key support for achieving local closed-loop compensation.

[0081] S2-2. Based on the predicted spectrum deviation, the adaptive feedforward algorithm is used to perform reverse calculation to obtain the control parameter fine-tuning amount;

[0082] Using the predicted spectral deviation output by the lightweight model as the input variable, an adaptive feedforward algorithm is invoked to perform back-calculation, calculating the real-time fine-tuning of control parameters required. For example, fine-tuning the RF frequency to compensate for mass shift caused by temperature drift, and fine-tuning the ion source electron energy to compensate for filament aging. The algorithm incorporates a correlation model between spectral deviation and control parameters, enabling it to deduce the direction and magnitude of control parameter adjustments to offset the deviation based on the input predicted spectral deviation value. Unlike traditional feedback regulation, this back-calculation process does not wait for the actual deviation to occur; instead, it generates compensation strategies in advance based on the predicted data. Furthermore, the algorithm dynamically optimizes the calculation weights according to changes in equipment operating conditions, ensuring that the output control parameter fine-tuning is both targeted and meets the equipment's safe operating thresholds.

[0083] S2-3. Perform feedforward compensation based on the fine-tuning amount of the control parameters and the current control parameters to obtain optimized control instructions;

[0084] By superimposing the fine-tuning amount of the control parameters with the current control parameters, an optimized control command that can counteract predicted disturbances in advance is generated, such as "At time t, the applied RF voltage should be V". rf +ΔV”, where V rfThe reference RF voltage is ΔV, and the compensation RF voltage fine-tuning amount is ΔV. The application of the feedforward compensation mechanism can effectively avoid the lag problem of traditional feedback regulation. Before the interference factors affect the actual detection results, they can be offset by adjusting the control parameters. The final output optimized control command can be directly sent to the reconfigurable RF drive module and intelligent signal chain SoC of the terminal layer to ensure that the device can still output high-precision ion flow detection signals stably under complex operating conditions.

[0085] In summary, steps S2-1 to S2-3, leveraging the real-time edge computing capabilities of the lightweight digital twin model, predict environmental interference, device aging, and other disturbances before actual spectral deviations occur in the physical equipment. Then, through the reverse deduction of an adaptive feedforward algorithm, a targeted parameter fine-tuning scheme is generated, ultimately outputting optimized control commands that can offset the effects of disturbances. This process not only avoids the lag defects of traditional feedback control but also fully utilizes the efficiency of the lightweight model and the adaptability of the algorithm, effectively improving the measurement accuracy and operational stability of the RGA equipment under complex operating conditions. It also provides reliable commands and data support for subsequent parameter calibration of the high-fidelity model.

[0086] As one possible implementation, in the above embodiments, step S3 may specifically include the following steps:

[0087] S3-1. Parse the optimized control command to obtain the parsed optimized control command;

[0088] The dual MCU chip integrates two independent microcontroller cores, possessing parallel computing and fast instruction parsing capabilities. After receiving optimized control instructions generated by the front end, the chip first processes the parameter information in the instructions (such as the RF voltage should be V). rf The system performs format verification and protocol conversion on parameters such as +ΔV, DC bias voltage, and scanning frequency range. Then, according to the preset control logic, the composite instruction is decomposed into subdivided control signals that can be recognized by subsequent functional modules. At the same time, the instruction priority sorting and execution timing planning are completed, and finally the parsed optimized control instruction is output, providing a standardized control basis for subsequent waveform generation, power amplification and other stages.

[0089] S3-2. Based on the optimized control instructions after analysis, generate RF digital waveforms and DC digital levels, and perform digital-to-analog conversion to obtain analog signals;

[0090] The DDS signal generator, relying on the parameter configuration in the optimized control instructions after analysis, first calls the built-in pre-distortion correction algorithm to compensate for the nonlinear distortion that may occur in the RF signal during subsequent power amplification, generating an RF digital waveform that conforms to the target amplitude, frequency, and phase characteristics; simultaneously, it outputs a stable DC digital level according to the instruction requirements. Subsequently, the DDS signal generator converts the generated RF digital waveform and DC digital level into a continuous analog electrical signal via a DAC. This analog signal combines high precision and low noise characteristics, and can directly drive the subsequent power amplification module.

