A medical instrument circuit management control system and control method

Abstract

The application relates to the technical field of medical instruments, and particularly discloses a medical instrument circuit management control system and a control method, which comprises the following steps: a data acquisition module is used to acquire voltage data, current data and temperature data of a circuit, and to construct a real-time data set; a data processing module is used to perform space-time registration, noise filtering and feature extraction on the real-time data set, and to generate frequency feature data; a fault judgment module is used to analyze the frequency feature data through an LSTM network, to identify a resonance type and a fault position, and to generate circuit output information; and a control module is used to acquire capacitor matching information of a compensation module and load end output information according to the circuit output information. Through the collaborative operation of multiple modules, the data acquisition and processing module realizes space-time registration and noise filtering of multi-source data, ensures the accuracy of input data, and provides a reliable basis for fault diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical instruments, and in particular to a medical instrument circuit management control system and control method. BACKGROUND

[0002] In the field of medical equipment, such as the circuit systems of precision instruments such as CT, MRI, and linear accelerators, resonance phenomena are key problems affecting the stability and reliability of the equipment.

[0003] Existing technologies have certain limitations in resonance detection and suppression. Some methods rely on manual experience for parameter adjustment, which is slow in response and low in precision. Some monitoring systems based on sensors can only detect a single type of resonance and cannot cover a wide frequency range, which is not conducive to long-term maintenance and performance optimization of the equipment. SUMMARY

[0004] The purpose of the present application is to provide a medical instrument circuit management control system and control method to solve the technical problems raised in the background.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] A medical instrument circuit management control system includes a compensation module and further includes:

[0007] A data acquisition module is configured to acquire voltage data, current data, and temperature data of the circuit and construct a real-time data set.

[0008] A data processing module is configured to perform time-space registration, noise filtering, and feature extraction on the real-time data set to generate frequency feature data.

[0009] A fault judgment module is configured to analyze the frequency feature data through an LSTM network, identify the resonance type and fault location, and generate circuit output information.

[0010] A control module is configured to obtain capacitor grouping information of the compensation module and load end output information according to the circuit output information.

[0011] Preferably, the data processing module includes:

[0012] A data matrix generation unit is configured to align the voltage data, current data, and temperature data through timestamps, compensate for sensor layout differences based on a spatial interpolation algorithm, and generate time-space synchronized voltage, current, and temperature matrices.

[0013] A data processing unit is configured to filter out high-frequency noise and retain fundamental frequency and low-order harmonic components through a threshold method on the time-space synchronized voltage, current, and temperature matrices.

[0014] A frequency identification unit is configured to extract frequency features from a time-space synchronized voltage, current and temperature matrix reserving a fundamental frequency and low-order harmonic components, obtain an FFT spectrum of a voltage signal, and identify a main resonance frequency and a sub-resonance frequency band.

[0015] A frequency output unit is configured to obtain a current waveform segment corresponding to the main resonance frequency and the sub-resonance frequency band output by the frequency identification unit, obtain an equivalent inductance and capacitance under the current waveform segment through current waveform integration, and generate a feature vector, and input the feature vector into a Kalman filter for dynamic correction to output optimized frequency feature data.

[0016] Preferably, the fault judgment module comprises:

[0017] A frequency input unit is configured to receive the frequency feature data and output resonance information labels and fault information labels corresponding to each resonance type and each fault level through a Softmax function;

[0018] A frequency training unit is configured to obtain a historical fault data set of a medical instrument, extract resonance type labels, fault level labels and corresponding frequency feature data for training from the historical fault data set, and construct a hybrid data training model through iterative training of the frequency feature data;

[0019] An information reasoning unit is configured to perform sliding window sampling on real-time collected to-be-tested frequency feature data based on the hybrid data training model, and output a fault probability matrix corresponding to each window through the LSTM network;

[0020] An information output unit is configured to obtain a located fault circuit node according to the fault probability matrix, and generate circuit output information containing a position coordinate and a confidence level based on the located fault circuit node.

[0021] Preferably, the obtaining of the historical fault data set of the medical instrument and the construction of the hybrid data training model based on the historical fault data set comprise the following steps:

[0022] Obtaining historical circuit data and fault type data based on the historical fault data set;

[0023] Obtaining a resonance frequency offset, a capacitance aging coefficient and an inductance temperature drift coefficient based on the historical circuit data;

[0024] Obtaining a fault type, a circuit component and an environmental parameter based on the fault type data;

[0025] Constructing a fault node-component relationship graph based on the fault type, the circuit component and the environmental parameter;

[0026] Map the resonance frequency offset, the capacitance aging coefficient, and the inductance temperature drift coefficient to a fault node-component relationship atlas to establish a hybrid data training model.

[0027] Preferably, the control module comprises:

[0028] The analysis unit is configured to analyze the circuit output information to obtain resonance suppression information.

[0029] The judgment unit is configured to determine the resonance type of the resonance suppression information, obtain a target capacitance value if the resonance type is series resonance, and generate a damping resistor activation instruction and a target resistance value if the resonance type is parallel resonance.

[0030] The adjustment unit is configured to generate a dynamic adjustment capacitance combination according to the resonance type to obtain capacitance combination information.

[0031] The compensation triggering unit is configured to obtain triggering of multi-stage damping units for collaborative compensation according to the capacitance combination information.

[0032] The supplementary parameter acquisition unit is configured to output a load end control signal, generate a driving signal through a PWM modulator, obtain control of IGBT switching frequency and duty cycle according to the driving signal, and obtain load voltage compensation parameters based on resonance suppression information.

[0033] Preferably, the obtaining triggering of multi-stage damping units for collaborative compensation according to the capacitance combination information comprises the following steps:

[0034] Establishing a multi-stage phase compensation model according to the capacitance combination information.

[0035] Generating a damping signal according to the multi-stage phase compensation model and performing phase lead compensation on the damping signal.

[0036] Superimposing the compensated signal to a current inner loop to generate a current inner loop reference value, and adjusting an output voltage based on the current inner loop reference value through a PID controller.

