A method for real-time monitoring and early warning of abnormalities in surgical equipment
By deploying multimodal data acquisition nodes in surgical equipment, and combining hardware filtering and software algorithm noise reduction, a temperature-electrical parameter coupling model was established. Feature enhancement and multi-feature fusion recognition algorithms were adopted to solve the problem of the influence of temperature on electrical variable monitoring, achieve accurate restoration of electrical variables and accurate identification of faults, and improve the operational stability and safety of surgical equipment.
Patent Information
- Application Number
- CN202611235186.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-14
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies have failed to effectively address the impact of temperature on chip electrical parameters, resulting in large errors in electrical variable monitoring. They are unable to fully identify ripple anomalies, electrical variable anomalies, and circuit anomalies, leading to insufficient monitoring accuracy, inaccurate fault identification, and delayed response, which affects the stability and safety of surgical equipment operation.
By deploying multimodal data acquisition nodes and combining hardware filtering and software dual-algorithm noise reduction, a temperature-electrical parameter coupling model is established. Feature enhancement and multi-feature fusion recognition algorithms are used to restore the true value of electrical variables and distinguish fault categories. A weighted fusion algorithm is used for collaborative judgment of multiple anomalies.
It achieves accurate restoration of the true value of electrical variables, reduces the misjudgment rate caused by temperature and ripple interference, improves the accuracy and comprehensiveness of fault identification, and ensures the operational stability and safety of surgical equipment.
Smart Images

Figure CN123101263A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electrical variable monitoring, specifically a method for real-time electrical variable monitoring and abnormal early warning of surgical equipment. Background Technology
[0002] Surgical equipment refers to medical devices and apparatus used for surgical procedures, diagnosis, or treatment. Real-time electrical variable monitoring and anomaly early warning methods utilize sensors, intelligent instruments, and data analysis technology to collect, transmit, and analyze the electrical parameters (such as voltage, current, power, and frequency) of surgical equipment in real time. Combined with preset thresholds or machine learning models, abnormal electrical signals (such as overload, short circuit, and voltage fluctuations) are instantly identified and trigger early warnings. Its core lies in dynamically tracking changes in electrical variables to detect potential equipment malfunctions or operational anomalies in advance, ensuring surgical safety. For example, in neurosurgery, if abnormal fluctuations in the electrosurgical current could cause tissue burns, the system will immediately alarm and suspend equipment output.
[0003] Existing technologies lack a temperature-electrical parameter coupling model, neglecting the impact of temperature on chip electrical parameters and failing to reproduce the true values of electrical variables. In extreme temperature scenarios such as high and low temperatures, the monitoring error of electrical variables is extremely large, failing to reflect the true operating status of the equipment. Furthermore, the absence of feature enhancement and multi-feature fusion recognition algorithms makes it impossible to resolve waveform confusion issues arising from similar faults, resulting in low accuracy in fault category differentiation, a high risk of misjudgment, and incorrect fault handling. These technologies primarily target single-type anomalies, failing to comprehensively identify ripple anomalies, electrical variable anomalies, and circuit anomalies, and easily overlooking scenarios with multiple anomalies overlapping. This can lead to insufficient monitoring accuracy, inaccurate fault identification, and delayed response, severely impacting the operational stability of surgical equipment and surgical safety. Summary of the Invention
[0004] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a real-time electrical variable monitoring and anomaly early warning method for surgical equipment. This method addresses the technical problems of ignoring the influence of temperature on chip electrical parameters, failing to restore the true value of electrical variables, and failing to fully identify ripple anomalies, electrical variable anomalies, and circuit anomalies. These problems result in insufficient monitoring accuracy, inaccurate fault identification, and delayed response, which seriously affect the operational stability of surgical equipment and surgical safety.
[0005] To address the above problems, a first aspect of the present invention provides a method for real-time electrical variable monitoring and abnormality early warning for surgical equipment, comprising the following steps:
[0006] Multimodal data acquisition nodes are deployed in the surgical equipment within the managed area to collect power ripple-related data, temperature adaptation-related data, circuit control-related data, and auxiliary monitoring data.
[0007] The residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction. The ripple signal in the acquired data is restored and the ripple frequency components are extracted. The frequency of the denoised ripple signal is analyzed by fast Fourier transform.
[0008] Based on the power ripple and temperature adaptation data of active and passive chips in the surgical equipment in the management area, and combined with the frequency analysis results of the ripple signal, a temperature-electrical parameter coupling model is established, and a temperature coupling model of chip operating voltage and on-resistance is established. The influence coefficient of temperature on electrical variables is analyzed, and the true value of electrical variables is restored and anomalies are identified.
[0009] Establish feature enhancement and multi-feature fusion recognition algorithms to distinguish fault categories based on the true values of electrical variables and anomaly recognition results, while adapting to transient fault responses;
[0010] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively. If two or more types of anomalies are identified at the same time, a weighted fusion algorithm is used to perform multi-anomaly collaborative judgment.
[0011] Optionally, in one example of the above aspects, residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction, the ripple signal in the acquired data is restored, and the ripple frequency components are extracted. The frequency analysis of the denoised ripple signal is then performed using Fast Fourier Transform, including the following steps:
[0012] Hardware filtering uses a second-order active low-pass filter combined with a notch filter to initially filter out high-frequency noise and power frequency interference.