[0091] S3-3. The analog signal is amplified to obtain a weak ion flow signal.

[0092] After receiving the analog electrical signal output from the DDS signal generator, the digital power amplifier efficiently amplifies the signal based on its own digital drive architecture, providing sufficient power to drive the quadrupole mass analyzer of the RGA device. The amplified analog electrical signal is applied to the quadrupole electrodes, forming a stable composite field of radio frequency and DC electric fields. This composite field filters ions entering the analyzer, allowing only ions that meet the target mass-to-charge ratio to pass through and collide with the detector. The detector converts the ion collision signal into a weak current signal, i.e., a weak ion current signal, providing raw data for subsequent signal acquisition.

[0093] S3-4. Perform mass scanning and signal acquisition operations based on the weak ion flow signal to obtain the ion flow signal;

[0094] The intelligent signal chain SoC integrates preamplifier, filter, analog-to-digital converter, and data processing units. The chip first performs low-noise preamplification and baseline noise filtering on the weak input ion flow signal. Then, according to a preset mass scan sequence, it acquires and quantizes ion signals corresponding to different mass-to-charge ratios point by point. Simultaneously, through a built-in signal calibration algorithm, it eliminates the influence of environmental interference and device drift on the signal, ultimately outputting an ion flow signal with high signal-to-noise ratio and high resolution. This signal can be directly used for subsequent parameter fine-tuning of high-fidelity digital twin models and measurement result output from RGA equipment.

[0095] In summary, steps S3-1 to S3-4, relying on the coordinated operation of multiple modules, achieve high efficiency in instruction parsing, accurate signal generation, stable power amplification, and high sensitivity in signal acquisition. This effectively avoids problems such as signal distortion and noise interference, significantly improves the accuracy of ion screening and the reliability of signal detection in the RGA equipment, and provides high-quality data support for the subsequent parameter calibration of the high-fidelity digital twin model.

[0096] As one possible implementation, in the above embodiments, step S4 may specifically include the following steps:

[0097] S4-1. Obtain the predicted spectrum using the predicted spectrum deviation;

[0098] Based on the predicted spectrum deviation output by the lightweight digital twin model, and combined with the theoretical standard spectrum corresponding to the current control parameters, the predicted spectrum is obtained through difference calculation. Specifically, by subtracting the predicted spectrum deviation value at the corresponding position from the ion current intensity corresponding to each mass-to-charge ratio in the theoretical standard spectrum, the expected spectrum shape of the RGA equipment output under the influence of disturbances such as current environmental interference and device aging can be obtained. The generated predicted spectrum fully retains the influence trend of disturbance factors on key features such as peak position, peak height, and full width at half maximum (FWHM), providing an accurate reference benchmark for subsequent comparison with the actual spectrum.

[0099] S4-2. Based on the ion current signal and the predicted spectrum, obtain the residual between the actual spectrum and the predicted spectrum;

[0100] The ion current signal acquired by the intelligent signal chain system-on-a-chip undergoes data conversion and preprocessing to generate a true spectrum with dimensions consistent with the predicted spectrum, ensuring complete alignment of the mass-to-charge ratio range and ion current intensity units. Subsequently, using point-by-point comparison or characteristic peak comparison methods, the difference in ion current intensity at the same mass-to-charge ratio position between the true and predicted spectra is calculated, forming a residual distribution curve. Simultaneously, statistical analysis is performed on the residual data to obtain quantitative indicators such as average residual and maximum residual. This residual not only intuitively reflects the degree of deviation between the lightweight model's prediction results and the actual operating state of the equipment, but also provides precise error feedback for subsequent parameter fine-tuning of the high-fidelity model's twin module.