[0037] Preferably, the control module further comprises:

[0038] The receiving unit is configured to receive a resonance fault report of a fault judgment module, and obtain a current frequency and a resonance frequency range according to the resonance fault report.

[0039] The calculation unit is configured to calculate a target capacitance value required to eliminate resonance according to the current frequency and the resonance frequency range.

[0040] The capacitance value acquisition unit is configured to traverse adjustable capacitances of a series compensation device to obtain capacitance values of the adjustable capacitances.

[0041] The matching unit is configured to select, according to a capacitance value matching algorithm, a to-be-controlled adjustable capacitance group from the adjustable capacitances, the to-be-controlled adjustable capacitance group having a combined total capacitance value closest to a target capacitance value.

[0042] The control instruction generation unit is configured to generate a control instruction to control the control switches in series with the adjustable capacitances in the to-be-controlled adjustable capacitance group to be turned off.

[0043] The verification unit is configured to re-collect circuit electrical parameters after the control switches are turned off, and verify whether resonance is eliminated.

[0044] The application further provides a medical instrument circuit management control method, comprising:

[0045] Obtaining voltage data, current data and temperature data of the circuit to construct a real-time data set;

[0046] Performing space-time registration, noise filtering and feature extraction on the real-time data set to generate frequency feature data;

[0047] Analyzing the frequency feature data through an LSTM network to identify a resonance type and a fault position and generate circuit output information;

[0048] Obtaining capacitance group information of a compensation module and load end output information according to the circuit output information.

[0049] Preferably, the step of performing space-time registration, noise filtering and feature extraction on the real-time data set to generate frequency feature data comprises:

[0050] Aligning the voltage data, current data and temperature data through timestamps and compensating for sensor layout differences based on a spatial interpolation algorithm to generate a space-time synchronous voltage, current and temperature matrix;

[0051] Filtering high-frequency noise from the space-time synchronous voltage, current and temperature matrix through a threshold method and retaining fundamental frequency and low-order harmonic components;

[0052] Extracting frequency features from the space-time synchronous voltage, current and temperature matrix retaining the fundamental frequency and low-order harmonic components to obtain an FFT frequency spectrum of a voltage signal and identify a main resonance frequency and a secondary resonance frequency band;

[0053] The main resonance frequency and the current waveform segment corresponding to the secondary resonance frequency band output by the frequency recognition unit are acquired, the equivalent inductance and capacitance under the current waveform segment are acquired through current waveform integration, and a feature vector is generated, the feature vector is input into a Kalman filter for dynamic correction, and optimized frequency feature data is output.

[0054] Preferably, the step of analyzing the frequency feature data through the LSTM network, identifying the resonance type and the fault position, and generating circuit output information comprises:

[0055] The frequency feature data is received, and the resonance information label and the fault information label corresponding to each resonance type and each fault level are output through a Softmax function;

[0056] A historical fault data set of the medical instrument is acquired, resonance type labels, fault level labels and corresponding frequency feature data for training are extracted from the historical fault data set, and a hybrid data training model is constructed through iterative training of the frequency feature data;

[0057] Based on the hybrid data training model, the frequency feature data to be measured collected in real time is subjected to sliding window sampling, and a fault probability matrix corresponding to each window is output through the LSTM network;

[0058] According to the fault probability matrix, a fault circuit node is located, and circuit output information containing position coordinates and confidence is generated based on the located fault circuit node.

[0059] The beneficial effects of the present application are: the present application realizes multi-source data space-time registration and noise filtering through the cooperative operation of multiple modules, and the data acquisition and processing module ensures the accuracy of the input data and provides a reliable basis for fault diagnosis. The fault judgment module based on the LSTM network and the hybrid data training model realizes accurate identification of the resonance type and the fault position, compared with the traditional method, the fault detection time is greatly shortened, and the positioning error is improved from the circuit board level to the component level. The control module adjusts the capacitance group and the multi-stage damping compensation dynamically, effectively suppresses the resonance, controls the voltage fluctuation within a safe range, and improves the response speed. The closed-loop feedback mechanism of the verification unit ensures the effectiveness of the control strategy, and in combination with the database recording function, a complete fault handling and tracing system is formed. The maintenance cost and downtime risk are significantly reduced, and strong technical support is provided for the safe and stable operation of the medical equipment. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 The system structure schematic diagram of the medical instrument circuit management control system provided by an embodiment of the present application is shown.

[0061] Figure 2A method flowchart of a medical instrument circuit management control method provided by an embodiment of the present application is shown.

[0062] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0063] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0064] As shown in the accompanying drawings, Figure 1 The present application provides a medical instrument circuit management control system, which comprises a compensation module and further comprises:

[0065] A data acquisition module 1 is configured to acquire voltage data, current data and temperature data of the circuit and to construct a real-time data set;

[0066] A data processing module 2 is configured to perform time-space registration, noise filtering and feature extraction on the real-time data set and to generate frequency feature data;

[0067] A fault judgment module 3 is configured to analyze the frequency feature data through an LSTM network, identify the resonance type and the fault position, and generate circuit output information;

[0068] A control module 4 is configured to acquire the capacitance matching information of the compensation module and the load end output information according to the circuit output information.

[0069] Specifically, the compensation module and the control module 4 are connected, and the compensation module is not shown in the figure. In the intelligent monitoring system of the medical device, the data acquisition module integrates Hall voltage sensors, Rogowski coil current transformers, miniature infrared thermal imaging arrays and EMI detection antennas through multi-source sensor fusion technology to perform multi-dimensional monitoring on the circuit operation state. Through the IEEE1588 precision clock protocol, sub-microsecond timestamp alignment is achieved, and the sensor layout difference is compensated by the bicubic interpolation algorithm to construct a three-dimensional real-time data set containing time-domain waveforms, frequency-domain features and spatial temperature fields. In the monitoring of the circuit of the medical linear accelerator, the module can acquire the voltage fluctuation of the accelerating tube power supply, the pulse current rise time and the magnetron cavity temperature in real time;

[0070] The data processing module eliminates the response delay according to the multi-sensor sampling rate difference method and generates a time-space consistent data cube in combination with the Kalman filter. Wavelet decomposition is performed through the base function, the switch noise is filtered by the front-stage Butterworth low-pass filter, and the power frequency harmonic is suppressed by the rear-stage soft threshold method.