[0013] The software dual-algorithm denoising uses a wavelet threshold denoising algorithm to suppress residual noise and restore the ripple signal in the power supply ripple-related data. The formula is as follows:
[0014] Let the ripple signal at time t in the collected power supply ripple correlation data be... The noise signal is The actual ripple signal is ,but ;right Wavelet decomposition was performed with 5 levels, using the db4 wavelet basis. The low-frequency components were obtained after decomposition. With high frequency components ;
[0015] High-frequency components are processed using an adaptive threshold function to suppress noise;
[0016] Perform inverse wavelet transform on the low-frequency and high-frequency components of the restored ripple signal to obtain the denoised true ripple signal. ;
[0017] right Discretization is performed to obtain discrete signals. The frequency components of the ripple signal after noise reduction are extracted by using Fast Fourier Transform (FFT), thus obtaining the frequency domain characteristics of the discrete signal after FFT transformation. ;
[0018] The spectral amplitude of the frequency domain characteristics after the discrete signal undergoes FFT transformation is calculated. Extract the fundamental frequency and harmonic frequencies of the ripple signal.
[0019] Optionally, in one example of the above aspects, based on the power ripple and temperature adaptation data of active and passive chips in the surgical equipment of the managed area, and combined with the ripple signal frequency analysis results, a temperature-electrical parameter coupling model is established, including the following steps:
[0020] Based on the temperature adaptation data of active and passive chips in the surgical equipment in the management area, the reference values of electrical parameters of active and passive chips at different temperatures are obtained. Based on the nonlinear relationship between the electrical parameters of the surgical equipment chips and temperature, and combined with the fundamental frequency and harmonic frequency after the ripple signal is restored, a temperature-electrical parameter coupling model is established.
[0021] Optionally, in one example of the above aspects, a temperature coupling model of chip operating voltage and on-resistance is established, the influence coefficient of temperature on electrical variables is analyzed, and the true value of electrical variables is restored and anomalies are identified, including the following steps:
[0022] Based on the temperature adaptation data of the active and passive chips corresponding to the core nodes of the surgical equipment circuit in the management area, the reference values of the electrical parameters of the active and passive chips at different temperatures are obtained, combined with the fundamental frequency and harmonic frequency after the ripple signal is restored.
[0023] Establish a voltage-temperature coupling model for active chips;
[0024] Based on the on-resistance of active / passive chips and their temperature characteristics, a temperature coupling model for the on-resistance of active chips is established:
[0025] The on-resistance of an active chip increases non-linearly with increasing temperature, which can be substituted into the general model.
[0026] Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, the operating voltage and on-resistance of the core nodes at temperature T are calculated as theoretical values. ;
[0027] The least squares method is used to fit the actual value to the theoretical value, and the temperature influence coefficients α and β are solved. Combined with the drift correction of the temperature influence coefficients, the final true value of the electrical variable is obtained. ;
[0028] The actual operating voltage and on-resistance of the chip are obtained after restoration. Based on the difference between the actual and theoretical values, preliminary anomaly identification is performed.
[0029] Optionally, in one example of the above aspects, a feature enhancement and multi-feature fusion identification algorithm is established to distinguish fault categories based on the true values of electrical variables and anomaly identification results, while adapting to transient fault responses, including the following steps:
[0030] For similar fault current waveforms, deep features are extracted using wavelet packet transform, and a feature enhancement algorithm is established to enhance feature discriminative power. The formula is as follows:
[0031] Let the original waveform of the fault current be... Wavelet packet decomposition was performed with four levels, and the db6 wavelet basis was selected to obtain the wavelet packet coefficients of each node. Where j is the decomposition level, k is the node number, and n is the discrete-time variable;
[0032] By using the SVM improved support vector machine algorithm, a multi-feature fusion identification algorithm is established, incorporating fault physical mechanism rules, and using the elements in the multi-dimensional feature vector F as input samples to construct a fault identification model;
[0033] Fault categories are distinguished through a fault identification model, while also adapting to transient fault responses.
[0034] Optionally, in one example of the above aspects, based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively, including the following steps:
[0035] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of distinguishing fault categories, a ripple anomaly identification scheme is integrated to identify ripple anomalies.
[0036] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, an electrical variable anomaly identification algorithm is integrated to identify electrical variable anomalies.
[0037] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant circuit control data and auxiliary monitoring data, a circuit electrical variable anomaly identification algorithm is established to identify electrical variable anomalies.
[0038] Optionally, in one example of the above aspects, identifying ripple anomalies includes the following steps:
[0039] Based on the amplitude and frequency components of the ripple signal, a ripple anomaly identification scheme is defined, including: amplitude criterion, frequency criterion, and co-criterion.
[0040] The amplitude criterion is: calculate the effective value of the ripple voltage after noise reduction. :
[0041] like The ripple amplitude was determined to be abnormal.
[0042] The frequency criterion is: if the ripple frequency components extracted by FFT contain harmonics that are not the inherent frequency of the power supply, or if the fundamental frequency drifts by more than ±5%, it is determined to be an abnormal ripple frequency.
[0043] The collaborative criterion is: if any of the above criteria is met and the duration is ≥100ms, it is judged as an abnormal ripple, which is associated with the risk of reduced power conversion efficiency;
[0044] The determination results of amplitude criteria, frequency criteria, and co-jurisdictional criteria of historical data are obtained. Combined with the processing results of electrical variable real value restoration and anomaly identification, as well as the fault category differentiation results and the ripple anomaly category corresponding to historical data, a DNN deep neural network is trained, and the ripple anomaly type is identified through the trained model.
[0045] Optionally, in one example of the above aspects, identifying electrical variable anomalies includes the following steps:
[0046] Based on the real electrical variables after compensation and correction, combined with trend prediction, static anomaly identification and dynamic trend prediction are carried out.
[0047] Static anomaly identification includes: calculating the deviation rate between the corrected actual electrical variable and the baseline value at the corresponding temperature; if the deviation rate between the corrected actual electrical variable and the baseline value at the corresponding temperature is greater than a threshold, it is determined to be a static anomaly;
[0048] Dynamic trend prediction includes: using a linear regression algorithm to fit the trend of the corrected electrical variables and predict the trend of changes in the electrical variables within a preset time period in the future;
[0049] If the absolute value of the slope k exceeds the preset rate threshold, it is judged as an abnormal dynamic trend, indicating that the electrical variable will deviate seriously.