[0101] S4-3. Input the residual between the real spectrum and the predicted spectrum into the high-fidelity digital twin model and fine-tune the model parameters through an online learning algorithm to obtain the fine-tuned model parameters;

[0102] The high-fidelity digital twin model comprises a physical sub-model and a degradation and interference sub-model, which work together to support the parameter fine-tuning process. First, the residual data of the real and predicted spectra are uploaded to the cloud-based high-fidelity digital twin model. Online learning algorithms (such as gradient descent, Bayesian update, or recursive least squares) decompose the error sources based on residual characteristics and fine-tune the twin model parameters accordingly: for the physical sub-model, simulation parameters of core physical processes such as ion source excitation, quadrupole screening, and detector response are corrected to improve the model's accuracy in replicating the basic operating characteristics of the equipment; for the degradation and interference sub-model, aging factors (such as electrode wear coefficients and component fatigue parameters) and interference coefficients (such as ambient temperature and electromagnetic interference influence weights) are optimized to enhance the model's adaptability to complex disturbances. Through multiple iterations, the residual between the model's simulation output and the real spectrum is reduced to a reasonable range, ultimately obtaining the fine-tuned model parameters. This allows the simulation characteristics of the high-fidelity model to closely match the actual operating state of the physical equipment, achieving self-evolution and precise iteration of the model.

[0103] A high-fidelity digital twin model is a full-dimensional device mapping model deployed in the cloud layer. Its core feature is to maximize the reproduction of the complete attributes and dynamic behavior of the physical RGA device, providing a high-precision virtual simulation platform for global optimization and long-term iteration. This model is built upon the physical device's design drawings, performance parameters, and full lifecycle operational data. It not only includes the core functional modules consistent with the lightweight model but also integrates fine physical details such as the device's material properties, processing errors, multi-component collaborative mechanisms, and nonlinear coupling effects under complex environments. It also supports the fusion analysis of massive historical data and real-time transmitted ion flow signals, continuously iterating model parameters through online learning algorithms. In practical applications, the model compares the real ion flow signals fed back from the edge layer with the prediction results of the lightweight model, calculates the residuals, and reverses its own key parameters such as interference factors and aging coefficients, achieving dynamic calibration between the digital twin and the physical entity. Furthermore, it can be used for in-depth applications such as control strategy optimization, fault simulation, and equipment life prediction based on the high-fidelity simulation environment. The technical value of this model lies in overcoming the resource limitations of the edge layer, providing a high-precision and scalable virtual experimental platform for global optimization, predictive maintenance, and large-scale management of the RGA system.

[0104] S4-4. Based on the fine-tuned model parameters and RGA running data, obtain the RGA control results;

[0105] By fusing and analyzing the finely tuned parameters of the high-fidelity model with historical operating data and condition data of similar RGA devices, a control strategy library for multi-device collaborative optimization is constructed. Based on this strategy library, on the one hand, relying on the accurate simulation capabilities of the physical sub-model, the optimal control parameters for the current RGA device, such as RF frequency, DC bias, and scan speed, are output to guide the device to achieve high-precision control; on the other hand, through the iterative parameters of the degradation and disturbance sub-model, the aging trend and potential failure risks of the device are predicted, and a health assessment report is generated; at the same time, universally applicable disturbance compensation rules and parameter configuration schemes are extracted to provide a reference for the large-scale operation and maintenance of similar RGA devices. The cloud platform, relying on this strategy library, further analyzes the operating data of a single device and similar devices, accurately identifies performance degradation trends (such as decreased filament emission efficiency and electron multiplier gain drift), proactively issues maintenance warnings, and periodically pushes globally optimized model parameters or control strategies to the edge layer to achieve continuous device optimization. The final RGA control results balance the accuracy of a single device with the universality of multiple devices, achieving global optimization of measurement accuracy, operational stability, and maintenance efficiency.

[0106] In summary, steps S4-1 to S4-4, relying on the twin architecture and online iteration capability of the high-fidelity digital twin model, achieve accurate replication of the physical characteristics and complex disturbance factors of the RGA equipment, effectively reducing the residual between the real spectrum and the predicted spectrum. This not only improves the measurement accuracy and operational stability of a single device, but also extracts a universal control strategy by integrating data from similar devices, providing reliable support for the large-scale precise control and intelligent operation and maintenance of RGA equipment.