[0071] The fault judgment module uses a bidirectional LSTM (Long Short-Term Memory) network combined with an attention mechanism to perform time series modeling on the feature vector. The training set contains fault samples (such as resonance, capacitor aging, and IGBT (Insulated Gate Bipolar Transistor) breakdown), and a fault-component-environment association network is constructed through a knowledge graph to locate the fault point to the filter capacitor and output fault codes and location codes. The control module sends capacitor configuration instructions to the series compensation module through the SPI bus based on the fault diagnosis results. For example, when the CT power supply resonates is detected, the system quickly switches the compensation capacitor array to shift the resonance frequency to a safe interval. At the same time, the output voltage and current are dynamically adjusted according to the load end requirements. The CT power supply refers to the power supply system of the CT (Computed Tomography) device, specifically referring to the power module that provides stable power to the high-voltage generator, X-ray tube, and other key components of the CT scanner. In the DR device (Digital Radiography), when the high-voltage generator voltage fluctuation exceeds the limit is detected, the system automatically adjusts the IGBT drive signal duty cycle to stabilize the output voltage within a safe range. This ensures the stable operation of medical equipment under complex working conditions.

[0072] In one embodiment, the data processing module includes a data matrix generation unit for aligning voltage data, current data, and temperature data by timestamp, compensating for sensor layout differences based on a spatial interpolation algorithm, generating a spatiotemporal synchronous voltage, current, and temperature matrix, and ensuring strict synchronization of voltage, current, and temperature data during the energy storage capacitor charging process through timestamp alignment technology. When voltage fluctuations occur at the trailing edge of the matrix at a certain time, and the temperature corresponding to the spatial coordinates rises sharply, the driving circuit element can be accurately located, and it can be determined that poor heat dissipation may cause resonance risk. For example, in a digital X-ray photography system (DR), when processing switch voltage, output current, and transformer oil temperature data, the generated spatiotemporal matrix shows that the temperature gradient corresponding to the moment when the voltage harmonic suddenly increases exceeds the preset temperature, and potential resonance risks caused by capacitor aging are warned in advance to avoid fault detection caused by asynchronous data.

[0073] The data processing unit is used for filtering high-frequency noise of the space-time synchronized voltage, current and temperature matrix through a threshold method, and retaining fundamental frequency and low harmonic components. Specifically, the wavelet transform is used for noise filtering processing of the synchronized data, the high-frequency noise is filtered through the threshold method, and the fundamental frequency and low harmonic components are retained. After filtering the noise above the preset hertz through the wavelet transform, the fundamental frequency amplitude fluctuation and the 2nd harmonic energy proportion are analyzed. For example, when the 2nd harmonic proportion is suddenly increased, combined with the rise of the time domain waveform distortion degree (THD), the system judges that the switching loss of the power MOSFET is abnormal, and timely triggers the parameter adjustment to avoid the flow control error caused by the harmonic distortion. In the high-voltage generator circuit of the CT device, when processing the direct current superimposed ripple of the device, the fundamental frequency and harmonic components are retained, the slight drift of the main resonance frequency is accurately identified, and key features are provided for subsequent resonance suppression.

[0074] The frequency identification unit is used for extracting frequency characteristics from the space-time synchronized voltage, current and temperature matrix retaining the fundamental frequency and low harmonic components, obtaining the FFT spectrum of the voltage signal, and identifying the main resonance frequency and the secondary resonance frequency band.

[0075] The frequency output unit is used for obtaining the current waveform segment corresponding to the main resonance frequency and the secondary resonance frequency band output by the frequency identification unit, obtaining the equivalent inductance and capacitance under the current waveform segment through current waveform integration, and generating a feature vector. The feature vector is input into a Kalman filter for dynamic correction, and the optimized frequency characteristic data is output. By adjusting the phase compensation parameters of the driving circuit, the resonance amplitude is effectively suppressed, and the pulse waveform distortion caused by resonance is avoided. By inputting the feature vector into the Kalman filter for dynamic correction, the optimized frequency characteristic data is output.

[0076] In one embodiment, the fault judgment module comprises:

[0077] A frequency input unit is configured to receive the frequency characteristic data and output, through a Softmax function, a resonance information label corresponding to each resonance type and a fault information label corresponding to each fault level. The frequency characteristic data includes parameters such as a main resonance frequency, a secondary resonance frequency band, an equivalent inductance, an equivalent capacitance, and a total harmonic distortion (THD). The frequency input unit performs multi-classification processing on the input data by using the Softmax function, which is derived from a multi-classification algorithm in machine learning and can map a multi-dimensional feature vector to a probability distribution of each resonance type (such as series resonance, parallel resonance, and hybrid resonance) and each fault level (from level I to level IV). The Softmax function can generate a resonance information label corresponding to each resonance type and a fault information label corresponding to each fault level. For example, when the input feature vector is (36.2 kHz, 9.2 μF, 2.8 mH, 12), the Softmax function can output a series resonance probability of 0.92 and a fault level III probability of 0.88, clearly defining the fault type and severity. The function is to convert abstract frequency characteristics into interpretable fault semantics, providing a clear basis for subsequent decision-making. In the CT high-voltage generator circuit monitoring, after the frequency input unit receives data with a main resonance frequency deviation and an equivalent capacitance decrease by a preset value, the Softmax function quickly classifies the data and determines that the series resonance is caused by capacitor aging, providing an accurate fault type signal for the control module.