[0050] The static anomaly identification and dynamic trend prediction results of historical data are obtained. The processing results of the restoration of the true value of electrical variables and anomaly identification, as well as the fault category differentiation results, and the corresponding electrical variable anomaly categories in historical data are used to train a DNN deep neural network. The trained model is then used to identify the anomaly type of electrical variables.
[0051] Optionally, in one example of the above aspects, identifying electrical variable anomalies includes the following steps:
[0052] Based on the normalized circuit loop electrical variables of the surgical equipment, the SVM (Support Vector Machine) algorithm is used to calculate the normalized voltage. Current ,power As input features, an SVM classification model is constructed, which is divided into normal, overvoltage, overcurrent, undervoltage, and poor contact.
[0053] The model output identifies the types of circuit anomalies, including overvoltage, overcurrent, undervoltage, and poor contact.
[0054] Compared with the prior art, the beneficial effects of the present invention are:
[0055] This invention combines ripple frequency analysis results, chip ripple data, and temperature data to establish a general temperature-electrical parameter coupling model and a specific model for chip operating voltage and on-resistance. It quantifies the influence of temperature on electrical variables, addressing the pain point of electrical variable distortion under wide temperature environments. By analyzing the influence coefficient of temperature on electrical variables, it achieves accurate restoration of the true values of electrical variables, eliminating dual interference from temperature and ripple, and ensuring that the monitoring data closely reflects the actual operating state of the equipment. Based on the restored true values of electrical variables, combined with static thresholds and dynamic trend criteria, it achieves accurate identification of abnormal chip electrical parameters, reducing the false positive and false negative rates caused by temperature drift and ripple interference, and is particularly suitable for complex temperature operating scenarios of surgical equipment.
[0056] This invention integrates the true values of electrical variables, anomaly identification results, and fault category differentiation results, combined with loop control data, to achieve accurate classification and identification of ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies, covering all-dimensional anomaly scenarios of equipment electrical systems; it adopts a weighted fusion algorithm to achieve collaborative judgment of multiple anomalies, overcomes the limitations of single anomaly identification, avoids the expansion of faults caused by the superposition of multiple anomalies, and improves the comprehensiveness and reliability of anomaly judgment. Attached Figure Description
[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0058] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation
[0059] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0060] Please see Figure 1 The first aspect of the present invention provides a method for real-time electrical variable monitoring and abnormal early warning of surgical equipment, comprising the following steps:
[0061] Multimodal data acquisition nodes are deployed in the surgical equipment within the managed area to collect power ripple-related data, temperature adaptation-related data, circuit control-related data, and auxiliary monitoring data.
[0062] The residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction. The ripple signal in the acquired data is restored and the ripple frequency components are extracted. The frequency of the denoised ripple signal is analyzed by fast Fourier transform.
[0063] Based on the power ripple and temperature adaptation data of active and passive chips in the surgical equipment in the management area, and combined with the frequency analysis results of the ripple signal, a temperature-electrical parameter coupling model is established, and a temperature coupling model of chip operating voltage and on-resistance is established. The influence coefficient of temperature on electrical variables is analyzed, and the true value of electrical variables is restored and anomalies are identified.
[0064] Establish feature enhancement and multi-feature fusion recognition algorithms to distinguish fault categories based on the true values of electrical variables and anomaly recognition results, while adapting to transient fault responses;
[0065] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively. If two or more types of anomalies are identified at the same time, a weighted fusion algorithm is used to perform multi-anomaly collaborative judgment.
[0066] In this embodiment, the power supply ripple related data includes: voltage ripple signal and current ripple signal, and synchronously acquired power supply input / output voltage, current, and active power; wherein, the sampling frequency of the current ripple signal is ≥1MHz;
[0067] Temperature adaptation-related data includes: temperature of the core circuit node, ambient temperature, and synchronously collected chip operating voltage, current, on-resistance, and insulation impedance; among which, the core circuit node is set as active chip and passive chip;
[0068] The relevant data for circuit control include: voltage, current, leakage current, and power factor of each branch circuit of the surgical equipment, and synchronously collect circuit contact resistance and circuit temperature;
[0069] Auxiliary monitoring data include: the sensor's own working status and the device's operating mode (such as high-frequency electrosurgical cutting / coagulation mode).
[0070] The sampling strategy is set to use a fixed high refresh rate (500kHz~1MHz) during normal operation. When the device starts up, switches modes, experiences sudden temperature changes, or experiences load fluctuations, the sampling rate is increased to 1MHz~2MHz to capture transient electrical variable characteristics and avoid missing transient anomalies.
[0071] This embodiment effectively suppresses power frequency interference, high-frequency noise, and residual noise in the acquired data by combining hardware filtering with software dual-algorithm noise reduction, significantly improving the accuracy of ripple signal restoration. Based on Fast Fourier Transform, frequency analysis is performed on the denoised ripple signal to accurately extract the fundamental and harmonic frequencies of the ripple, which can directly diagnose power quality problems and eliminate ripple interference for subsequent electrical parameter correction and coupling modeling. The accurate restoration and frequency analysis of the ripple signal provide a core basis for identifying power supply ripple anomalies, enabling early detection of potential problems such as decreased power conversion efficiency and load power drift, reducing the risk of equipment failure caused by ripple anomalies.
[0072] This embodiment combines ripple frequency analysis results, chip ripple data, and temperature data to establish a general temperature-electrical parameter coupling model and a specific model for chip operating voltage and on-resistance. It quantifies the influence of temperature on electrical variables, addressing the pain point of electrical variable distortion under wide temperature environments. By analyzing the influence coefficient of temperature on electrical variables, it achieves accurate restoration of the true values of electrical variables, eliminating dual interference from temperature and ripple, and ensuring that the monitoring data closely reflects the actual operating state of the equipment. Based on the restored true values of electrical variables, combined with static thresholds and dynamic trend criteria, it achieves accurate identification of abnormal chip electrical parameters, reducing the false positive and false negative rates caused by temperature drift and ripple interference, and is particularly suitable for complex temperature operating scenarios of surgical equipment.