[0107] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A digital twin-based RGA control system, characterized in that, include: The system comprises a terminal layer, an edge layer, and a cloud layer, wherein the terminal layer includes an enhanced sensor array and a function execution module; The enhanced sensor array is used to collect operating status data; The edge layer is used to perform disturbance prediction and feedforward compensation sequentially using a lightweight digital twin model and an adaptive feedforward algorithm based on the working status data and current control parameters, in order to obtain optimized control commands. The function execution module is used to acquire ion flow signals using the optimized control instructions; The cloud layer is used to fine-tune the key parameters of the high-fidelity digital twin model based on the ion flow signal using an online learning algorithm, and obtain the RGA control results.

2. The RGA control system based on digital twin according to claim 1, characterized in that, The functional execution module includes a reconfigurable radio frequency drive module and an intelligent signal chain system-on-a-chip. The reconfigurable radio frequency drive module includes a dual microcontroller chip, a direct digital frequency synthesis signal generator, and a digital power amplifier. The dual microcontroller chip includes a real-time control chip and an intelligent computing chip.

3. The RGA control system based on digital twin according to claim 1, characterized in that, The enhanced sensor array includes an ion source, a quadrupole, a detector, a distributed high-precision temperature sensor, a micro-vibration sensor, and an internal pressure gradient sensor.

4. The RGA control system based on digital twin according to claim 1, characterized in that, The edge layer uses a multi-core ARM processor or a heterogeneous system-on-a-chip.

5. The RGA control system based on digital twin according to claim 1, characterized in that, The high-fidelity digital twin model includes a physical sub-model and a degradation and interference sub-model.

6. A control method for an RGA control system based on digital twins, employing the RGA control system based on digital twins as described in any one of claims 1-5, characterized in that, include: S1. Collect working status data; S2. Based on the working status data and current control parameters, use a lightweight digital twin model and an adaptive feedforward algorithm to perform disturbance prediction and feedforward compensation in sequence to obtain optimized control commands; S3. Obtain the ion flow signal using the optimized control command; S4. Based on the ion flow signal, fine-tune the key parameters of the high-fidelity digital twin model using an online learning algorithm to obtain the RGA control results.

7. The control method for an RGA control system based on digital twins according to claim 6, characterized in that, S2. Based on the operating status data and current control parameters, disturbance prediction and feedforward compensation are performed sequentially using a lightweight digital twin model and an adaptive feedforward algorithm to obtain optimized control commands, including: The operating status data and current control parameters are input into a lightweight digital twin model for disturbance prediction, and the predicted spectrum deviation is obtained. Based on the predicted spectrum deviation, an adaptive feedforward algorithm is used to perform reverse calculation to obtain the fine-tuning amount of the control parameters. Based on the fine-tuning amount of the control parameters and the current control parameters, feedforward compensation is performed to obtain optimized control commands.

8. The control method for an RGA control system based on digital twins according to claim 6, characterized in that, S3. Using the optimized control command, acquire the ion flow signal, including: The optimized control command is parsed to obtain the parsed optimized control command; Based on the parsed optimized control instructions, an RF digital waveform and a DC digital level are generated, and a digital-to-analog conversion is performed to obtain an analog signal; The analog signal is amplified to obtain a weak ion flow signal. Mass scanning and signal acquisition operations are performed based on the weak ion flow signal to obtain the ion flow signal.

9. The control method for an RGA control system based on digital twins according to claim 7, characterized in that, S4. Based on the ion flow signal, fine-tune the key parameters of the high-fidelity digital twin model using an online learning algorithm to obtain the RGA control results, including: The predicted spectrum is obtained using the predicted spectrum deviation; Based on the ion flow signal and the predicted spectrum, the residual between the actual spectrum and the predicted spectrum is obtained; The residual between the real spectrum and the predicted spectrum is input into the high-fidelity digital twin model, and the model parameters are fine-tuned through an online learning algorithm to obtain the fine-tuned model parameters. Based on the fine-tuned model parameters and RGA running data, the RGA control results are obtained.

10. The control method for an RGA control system based on digital twins according to claim 9, characterized in that, The high-fidelity digital twin model includes a physical sub-model and a degradation and interference sub-model.

Images