[0078] A frequency training unit is configured to obtain a historical fault data set of a medical instrument, extract resonance type labels, fault level labels, and corresponding frequency characteristic data for training from the historical fault data set, and construct a hybrid data training model by iteratively training the frequency characteristic data. The historical fault data set covers thousands of fault cases and includes characteristic parameters (such as resonance frequency deviation, capacitor aging coefficient, and inductance temperature drift coefficient) of different types of resonance (such as series resonance and parallel resonance) and corresponding fault types, circuit components, and environmental parameters. The unit uses a hybrid architecture combining LSTM (LSTM) and random forest. LSTM is used to capture dynamic dependencies in time series data (such as parameter drift trends over time), and random forest is used to handle nonlinear relationships between high-dimensional features (such as the correlation between temperature rise and capacitor capacity decrease). Through training, the model can learn the typical patterns of capacitor aging. In the training process of the respirator power supply circuit, the model finds that when the switching frequency drift exceeds a preset value and the temperature rises above a preset temperature, MOSFET breakdown will occur within a predetermined time. Based on this model, an early warning can be issued, and potential risks can be identified earlier than traditional threshold-based methods.

[0079] The information reasoning unit is used for real-time reasoning and fault positioning. The mixed data training model is used to perform sliding window sampling on the real-time collected frequency characteristic data, and a fault probability matrix is output through the LSTM network. The information reasoning unit performs sliding window sampling on the frequency characteristic data to be tested based on the mixed data training model to capture transient fault characteristics. Through the time series analysis capability of the LSTM network, the unit generates a fault probability matrix, which is a three-dimensional tensor structure (time step, fault type, and probability value), and reflects the occurrence probability of different fault types at each time. For example, in the monitoring of a linear accelerator modulator circuit, the unit detects that the 3rd harmonic energy of the pulse voltage continuously increases from the fundamental frequency to in multiple windows, and the LSTM network outputs a rapidly increasing probability of "pulse transformer inter-turn short circuit". Combined with abnormal fluctuations in equivalent inductance, the unit accurately locates the partial discharge fault of the winding.

[0080] The information output unit is used for obtaining a fault circuit node based on the fault probability matrix, and generating circuit output information containing position coordinates and confidence based on the fault circuit node. The information output unit generates circuit output information containing fault position coordinates and confidence based on the fault probability matrix. The position coordinates are pre-mapped based on the circuit topology structure and the sensor layout, for example, the element coordinates (X, Y, Z) on the PCB correspond to the positions of specific capacitors, inductors, or transistors. The confidence is calculated by multi-feature matching degree. When the multi-dimensional evidences such as frequency offset, parameter drift, and temperature anomaly are consistent, the unit outputs the information of "capacitor aging, position coordinates, and confidence" in the fault processing of the high-voltage circuit of the DR device, and the maintenance personnel can directly locate to the specific area of the circuit board to replace the capacitor, thereby shortening the maintenance time from the traditional manual troubleshooting. In addition, the unit can also synchronize the fault information to the cloud operation and maintenance system to generate an electronic work order containing fault time, type, position, and historical processing record, thereby realizing digital management of the maintenance process.

[0081] In one embodiment, the obtaining of the historical fault data set of the medical instrument and the construction of the mixed data training model based on the historical fault data set include the following steps:

[0082] According to the historical fault data set, historical circuit data and fault type data are obtained from the historical fault data set of medical instruments, the historical circuit data covers voltage, current, temperature and other original monitoring sampling values, such as the DC ripple waveform recorded by the CT high-voltage generator, the charge-discharge current curve of the energy storage capacitor of the cardiac defibrillator, etc., which are used to extract key features such as resonance frequency offset, capacitor aging coefficient, inductance temperature drift coefficient, etc. The resonance frequency offset is calculated by the difference between the actual measured value and the theoretical value, which reflects the degree of circuit parameter drift, and the inductance temperature drift coefficient is the inductance value change rate under unit temperature change, such as the influence of temperature on the stability of inductance when the temperature of a certain inductance increases. These features provide quantitative basis for fault modeling, for example, in the analysis of cardiac defibrillator circuit, when the resonance frequency offset exceeds the preset value and the capacitor aging coefficient is lower than the preset aging value, the historical data shows that the discharge energy is insufficient, which can be used as the warning threshold of resonance fault.

[0083] According to the historical circuit data, the resonance frequency offset, the capacitor aging coefficient and the inductance temperature drift coefficient are obtained.

[0084] According to the fault type data, the fault type, circuit component and environmental parameter are obtained from the fault type data, the fault type is predefined as resonance fault, element breakdown, poor contact and other classifications, the circuit component is specific to elements such as capacitor, and the environmental parameter includes temperature, humidity, power grid voltage fluctuation and other external conditions. In the historical data of DR device filament heating circuit, correlation analysis shows that the environmental temperature rises and the power grid voltage fluctuation increases, indicating that high temperature and voltage instability synergistically exacerbate element wear; breathing machine power module data shows. These correlation relationships bind the fault phenomenon with multi-dimensional factors, for example, in the anesthesia machine flow control circuit, the harmonic distortion anomaly of the signal is associated with the rise of environmental temperature, which can more accurately identify temperature-sensitive faults.

[0085] According to the fault type, circuit component and environmental parameter, a fault node-component relationship graph is constructed, based on the fault type, circuit component and environmental parameter, a fault node-component relationship graph is constructed, the node represents the fault type or component, and the weight of the edge represents the correlation strength. For example: in the graph of CT high-voltage circuit, the edge weights of "series resonance" node and "high-voltage capacitor group" "rectifier diode" are 0.85 and 0.78 respectively, indicating that these two types of components are the main causes of resonance fault; the MRI radio frequency system graph shows that the correlation degrees of "radio frequency leakage" fault and "coil connector" "matching capacitor" are 0.72 and 0.65 respectively, guiding maintenance personnel to preferentially check these high-correlation components. The graph can also integrate environmental parameter nodes, such as the edge weight of "temperature greater than 35℃" and "capacitor aging" is 0.6, which directly presents the inducing effect of environment on fault, and assists in analyzing complex fault chain.

[0086] The resonance frequency offset, the capacitance aging coefficient, and the inductance temperature drift coefficient are mapped to a fault node-component relationship graph, a mixed data training model is established, and the resonance frequency offset, the capacitance aging coefficient, and other time sequence characteristics are fused with the fault node-component relationship graph to construct the mixed data training model. The model uses an LSTM network to process time sequence data and capture the drift trend of parameters over time, such as the monthly decline of the capacitance aging coefficient; meanwhile, a graph neural network is used to analyze the graph structure and mine the potential association between components. In the modulator circuit of a linear accelerator, for example, in the case of a newly added DR device capacitance burst, the model automatically strengthens the association weight between the “overvoltage” node and the “capacitor group” component, and improves the future prediction ability of similar faults.