[0073] This embodiment extracts deep fault features through a feature enhancement algorithm, solving the problem of similar current waveforms and non-unique features in faults such as coil demagnetization and mechanism jamming in surgical equipment, thus improving the accuracy of fault category differentiation. The multi-feature fusion recognition algorithm takes into account both data-driven and physical logic, using sample data to improve recognition accuracy while avoiding the "black box" problem of artificial intelligence models, thereby improving the verifiability of fault recognition results and reducing the resistance to engineering implementation. The lightweight design of the algorithm is adapted to local edge computing, achieving millisecond-level response to transient faults, solving the communication latency problem caused by the reliance on cloud platforms in traditional systems, and meeting the needs of rapid handling of transient faults in surgical equipment.
[0074] This embodiment integrates the actual values of electrical variables, anomaly identification results, and fault category differentiation results, combined with loop control data, to achieve accurate classification and identification of ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies, covering all-dimensional anomaly scenarios of the equipment's electrical system. A weighted fusion algorithm is used to achieve collaborative judgment of multiple anomalies, overcoming the limitations of single anomaly identification, avoiding fault expansion caused by the superposition of multiple anomalies, and improving the comprehensiveness and reliability of anomaly judgment. The collaborative judgment results provide accurate basis for graded early warning and dynamic adjustment, enabling graded handling of "minor anomalies requiring fine-tuning, and severe anomalies requiring shutdown," balancing the continuity of surgical equipment operation with surgical safety, and avoiding unnecessary surgical interruptions.
[0075] In one embodiment of the present invention, residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction, the ripple signal in the acquired data is restored, and the ripple frequency components are extracted. The frequency analysis of the denoised ripple signal is then performed using Fast Fourier Transform, including the following steps:
[0076] Hardware filtering uses a second-order active low-pass filter (cutoff frequency 1.5MHz) combined with a notch filter (to suppress 50Hz power frequency interference) to initially filter out high-frequency noise and power frequency interference.
[0077] The software dual-algorithm denoising uses a wavelet threshold denoising algorithm to suppress residual noise and restore the ripple signal in the power supply ripple-related data. The formula is as follows:
[0078] Let the ripple signal at time t in the collected power supply ripple correlation data be... The noise signal is The actual ripple signal is ,but ;right Wavelet decomposition was performed with 5 levels, using the db4 wavelet basis. The low-frequency components were obtained after decomposition. With high frequency components ;
[0079] High-frequency components are processed using an adaptive threshold function to suppress noise.
[0080]
[0081] in, The threshold value for the high-frequency components of the j-th layer is... Here, N is the noise estimate, and N is the length of the discrete signal.
[0082] Perform inverse wavelet transform on the low-frequency and high-frequency components of the restored ripple signal to obtain the denoised true ripple signal. ;
[0083] right Discretization is performed to obtain discrete signals. The frequency components of the ripple signal after noise reduction are extracted by using Fast Fourier Transform (FFT), thus obtaining the frequency domain characteristics of the discrete signal after FFT transformation. ;
[0084] The spectral amplitude of the frequency domain characteristics after the discrete signal undergoes FFT transformation is calculated. The fundamental and harmonic frequencies of the ripple signal are extracted. These frequencies can be used for power quality diagnostics.
[0085] In one embodiment of the present invention, based on the power ripple-related data and temperature adaptation-related data of active and passive chips in the surgical equipment in the management area, and combined with the ripple signal frequency analysis results, a temperature-electrical parameter coupling model is established, including the following steps:
[0086] Based on temperature adaptation data of active and passive chips in surgical equipment within the managed area, baseline electrical parameter values for active chips (such as driver chips) and passive chips (such as resistors and capacitors) at different temperatures (-10℃ to 85℃, covering the possible operating temperature range of the surgical equipment) were obtained. Furthermore, considering the non-linear relationship between the electrical parameters of the surgical equipment chips and temperature, and combining the fundamental and harmonic frequencies after ripple signal reconstruction, a temperature-electrical parameter coupling model was established. The formula is as follows:
[0087]
[0088] Where X(T) represents the electrical parameters of the chip at temperature T (including operating voltage U or on-resistance R); X0 represents the room temperature reference value (T0=25℃, the normal operating temperature of surgical equipment), i.e., the measured value of the electrical parameters at T=T0; T represents the real-time temperature of the chip core node (in ℃), T∈[-10, 85], and T0 represents the preset standard temperature of the chip; α represents the first-order temperature influence coefficient (linear term coefficient), reflecting the degree of linear influence of temperature on electrical parameters; β represents the second-order temperature influence coefficient (nonlinear term coefficient), reflecting the degree of nonlinear influence of temperature on electrical parameters.
[0089] In this embodiment, the surgical device chip includes an active chip and a passive chip; the α and β coefficients of the active chip and the passive chip are different and need to be calibrated separately; the ripple frequency analysis results have been incorporated into the model through electrical parameter correction to ensure that the model is not affected by ripple interference.
[0090] In one embodiment of the present invention, a temperature coupling model of chip operating voltage and on-resistance is established, the influence coefficient of temperature on electrical variables is analyzed, and the true value of electrical variables is restored and anomalies are identified, including the following steps:
[0091] Based on the temperature adaptation data of the active and passive chips corresponding to the core nodes of the surgical equipment circuit in the management area, the reference values of the electrical parameters of the active and passive chips at different temperatures are obtained, combined with the fundamental frequency and harmonic frequency after the ripple signal is restored.