[0087] In one embodiment, the control module comprises:

[0088] The analysis unit is configured to analyze the circuit output information and obtain resonance suppression information. The circuit output information received by the analysis unit from the fault judgment module includes fault type, location coordinates, confidence, and frequency characteristic parameters. The unit converts unstructured fault descriptions into structured resonance suppression information through semantic analysis and parameter extraction, such as extracting key parameters such as target capacitance value adjustment amount and damping resistance activation condition from “series resonance, capacitance C5 aging, frequency offset decline”. The target resistance value of the damping resistance required for resonance energy calculation is obtained. In the CT high-voltage circuit case, the analysis unit obtains accurate parameters for subsequent control actions from the “series resonance, location C5, aging coefficient decline” information.

[0089] The judgment unit is configured to judge the resonance type of the resonance suppression information. If the resonance type is series resonance, the target capacitance value is obtained; if the resonance type is parallel resonance, the damping resistance activation instruction and the target resistance value are generated. Based on the analyzed resonance suppression information, the judgment unit first determines the resonance type through a preset threshold. If it is series resonance, the target capacitance value is calculated by calling a capacitance compensation algorithm, which considers factors such as capacitance aging coefficient, frequency offset, and circuit quality factor; if it is parallel resonance, the damping resistance activation instruction and the target resistance value are generated. The resistance value is obtained through the resonance impedance formula Determination, wherein, is the resistance value, L is the equivalent inductance, and C is the equivalent capacitance. For example, in the cardiac defibrillator circuit, after the judgment unit detects a series resonance probability of 0.92, the target capacitance value is calculated as 125 μF (the original 100 μF capacitance has aged to 80 μF, and needs to be compensated to 125% of the nominal value), and the capacitance group adjustment process is triggered. For parallel resonance (probability 0.88) of the breathing machine power module, the unit generates the instruction “activate damping resistance R7, resistance value 33 Ω”, and by closing the corresponding relay to access the damping network, the resonance peak current is suppressed;

[0090] An adjustment unit is configured to generate a dynamic adjustment capacitor combination according to the resonance type, obtain capacitor combination information, and generate a dynamic capacitor combination instruction according to the resonance type and target parameters. The combination scheme is based on a binary coded control relay array, which realizes the combination on-off of multiple adjustable capacitors, such as the combination of C1 (1 μF), C2 (2 μF), C3 (5 μF), etc., covering a compensation range of 0.1 μF to 10 μF. In the DR device filament heating circuit, the target capacitance value is 10.8 μF, and the adjustment unit selects the combination of C2 (2 μF) + C3 (5 μF) + C4 (3.8 μF) (code 01110) by traversing the pre-stored combination, and the total capacity is 10.8 μF, with an error of ≤±0.5%. For the 9.5 μF compensation requirement of the linear accelerator modulator, the unit outputs code 00111, turns on C3 (3.3 μF) + C4 (4.7 μF) + C5 (1.5 μF), and realizes accurate matching. During the adjustment process, the unit monitors the equivalent capacitance value after the combination in real time, corrects the combination scheme through closed-loop feedback, and ensures the compensation accuracy.

[0091] A compensation triggering unit is configured to obtain a trigger multi-stage damping unit for collaborative compensation according to the capacitor combination information, and the compensation triggering unit activates the multi-stage damping unit to work collaboratively according to the capacitor combination information. The damping unit is composed of an RC network and is divided into three frequency bands: L1 (1-10 kHz), L2 (10-50 kHz), and L3 (50-100 kHz), each of which corresponds to a different resonance frequency range. When the MRI radio frequency system detects a 23.5 kHz parallel resonance, the unit triggers L2 and L3 stage dampings at the same time, accesses a series network of a 10 Ω resistor and a 0.01 μF capacitor, and reduces the resonance loop energy through impedance matching. In the case of 18 kHz resonance in the anesthesia machine flow control circuit, the L1 stage damping is activated to reduce the total harmonic distortion (THD) and effectively suppress the current waveform distortion. During the compensation process, the unit dynamically adjusts the damping stages through real-time spectrum analysis to realize accurate suppression of wideband resonance.

[0092] The supplementary parameter acquisition unit is configured to output a load end control signal, generate a drive signal through a PWM modulator, acquire a control IGBT switching frequency and duty cycle according to the drive signal, and acquire a load voltage compensation parameter based on resonance suppression information. The supplementary parameter acquisition unit generates a load end control signal, adjusts the IGBT switching frequency and duty cycle through the PWM modulator, and compensates for voltage fluctuations caused by resonance. The control algorithm is based on proportional-integral-derivative adjustment, the input is the load voltage deviation, and the output is the frequency and duty cycle of the PWM signal. For example, in the CT rotating anode circuit, the PWM frequency of the unit will rise and the duty cycle will decrease due to voltage fluctuations caused by resonance. The voltage is stabilized within the preset range through capacitance compensation. In addition, the overvoltage protection threshold and current limiting parameter can also be dynamically adjusted according to the resonance suppression information.

[0093] In one embodiment, the method of acquiring a trigger multi-stage damping unit to compensate according to the capacitor matching information comprises the following steps:

[0094] A multi-stage phase compensation model is established according to the capacitor matching information. The multi-stage phase compensation model establishment unit receives capacitor matching information from the adjustment unit, which is presented in binary code form. The unit constructs a transfer function matrix of circuit impedance characteristics based on this. The matrix elements include equivalent resistance, inductance, capacitance value and frequency response parameters. Time domain capacitance values are mapped to frequency domain phase compensation coefficients through Laplace transform. For example, for a 10 μF capacitor, the model calculates that 0.7 order lead compensation is required at a frequency point of 35 kHz, corresponding to a phase compensation angle of 18°; for a 15 μF capacitor, the compensation order is increased to 1.2 orders, and the phase compensation angle is increased to 30°. These coefficients are stored in a multi-dimensional lookup table, which includes capacitance value, frequency point, temperature coefficient, etc. Real-time compensation parameters are quickly obtained through linear interpolation.

[0095] A damping signal is generated according to the multi-stage phase compensation model, and the damping signal is phase-advanced. The damping signal generation and phase-advanced unit outputs parameters according to the multi-stage phase compensation model, drives a multi-stage RC network to generate a band-pass damping signal in a specific frequency range. For example: the network is divided into three levels: L1 (1-10 kHz), L2 (10-50 kHz), and L3 (50-100 kHz), each level is controlled by a relay. In a cardiac defibrillator circuit, when a 28 kHz resonance is detected, the unit activates the L2 level network to generate a damping signal with a center frequency of 28 kHz. Phase-advanced is achieved through a phase-shifting circuit, which achieves a 30° lead at this frequency point to compensate for the phase lag caused by resonance. The compensated signal is amplified by power and injected into the resonance circuit. In the case of a respirator power module, the unit uses a lead network composed of resistance and capacitance to achieve phase compensation by adjusting the operational amplifier gain, which increases the system phase margin and significantly enhances the system stability.

[0096] The compensated signal is superimposed to the current inner loop, and a current inner loop reference value is generated, and the output voltage is adjusted by a PID controller based on the current inner loop reference value. The current inner loop superimposes the phase-advanced compensated damping signal with the original current command to generate a new current inner loop reference value. For example, when the original current command is 5A and the compensation signal amplitude is 0.5A, the reference value after superposition becomes 5.5A. The reference value is input to the PID controller, and the controller parameters are dynamically adjusted according to the system response characteristics. In the filament circuit of the DR device, the resonance causes the current fluctuation to be ±4%, and after superimposing the compensation signal, the PID controller reduces the voltage fluctuation from ±4% to ±1.2% by adjusting the IGBT switching duty cycle, ensuring that the filament temperature is stable within ±2℃. In the case of a linear accelerator modulator, the pulse width fluctuation is ±2% due to resonance, and by superimposing the compensation signal, the PID controller improves the pulse width stability and significantly improves the beam quality. The controller uses an adaptive algorithm to automatically increase the Kp value when the system response is detected to be slow, and to speed up the system response; when overshoot occurs, the Ki value is reduced to suppress integral saturation. This dynamic adjustment mechanism enables the system to maintain optimal performance under different working conditions.

[0097] In one embodiment, the control module further comprises:

[0098] The receiving unit receives the resonance fault report of the fault judgment module, obtains the current frequency and the resonance frequency range according to the resonance fault report, and the receiving unit serves as the input interface of the control module and is responsible for receiving the resonance fault report output by the fault judgment module. The key parameters related to frequency are extracted from the report, including the real-time working frequency of the current circuit (i.e. the actual operating frequency of the circuit when the fault occurs) and the frequency range of the resonance (i.e. the frequency interval in which the resonance phenomenon occurs). The receiving unit converts the unstructured description of the current frequency and the resonance frequency range in the report into standardized digital signals through a preset analysis rule. For example, the information “current frequency 35kHz, resonance frequency range 34.8-35.2kHz” is extracted from the report to form structured data containing these two types of frequency parameters, which are temporarily stored in an internal register and trigger the subsequent calculation process.

[0099] The calculation unit calculates the target capacitance value required to eliminate the resonance according to the current frequency and the resonance frequency range. The calculation unit uses the circuit resonance formula wherein, represents the target capacitance value, and represents the natural constant, The equivalent inductance is equivalent to the circuit. For example, in the cardiac defibrillator circuit, when the resonance frequency of 28 kHz is detected, the calculation unit calculates the target capacitance value of 125 μF according to the pre-stored equivalent inductance value (such as 1.2 mH) to compensate for the capacity attenuation caused by the aging of the capacitor. The calculation unit also considers the influence of temperature on the capacitance value, introduces the temperature coefficient α for correction, and in the case of a breathing machine power module, the ambient temperature is 40℃ (the reference temperature is 25℃), and the capacitor temperature coefficient α = 200ppm / ℃. The calculation unit corrects the target capacitance value from 8.2 μF to 8.225 μF, improving the compensation accuracy.

[0100] The capacitance value acquisition unit is used to traverse the adjustable capacitors of the series compensation device and acquire the capacitance values of each adjustable capacitor. The capacitance value acquisition unit is responsible for traversing the adjustable capacitor array in the series compensation device and reading the real-time value of each capacitor through the I2C bus. The array is composed of multiple adjustable capacitors with different nominal values and corresponding relay switches.

[0101] The matching unit is used to select the to-be-controlled adjustable capacitor group from the adjustable capacitors according to the capacitance value matching algorithm, and the matching unit uses the dynamic programming-based capacitance value matching algorithm to find the optimal combination in the adjustable capacitor array, so that the total capacitance value after combination is closest to the target capacitance value.

[0102] The control instruction generation unit is used to generate control instructions to control the control switches in series with each adjustable capacitor in the to-be-controlled adjustable capacitor group to be turned off. The control instruction generation unit generates corresponding binary control instructions according to the to-be-controlled adjustable capacitor group output by the matching unit to control the on-off state of the solid-state relay in series with each adjustable capacitor.

[0103] The verification unit is used for re-acquiring the circuit electrical parameters after disconnection, verifying whether the resonance is eliminated, and if not, repeatedly performing the matching algorithm according to the capacitance value to filter out the to-be-controlled adjustable capacitance group with the total capacitance value closest to the target capacitance value from the adjustable capacitance, generating the control instruction, controlling the control switch connected in series with each adjustable capacitance in the to-be-controlled adjustable capacitance group to be disconnected, and stopping until the resonance is eliminated. The adjustment process and result of the capacitance group are recorded to the database. After the control instruction is executed, the verification unit re-acquires the circuit electrical parameters (such as voltage, current, and frequency), verifies whether the resonance is eliminated by analyzing the total harmonic distortion (THD) and the phase margin. The THD < 5% and the phase margin > 45° are set as the judgment standard of the resonance elimination. The unit evaluates the adjustment effect by comparing the change rate of the electrical parameters before and after the adjustment, and dynamically adjusts the tolerance value by using the PID algorithm. In the case of the linear accelerator modulator, the frequency decreases after the first adjustment, but the THD is still not in the preset range. The verification unit instructs to continue the iterative adjustment. After three times of optimization, the THD is reduced to the preset range, and it is confirmed that the resonance is eliminated. The verification unit also records the capacitance group of each adjustment, the electrical parameter change, the adjustment time and other information to the database, forming a complete fault processing log to provide data support for subsequent analysis.