[0092] Establish an active chip operating voltage-temperature coupling model:
[0093] The operating voltage of active chips (such as driver chips) is significantly affected by temperature. When the temperature rises, the voltage is prone to drift. This can be obtained by combining the modeling dataset and substituting it into a general model.
[0094]
[0095] in, This is the operating voltage of the core node of the circuit at temperature T. This is the reference operating voltage value for the core node of the circuit at T0=25℃; T0 is the preset standard temperature of the chip. The first-order temperature influence coefficient of the core node voltage in the circuit. The range was determined experimentally to be 1.5 × 10⁻⁶. -3 / ℃ to 3.5×10 -3 / ℃; This represents the second-order temperature influence coefficient of the voltage at the core node of the circuit. The range was calibrated experimentally to 5 × 10⁻⁶. -6 / ℃ 2 Up to 15×10 -6 / ℃ 2 ; and Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, experimental calibration was performed.
[0096] The absolute values of the first-order temperature influence coefficient and the second-order temperature influence coefficient of the core node voltage of the passive chip are smaller than the corresponding coefficient settings of the active chip, which is consistent with the characteristic that the temperature stability of passive devices is better than that of active devices.
[0097] Based on the on-resistance of active / passive chips and their temperature characteristics, a temperature coupling model for the on-resistance of active chips is established:
[0098] The on-resistance of an active chip increases non-linearly with increasing temperature. Substituting this into the general model, we get:
[0099]
[0100] in, This is the on-resistance of the core node of the circuit at temperature T. This is the reference value for the on-resistance of the core node of the circuit at T0=25℃; The first-order temperature influence coefficient of the resistance at the core node of the circuit was determined experimentally, and its range is 2 × 10⁻⁶. -3 / ℃ to 4×10 -3 / ℃; The second-order temperature influence coefficient of the resistance at the core node of the circuit was determined experimentally, and its range is 6 × 10⁻⁶. -6 / ℃ 2 Up to 18×10 -6 / ℃ 2 ; and Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, experimental calibration was performed.
[0101] Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, calculate the operating voltage of the core node at temperature T. and the on-resistance of the core node of the circuit at temperature T As a theoretical value ;
[0102] The least squares method is used to fit the actual value to the theoretical value, and the temperature influence coefficients α and β are solved. Combined with the drift correction of the temperature influence coefficients, the final true value of the electrical variable is obtained. The corrected formula is:
[0103]
[0104] Where Δα and Δβ are the coefficient drift caused by the ripple frequency, and ΔT = T - T0;
[0105] The actual operating voltage and on-resistance of the chip are obtained after restoration. Based on the difference between the actual and theoretical values, preliminary anomaly identification is performed.
[0106] In this embodiment, the actual value of the chip's operating voltage and the actual value of its on-resistance are used as the core basis for anomaly identification. If the difference between the actual value and the theoretical value is greater than a preset threshold, it is initially identified as an abnormal state. The preset threshold for the difference between the actual value and the theoretical value is set to 5% of the theoretical value.
[0107] In one embodiment of the present invention, a feature enhancement and multi-feature fusion identification algorithm is established to distinguish fault categories based on the true values of electrical variables and the anomaly identification results, while adapting to transient fault responses, including the following steps:
[0108] Establish feature enhancement and multi-feature fusion recognition algorithms to achieve accurate fault differentiation, balancing accuracy and verifiability, while adapting to transient fault response;
[0109] For similar fault current waveforms, deep features are extracted using wavelet packet transform, and a feature enhancement algorithm is established to enhance feature discriminative power. The formula is as follows:
[0110] Let the original waveform of the fault current be... Wavelet packet decomposition was performed with four levels, and the db6 wavelet basis was selected to obtain the wavelet packet coefficients of each node. Where j is the decomposition level, k is the node number, and n is the discrete-time variable;
[0111] Calculate the energy characteristics of wavelet packet coefficients at each node. As a core feature for fault differentiation:
[0112]
[0113] Simultaneously, the time-domain features of the current waveform are extracted, including: peak value. Rising slope , slope of falling edge Waveform distortion rate Construct multi-dimensional feature vectors ;
[0114] A multi-feature fusion identification algorithm is established using an improved SVM (Support Vector Machine) algorithm, incorporating fault physical mechanism rules. Elements from the multi-dimensional feature vector F are used as input samples to construct a fault identification model, as shown in the formula:
[0115] Objective function: ;
[0116] Constraints: , ;
[0117] Where w is the classification hyperplane normal vector, b is the bias term, C is the penalty coefficient (set to 10), xi is the i-th input sample, ξi is the slack variable of the i-th input sample, i∈(1,2,...,m), and D is the physical mechanism constraint coefficient, which is set to 5 in this embodiment. Multi-dimensional feature vector The model prediction value of the i-th input sample. Multi-dimensional feature vector The i-th input sample corresponds to the label value based on the physical mechanism, and m is the number of samples;
[0118] Fault categories are distinguished through a fault identification model, while also adapting to transient fault responses.
[0119] The fault identification model in this embodiment not only utilizes sample data to improve identification accuracy, but also avoids the "black box" problem through physical mechanism constraints, thus achieving accurate fault differentiation. At the same time, the model has a small computational load and can run in real time on local edge nodes (response time ≤ 1ms), adapting to transient fault response requirements.
[0120] In one embodiment of the present invention, based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively, including the following steps:
[0121] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of distinguishing fault categories, a ripple anomaly identification scheme is integrated to identify ripple anomalies.
[0122] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, an electrical variable anomaly identification algorithm is integrated to identify electrical variable anomalies.
[0123] Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant circuit control data and auxiliary monitoring data, a circuit electrical variable anomaly identification algorithm is established to identify circuit electrical variable anomalies.
[0124] In one embodiment of the present invention, identifying ripple anomalies includes the following steps:
[0125] Based on the amplitude and frequency components of the ripple signal, a ripple anomaly identification scheme is defined, including: amplitude criterion, frequency criterion, and co-criterion.
[0126] The amplitude criterion is: calculate the effective value of the ripple voltage after noise reduction. :
[0127] The formula for calculating the effective value of ripple voltage is: ,in, The ripple voltage is the effective value, and T is the sampling period; This is a real ripple signal;
[0128] like ( Set the power supply according to the rated parameters of the surgical equipment. For example, if the rated power supply parameter is ≤50mV, then... If set to 50mV, it is determined to be an abnormal ripple amplitude.
[0129] The frequency criterion is: if the ripple frequency components extracted by FFT contain harmonics that are not the inherent frequency of the power supply (such as high-frequency ripple harmonic frequency > 10MHz), or if the fundamental frequency drifts by more than ±5%, it is determined to be an abnormal ripple frequency.
[0130] The collaborative criterion is: if any of the above criteria is met and the duration is ≥100ms (to avoid misjudging transient fluctuations), it is judged as an abnormal ripple, which is associated with the risk of reduced power conversion efficiency;
[0131] The determination results of amplitude criteria, frequency criteria, and co-jurisdictional criteria of historical data are obtained. Combined with the processing results of electrical variable real value restoration and anomaly identification, as well as the fault category differentiation results and the ripple anomaly category corresponding to historical data, a DNN deep neural network is trained, and the ripple anomaly type is identified through the trained model.
[0132] In one embodiment of the present invention, identifying electrical variable anomalies includes the following steps:
[0133] Based on the real electrical variables after compensation and correction, combined with trend prediction, static anomaly identification and dynamic trend prediction are carried out.
[0134] Static anomaly identification includes: calculating the deviation rate between the corrected actual electrical variable and the baseline value at the corresponding temperature; the formula is:
[0135]
[0136] in, The deviation rate, This is the reference value of the electrical variable during the sampling period T. The corrected actual electrical variable; if the deviation rate between the corrected actual electrical variable and the reference value at the corresponding temperature is greater than the threshold (δ threshold is set to ±5%, which can be adjusted according to the chip type), it is judged as a static anomaly;
[0137] Dynamic trend prediction includes: using a linear regression algorithm to fit the trend of the corrected electrical variables, and predicting the trend of changes in the electrical variables within a preset time period. The formula is as follows:
[0138] Let the corrected electrical variable at time t be... t∈(1,2,...,H), where H is the number of sampling points within the sliding window. In this embodiment, n=50, and the fitted straight line equation is: ;
[0139] The formula for calculating the slope k is: b is the intercept;
[0140] If the absolute value of the slope k exceeds the preset rate threshold, it is judged as an abnormal dynamic trend, indicating that the electrical variable will deviate seriously.
[0141] The static anomaly identification and dynamic trend prediction results of historical data are obtained. The processing results of electrical variable real value restoration and anomaly identification, as well as the fault category differentiation results and the electrical variable anomaly categories corresponding to historical data, are combined to train a DNN deep neural network. The trained model is then used to identify the electrical variable anomaly type.
[0142] In one embodiment of the present invention, identifying abnormal circuit electrical variables includes the following steps:
[0143] Based on the normalized circuit loop electrical variables of the surgical equipment, the SVM (Support Vector Machine) algorithm is used to calculate the normalized voltage. Current ,power As input features, an SVM classification model is constructed, which is divided into normal, overvoltage, overcurrent, undervoltage, and poor contact.
[0144] SVM objective function: Constraints: , i∈(1,2,...,m);
[0145] Where w is the classification hyperplane normal vector, b is the bias term, C is the penalty coefficient (set to 10), ξi is the slack variable, yi is the sample label (1 for normal, -1 for abnormal), xi is the i-th input sample, and m is the number of samples; through the model output, the circuit abnormality type is identified, including: overvoltage, overcurrent, undervoltage, and poor contact abnormality.
[0146] In this embodiment, for the identified circuit anomalies, PID control is used to achieve precise adjustment of the circuit electrical variables; by adjusting the voltage regulating module in the distribution box, the circuit voltage is quickly brought back to the target value; for different anomaly types (such as overcurrent and undervoltage), the Kp, Ti, and Td parameters are adaptively adjusted to improve the adjustment effect; for poor contact anomalies, contact resistance monitoring and prompts are triggered simultaneously.
[0147] In one embodiment of the present invention, a weighted fusion algorithm is used for collaborative judgment of multiple anomalies, including the following steps:
[0148] If two or more anomalies occur simultaneously (such as ripple anomaly + loop overcurrent, temperature anomaly + chip current anomaly), a weighted fusion algorithm is used to calculate the anomaly comprehensive index, as shown in the formula:
[0149]
[0150] Wherein, EI is the comprehensive anomaly index, w1, w2, and w3 are the weights of ripple anomaly, temperature electrical variable anomaly, and loop anomaly, respectively. In this embodiment, the weights are set according to the severity of the anomaly: w1=0.3, w2=0.3, w3=0.4. e1, e2, and e3 are the anomaly values of the three types of anomalies, which are assigned by experts according to the severity of different anomaly types. In this embodiment, 0≤ei≤1, and the more severe the anomaly, the closer ei is to 1. If EI≥0.7, it is judged as a serious anomaly, and the highest level of warning and emergency response is immediately triggered.