[0104] As shown in Figure 2 The present application also provides a medical instrument circuit management control method, comprising:

[0105] S1, acquiring voltage data, current data and temperature data of the circuit to construct a real-time data set;

[0106] S2, performing time-space registration, noise filtering and feature extraction on the real-time data set to generate frequency feature data;

[0107] S3, analyzing the frequency feature data through an LSTM network to identify the resonance type and the fault position, and generating circuit output information;

[0108] S4, acquiring capacitance group information of a compensation module and load end output information according to the circuit output information.

[0109] In one embodiment, the step of performing time-space registration, noise filtering and feature extraction on the real-time data set to generate frequency feature data comprises:

[0110] S201, aligning the voltage data, current data and temperature data through timestamps, compensating for the sensor layout difference based on a spatial interpolation algorithm, and generating a time-space synchronous voltage, current and temperature matrix;

[0111] S202, filtering out high-frequency noise from the time-space synchronous voltage, current and temperature matrix through a threshold method, and retaining the fundamental frequency and low-order harmonic components;

[0112] S203, extract frequency characteristics from the space-time synchronization voltage, current and temperature matrix reserved for the fundamental frequency and low harmonic components, obtain the FFT spectrum of the voltage signal, and identify the main resonance frequency and the sub-resonance frequency band;

[0113] S204, obtain the current waveform segment corresponding to the main resonance frequency and the sub-resonance frequency band output by the frequency identification unit, obtain the equivalent inductance and capacitance under the current waveform segment through current waveform integration, and generate a feature vector. The feature vector is input into a Kalman filter for dynamic correction, and the optimized frequency characteristic data is output.

[0114] In one embodiment, the step of analyzing the frequency characteristic data through the LSTM network, identifying the resonance type and the fault location, and generating the circuit output information comprises:

[0115] S301, receive the frequency characteristic data, and output the resonance information label and the fault information label corresponding to each resonance type and each fault level through the Softmax function;

[0116] S302, obtain the historical fault data set of the medical instrument, extract the resonance type label, the fault level label and the corresponding frequency characteristic data for training from the historical fault data set, and construct a mixed data training model by iteratively training the frequency characteristic data;

[0117] S304, based on the mixed data training model, perform sliding window sampling on the real-time collected frequency characteristic data to be tested, and output the fault probability matrix corresponding to each window through the LSTM network;

[0118] S305, according to the fault probability matrix, obtain the positioning fault circuit node, and generate the circuit output information containing the position coordinates and the confidence based on the positioning fault circuit node.

[0119] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, value library or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0120] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, so that a process, device, article or method that includes a list of elements not only includes those elements, but also includes other elements not expressly listed or inherent to such process, device, article or method. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, device, article or method that includes the element.

[0121] The above only describes the preferred embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent result or equivalent process transformation obtained by using the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A medical instrument circuit management control system comprising a compensation module, characterized by, Also include: Data acquisition module, for acquiring voltage data, current data and temperature data of the circuit, constructing real-time data set; Data processing module, for time and space registration, noise filtering and feature extraction of the real-time data set, generating frequency feature data; Fault judgment module, for analyzing the frequency feature data through the LSTM network, identifying the resonance type and the fault position, and generating the circuit output information; Control module, for obtaining the capacitance group information and the load end output information of the compensation module according to the circuit output information, the control module comprising: Analysis unit, for analyzing the circuit output information to obtain resonance suppression information; Judgment unit, for judging the resonance type of the resonance suppression information, if the resonance type is series resonance, obtaining the target capacitance value, if the resonance type is parallel resonance, generating the damping resistance activation instruction and the target resistance value; Adjustment unit, for generating dynamic adjustment capacitance group according to the resonance type, obtaining capacitance group information; Compensation triggering unit, for triggering multi-stage damping unit cooperative compensation according to the capacitance group information, the compensation triggering unit comprising: According to the capacitance group information, a multi-stage phase compensation model is established; According to the multi-stage phase compensation model, a damping signal is generated, and the damping signal is phase-advanced compensated; The compensated signal is superimposed to the current inner loop, and a current inner loop reference value is generated, and the output voltage is adjusted by the PID controller based on the current inner loop reference value; Supplementary parameter acquisition unit, for outputting load end control signal, generating driving signal through PWM modulator, obtaining control IGBT switching frequency and duty cycle according to the driving signal, and obtaining load voltage compensation parameters based on resonance suppression information.

2. A medical instrument circuit management control system according to claim 1, wherein, The data processing module comprises: Data matrix generation unit, for aligning voltage data, current data and temperature data through timestamp, and compensating sensor layout difference based on spatial interpolation algorithm, generating time and space synchronized voltage, current and temperature matrix; Data processing unit, for filtering high frequency noise from time and space synchronized voltage, current and temperature matrix through threshold method, and retaining fundamental frequency and low harmonic components; Frequency identification unit, for extracting frequency features from time and space synchronized voltage, current and temperature matrix retaining fundamental frequency and low harmonic components, obtaining FFT frequency spectrum of voltage signal, and identifying main resonance frequency and secondary resonance frequency band; Frequency output unit, for obtaining current waveform segment corresponding to main resonance frequency and secondary resonance frequency band output by frequency identification unit, obtaining equivalent inductance and capacitance under current waveform segment through current waveform integration, and generating feature vector, inputting the feature vector into Kalman filter for dynamic correction, and outputting optimized frequency feature data.