[0151] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for real-time electrical variable monitoring and abnormal early warning of surgical equipment, characterized in that, Includes the following steps: Multimodal data acquisition nodes are deployed in the surgical equipment within the managed area to collect power ripple-related data, temperature adaptation-related data, circuit control-related data, and auxiliary monitoring data. The residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction. The ripple signal in the acquired data is restored and the ripple frequency components are extracted. The frequency of the denoised ripple signal is analyzed by fast Fourier transform. Based on the power ripple and temperature adaptation data of active and passive chips in the surgical equipment in the management area, and combined with the frequency analysis results of the ripple signal, a temperature-electrical parameter coupling model is established, and a temperature coupling model of chip operating voltage and on-resistance is established. The influence coefficient of temperature on electrical variables is analyzed, and the true value of electrical variables is restored and anomalies are identified. Establish feature enhancement and multi-feature fusion recognition algorithms to distinguish fault categories based on the true values of electrical variables and anomaly recognition results, while adapting to transient fault responses; Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively. If two or more types of anomalies are identified at the same time, a weighted fusion algorithm is used to perform multi-anomaly collaborative judgment.
2. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, The residual noise in the acquired data is suppressed by combining hardware filtering with software dual-algorithm noise reduction. The ripple signal in the acquired data is restored, and the ripple frequency components are extracted. The frequency analysis of the denoised ripple signal is performed by Fast Fourier Transform, including the following steps: Hardware filtering uses a second-order active low-pass filter combined with a notch filter to initially filter out high-frequency noise and power frequency interference. The software dual-algorithm denoising uses wavelet threshold denoising algorithm to suppress residual noise and restore the ripple signal in the power supply ripple related data; Let the ripple signal at time t in the collected power supply ripple correlation data be... The noise signal is The actual ripple signal is ,but ;right Wavelet decomposition was performed with 5 levels, using the db4 wavelet basis. The low-frequency components were obtained after decomposition. With high frequency components ; High-frequency components are processed using an adaptive threshold function to suppress noise. in, The threshold value for the high-frequency components of the j-th layer is... Here, N is the noise estimate, and N is the length of the discrete signal. Perform inverse wavelet transform on the low-frequency and high-frequency components of the restored ripple signal to obtain the denoised true ripple signal. ; right Discretization is performed to obtain discrete signals. The frequency components of the ripple signal after noise reduction are extracted by using Fast Fourier Transform (FFT), thus obtaining the frequency domain characteristics of the discrete signal after FFT transformation. ; The spectral amplitude of the frequency domain characteristics after the discrete signal undergoes FFT transformation is calculated. Extract the fundamental frequency and harmonic frequencies of the ripple signal.
3. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, Based on the power ripple and temperature adaptation data of active and passive chips in the surgical equipment within the management area, and combined with the ripple signal frequency analysis results, a temperature-electrical parameter coupling model is established, including the following steps: Based on temperature adaptation data of active and passive chips in surgical equipment within the managed area, baseline electrical parameter values for active and passive chips at different temperatures were obtained. Given the non-linear relationship between the electrical parameters of surgical equipment chips and temperature, and combining the fundamental and harmonic frequencies obtained from ripple signal reconstruction, a temperature-electrical parameter coupling model was established. The formula is as follows: Where X(T) is the electrical parameter of the chip at temperature T; X0 is the room temperature reference value, that is, the measured value of the electrical parameter when T=T0; T is the real-time temperature of the chip core node, T∈[-10, 85], and T0 is the preset standard temperature of the chip; α is the first-order temperature influence coefficient, reflecting the degree of linear influence of temperature on electrical parameters; β is the second-order temperature influence coefficient, reflecting the degree of nonlinear influence of temperature on electrical parameters.
4. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, A temperature coupling model of chip operating voltage and on-resistance is established to analyze the influence coefficient of temperature on electrical variables, and to restore the true values of electrical variables and identify anomalies, including the following steps: Based on the temperature adaptation data of the active and passive chips corresponding to the core nodes of the surgical equipment circuit in the management area, the reference values of the electrical parameters of the active and passive chips at different temperatures are obtained, combined with the fundamental frequency and harmonic frequency after the ripple signal is restored. Establish an active chip operating voltage-temperature coupling model: in, This is the operating voltage of the core node of the circuit at temperature T. This is the reference operating voltage value for the core node of the circuit at T0=25℃; T0 is the preset standard temperature of the chip. The first-order temperature influence coefficient of the core node voltage in the circuit; The second-order temperature influence coefficient of the voltage at the core node of the circuit; and Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, experimental calibration was performed. Based on the on-resistance of active / passive chips and their temperature characteristics, a temperature coupling model for the on-resistance of active chips is established: The on-resistance of an active chip increases non-linearly with increasing temperature. Substituting this into the general model, we get: in, This is the on-resistance of the core node of the circuit at temperature T. This is the reference value for the on-resistance of the core node of the circuit at T0=25℃; This is the first-order temperature influence coefficient of the resistance at the core node of the circuit. The second-order temperature influence coefficient of the resistance at the core node of the circuit; and Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, experimental calibration was performed. Based on the fact that the core nodes of the surgical equipment circuit correspond to active and passive chips, calculate the operating voltage of the core node at temperature T. and the on-resistance of the core node of the circuit at temperature T As a theoretical value ; The least squares method is used to fit the actual value to the theoretical value, and the temperature influence coefficients α and β are solved. Combined with the drift correction of the temperature influence coefficients, the final true value of the electrical variable is obtained. The corrected formula is: Where Δα and Δβ are the coefficient drift caused by the ripple frequency, and ΔT = T - T0; The actual operating voltage and on-resistance of the chip are obtained after restoration. Based on the difference between the actual and theoretical values, preliminary anomaly identification is performed.
5. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, A feature enhancement and multi-feature fusion identification algorithm is established to distinguish fault categories based on the true values of electrical variables and anomaly identification results, while adapting to transient fault responses. The algorithm includes the following steps: For similar fault current waveforms, deep features are extracted through wavelet packet transform, and a feature enhancement algorithm is established to enhance feature discriminative power. Let the original waveform of the fault current be... Wavelet packet decomposition was performed with four levels, and the db6 wavelet basis was selected to obtain the wavelet packet coefficients of each node. Where j is the decomposition level, k is the node number, and n is the discrete-time variable; Calculate the energy characteristics of wavelet packet coefficients at each node. As a core feature for fault differentiation: Simultaneously, the time-domain features of the current waveform are extracted, including: peak value. Rising slope , slope of falling edge Waveform distortion rate Construct multi-dimensional feature vectors ; A multi-feature fusion identification algorithm is established using an improved SVM (Support Vector Machine) algorithm, incorporating fault physical mechanism rules. Elements from the multi-dimensional feature vector F are used as input samples to construct a fault identification model, as shown in the formula: Objective function: ; Constraints: , ; Where w is the classification hyperplane normal vector, b is the bias term, C is the penalty coefficient, xi is the i-th input sample, ξi is the slack variable of the i-th input sample, i∈(1,2,...,m), and D is the physical mechanism constraint coefficient. Multi-dimensional feature vector The model prediction value of the i-th input sample. Multi-dimensional feature vector The i-th input sample corresponds to the label value based on the physical mechanism, and m is the number of samples; Fault categories are distinguished through a fault identification model, while also adapting to transient fault responses.
6. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, Based on the processing results of restoring the true values of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant loop control data and auxiliary monitoring data, ripple anomalies, electrical variable anomalies, and loop electrical variable anomalies are identified respectively, including the following steps: Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of distinguishing fault categories, a ripple anomaly identification scheme is integrated to identify ripple anomalies. Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, an electrical variable anomaly identification algorithm is integrated to identify electrical variable anomalies. Based on the processing results of restoring the true value of electrical variables and identifying anomalies, as well as the results of fault category differentiation, and combined with relevant circuit control data and auxiliary monitoring data, a circuit electrical variable anomaly identification algorithm is established to identify circuit electrical variable anomalies.
7. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 6, characterized in that, Identifying ripple anomalies includes the following steps: Based on the amplitude and frequency components of the ripple signal, a ripple anomaly identification scheme is defined, including: amplitude criterion, frequency criterion, and co-criterion. The amplitude criterion is: calculate the effective value of the ripple voltage after noise reduction. : The formula for calculating the effective value of ripple voltage is: ,in, The ripple voltage is the effective value, and T is the sampling period; This is a real ripple signal; like The ripple amplitude was determined to be abnormal. The frequency criterion is: if the ripple frequency components extracted by FFT contain harmonics that are not the inherent frequency of the power supply, or if the fundamental frequency drifts by more than ±5%, it is determined to be an abnormal ripple frequency. The collaborative criterion is: if any of the above criteria is met and the duration is ≥100ms, it is judged as an abnormal ripple, which is associated with the risk of reduced power conversion efficiency; The determination results of amplitude criteria, frequency criteria, and co-jurisdictional criteria of historical data are obtained. Combined with the processing results of electrical variable real value restoration and anomaly identification, as well as the fault category differentiation results and the ripple anomaly category corresponding to historical data, a DNN deep neural network is trained, and the ripple anomaly type is identified through the trained model.
8. The method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 6, characterized in that, Identifying electrical variable anomalies includes the following steps: Based on the real electrical variables after compensation and correction, combined with trend prediction, static anomaly identification and dynamic trend prediction are carried out. Static anomaly identification includes: calculating the deviation rate between the corrected actual electrical variable and the baseline value at the corresponding temperature; if the deviation rate between the corrected actual electrical variable and the baseline value at the corresponding temperature is greater than a threshold, it is determined to be a static anomaly; Dynamic trend prediction includes: using a linear regression algorithm to fit the trend of the corrected electrical variable and predicting the trend of the electrical variable's change over a preset time period; let the corrected electrical variable at time t be... t∈(1,2,...,H), where H is the number of sampling points within the sliding window, and the fitted line equation is: The formula for calculating the slope k is: b is the intercept; If the absolute value of the slope k exceeds the preset rate threshold, it is judged as an abnormal dynamic trend, indicating that the electrical variable will deviate seriously. The static anomaly identification and dynamic trend prediction results of historical data are obtained. The processing results of the restoration of the true value of electrical variables and anomaly identification, as well as the fault category differentiation results, and the corresponding electrical variable anomaly categories in historical data are used to train a DNN deep neural network. The trained model is then used to identify the anomaly type of electrical variables.
9. A method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 6, characterized in that, Identifying abnormal circuit electrical variables includes the following steps: Based on the normalized circuit loop electrical variables of the surgical equipment, the SVM (Support Vector Machine) algorithm is used to calculate the normalized voltage. Current ,power As input features, an SVM classification model is constructed, which is divided into normal, overvoltage, overcurrent, undervoltage, and poor contact. SVM objective function: Constraints: , i∈(1,2,...,m); Where w is the classification hyperplane normal vector, b is the bias term, C is the penalty coefficient (set to 10), ξi is the slack variable, yi is the sample label, xi is the i-th input sample, and m is the number of samples; through the model output, the circuit anomaly type is identified, including: overvoltage, overcurrent, undervoltage, and poor contact anomaly.
10. A method for real-time electrical variable monitoring and abnormal early warning of surgical equipment according to claim 1, characterized in that, The weighted fusion algorithm is used for collaborative judgment of multiple anomalies, including the following steps: If two or more types of anomalies occur simultaneously, a weighted fusion algorithm is used to calculate the anomaly composite index, as shown in the formula: Among them, EI is the comprehensive anomaly index, w1, w2, and w3 are the weights of ripple anomaly, temperature electrical variable anomaly, and loop anomaly, respectively, and e1, e2, and e3 are the anomaly values of the three types of anomalies. The anomaly values of the three types of anomalies are assigned by experts according to the severity of different anomaly types. If EI ≥ 0.7, it is judged as a serious anomaly, and the highest level of warning and emergency response is immediately triggered.