3. The medical instrument circuit management control system of claim 1, wherein, The fault judgment module comprises: Frequency input unit, for receiving the frequency feature data, and outputting resonance information label and fault information label corresponding to each resonance type and each fault grade through Softmax function; The frequency training unit is configured to obtain a historical fault data set of the medical instrument, extract a resonance type label, a fault level label, and corresponding frequency feature data for training from the historical fault data set, and construct a hybrid data training model by iteratively training the frequency feature data. The information reasoning unit is configured to perform sliding window sampling on the real-time collected frequency feature data based on the hybrid data training model, and output a fault probability matrix corresponding to each window through the LSTM network. The information output unit is configured to obtain a located fault circuit node according to the fault probability matrix, and generate circuit output information including a position coordinate and a confidence level based on the located fault circuit node.

4. A medical instrument circuit management control system according to claim 3, wherein, The method for obtaining the historical fault data set of the medical instrument and constructing the hybrid data training model based on the historical fault data set comprises the following steps: Obtain historical circuit data and fault type data based on the historical fault data set. Obtain a resonance frequency offset, a capacitor aging coefficient, and an inductance temperature drift coefficient based on the historical circuit data. Obtain a fault type, a circuit component, and an environmental parameter based on the fault type data. Construct a fault node-component relationship graph based on the fault type, the circuit component, and the environmental parameter. Map the resonance frequency offset, the capacitor aging coefficient, and the inductance temperature drift coefficient to the fault node-component relationship graph to establish the hybrid data training model.

5. The medical instrument circuit management control system of claim 1, wherein, The control module further comprises: The receiving unit is configured to receive a resonance fault report of the fault judgment module, and obtain a current frequency and a resonance frequency range based on the resonance fault report. The calculation unit is configured to calculate a target capacitor value required to eliminate resonance based on the current frequency and the resonance frequency range. The capacitor value obtaining unit is configured to traverse the adjustable capacitors of the series compensation device, and obtain capacitor values of the adjustable capacitors. The matching unit is configured to filter out, from the adjustable capacitors, a to-be-controlled adjustable capacitor group whose combined total capacitor value is closest to the target capacitor value based on a capacitor value matching algorithm. The control instruction generation unit is configured to generate a control instruction to control a control switch connected in series with each adjustable capacitor in the to-be-controlled adjustable capacitor group to be turned off. The verification unit is configured to re-collect circuit electrical parameters after turning off, verify whether resonance is eliminated, and if not, repeatedly perform the steps of filtering out, from the adjustable capacitors, the to-be-controlled adjustable capacitor group whose combined total capacitor value is closest to the target capacitor value based on the capacitor value matching algorithm, generating the control instruction to control the control switch connected in series with each adjustable capacitor in the to-be-controlled adjustable capacitor group to be turned off, until resonance is eliminated, and record the capacitor group adjustment process and results to a database.

6. A medical instrument circuit management control method, characterized by, The method comprises the following steps: Obtain voltage data, current data, and temperature data of the circuit to construct a real-time data set. Perform time-space registration, noise filtering, and feature extraction on the real-time data set to generate frequency feature data. Analyze the frequency feature data through an LSTM network to identify a resonance type and a fault position, and generate circuit output information. Obtain capacitor group information of a compensation module and load end output information based on the circuit output information. The method for obtaining the capacitor group information of the compensation module and the load end output information based on the circuit output information comprises the following steps: Analyzing the circuit output information to obtain resonance suppression information; Judging the resonance type of the resonance suppression information, if the resonance type is series resonance, obtaining a target capacitance value, if the resonance type is parallel resonance, generating a damping resistor activation instruction and a target resistance value; Generating a dynamically adjusted capacitance package according to the resonance type to obtain capacitance package information; According to the capacitance package information, triggering multi-stage damping unit cooperative compensation; Outputting a load end control signal, generating a drive signal through a PWM modulator, obtaining control IGBT switching frequency and duty cycle according to the drive signal, and obtaining load voltage compensation parameters based on resonance suppression information; The step of obtaining the trigger multi-stage damping unit cooperative compensation according to the capacitance package information comprises the following steps: Establishing a multi-stage phase compensation model according to the capacitance package information; Generating a damping signal according to the multi-stage phase compensation model, and performing phase lead compensation on the damping signal; Superimposing the compensated signal to the current inner loop, generating a current inner loop reference value, and adjusting the output voltage through a PID controller based on the current inner loop reference value.

7. The method of claim 6, wherein, The step of performing spatio-temporal registration, noise filtering and feature extraction on the real-time data set to generate frequency feature data comprises: Aligning voltage data, current data and temperature data through timestamps, and compensating for sensor layout differences based on a spatial interpolation algorithm to generate a spatio-temporal synchronous voltage, current and temperature matrix; Filtering high-frequency noise from the spatio-temporal synchronous voltage, current and temperature matrix through threshold method, and retaining fundamental frequency and low-order harmonic components; Extracting frequency features from the spatio-temporal synchronous voltage, current and temperature matrix retaining fundamental frequency and low-order harmonic components, obtaining FFT spectrum of voltage signal, and identifying main resonance frequency and sub-resonance frequency band; Obtaining the current waveform segment corresponding to the main resonance frequency and sub-resonance frequency band output by the frequency identification unit, obtaining equivalent inductance and capacitance under the current waveform segment through current waveform integration, and generating a feature vector, inputting the feature vector into a Kalman filter for dynamic correction, and outputting optimized frequency feature data.

8. The method of claim 6, wherein, The step of analyzing the frequency feature data through the LSTM network, identifying the resonance type and the fault location, and generating the circuit output information comprises: Receiving the frequency feature data, and outputting resonance information labels and fault information labels corresponding to each resonance type and each fault level through a Softmax function; Obtaining a historical fault data set of a medical instrument, training resonance type labels, fault level labels and corresponding extracted frequency feature data from the historical fault data set, and constructing a hybrid data training model through iterative training of frequency feature data; Based on the hybrid data training model, performing sliding window sampling on the real-time collected frequency feature data to be tested, and outputting a fault probability matrix corresponding to each window through the LSTM network; According to the fault probability matrix, obtaining a positioning fault circuit node, and generating circuit output information containing position coordinates and confidence based on the positioning fault circuit node.

Images