Operation power system prediction maintenance method and system based on digital twinning

Abstract

The invention relates to the technical field of medical equipment maintenance, in particular to an operation power system prediction maintenance method and system based on digital twinning, and the method comprises the steps: carrying out the digital modeling of an operation power system, and forming a digital twinning system; acquiring real-time state information from three channels of a power supply, a surgical instrument and a cooling system; constructing a heterogeneous data characterization mechanism based on a matrix manifold, and mapping multichannel data to a unified feature space; establishing a multi-channel fusion algorithm of adaptive feature weight optimization; constructing an aging degree degradation model; outputting the aging degree of the system; according to the method, the problem of multi-source heterogeneous data fusion is innovatively solved, the prediction precision and the maintenance efficiency are improved, accurate evaluation of the health state of the operation power system, accurate prediction of performance degradation and intelligent optimization of maintenance decision are achieved, and the equipment reliability and the operation safety are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of medical equipment maintenance technology, specifically to a predictive maintenance method and system for surgical power systems based on digital twins, used to achieve health status monitoring, degradation trend analysis, and predictive maintenance decision-making for surgical power systems. Background Technology

[0002] The surgical power system is one of the core pieces of equipment in modern surgery, primarily used to provide precise and stable power output for various surgical instruments. This system typically consists of a main unit, a control handle, and a foot switch. The main unit drives the control handle, which is then controlled by the foot switch. The control handle connects to an adapter blade to remove soft and hard tissues and bone in ear, nose, and throat (otology, rhinology, sinus surgery, nasopharynx / larynx), head and neck surgery, or other surgical procedures.

[0003] Modern surgical power systems typically have multiple operating modes, such as forward cutting, reverse cutting, and reciprocating cutting, and are equipped with a flushing system pump. Key performance parameters of the system include rotational speed accuracy (no-load and load speed errors), mechanical stability (such as radial runout and axial movement), and adjustable rotational speed range. These parameters directly affect the precision, safety, and efficiency of the surgery.

[0004] Currently, the maintenance of surgical power systems is mainly based on three methods: scheduled maintenance, post-failure repair, and simple condition monitoring. Scheduled maintenance involves inspecting and maintaining the equipment at fixed intervals, often without considering the actual usage and aging of the equipment, which can easily lead to unnecessary maintenance costs or failure to detect potential faults in a timely manner. Post-failure repair involves fixing the equipment after it has already failed, which may result in surgical interruption or even endanger patient safety. Simple condition monitoring can only provide limited equipment status information and lacks the ability to predict future performance degradation.

[0005] Surgical power systems involve the coordinated operation of multiple subsystems, including power supply, drive, and irrigation pumps. The systems are highly complex, with intricate interactions between components. Traditional maintenance methods struggle to fully assess the system's health, especially for surgical power systems requiring high precision (such as radial runout ≤0.10mm and axial runout no greater than 0.5mm). Even slight performance degradation can impact surgical outcomes.

[0006] With the development of medical equipment technology, traditional maintenance methods can no longer meet the maintenance needs of modern surgical power systems. Digital twin technology, as an emerging virtual-physical mapping technology, enables real-time monitoring, analysis, and prediction by establishing digital models of physical entities, providing a new technical approach for predictive maintenance of equipment. However, existing digital twin applications still face key technical challenges when dealing with complex medical equipment such as surgical power systems, including difficulties in fusing multi-source heterogeneous data, insufficient accuracy of predictive models, and inaccurate maintenance decisions. Summary of the Invention

[0007] In view of the shortcomings of the prior art, the present invention provides a predictive maintenance method and system for surgical power systems based on digital twins, which aims to achieve accurate assessment of the health status of surgical power systems, accurate prediction of performance degradation, and intelligent optimization of maintenance decisions, thereby improving equipment reliability, reducing maintenance costs, and ensuring surgical safety.

[0008] This invention proposes a method and system for predictive maintenance of surgical dynamic systems based on digital twins, comprising: A digital model of the surgical power system is created, and a physical model of the surgical power system is established. By collecting information from surgical power sensors, the physical model of the surgical power system is mapped to form a mapping-driven digital twin system of surgical power. Real-time status information of the surgical power system is collected from the power supply channel, surgical instrument channel, and cooling system channel; A heterogeneous data representation mechanism based on matrix manifold is constructed to map the real-time status information of the power supply channel, surgical instrument channel and cooling system channel to a unified feature space to form feature representations; An adaptive feature weight optimization multi-channel fusion algorithm is established to dynamically fuse the feature representations to obtain fused features; Based on the aforementioned fusion features, an aging and degradation model of the surgical power system is constructed. The real-time status information of the surgical power system is input into the aging and degradation model of the surgical power system, and the aging degree of the surgical power system is output. A predictive maintenance plan is generated based on the aging degree of the surgical power system.

[0009] Preferably, the digital modeling of the surgical power system includes: Establish a static geometric model that includes the power supply unit, transmission mechanism, and cooling system; Construct a multi-physics domain dynamic behavior model encompassing the mechanical, thermal, and electrical domains; Design a sensor deployment strategy and establish a real-time data synchronization channel; The real-time data synchronization channel enables bidirectional interaction between the physical system and the digital model, supporting automatic parameter calibration.

[0010] Preferably, the acquisition of real-time status information of the surgical power system from the power supply channel, surgical instrument channel, and cooling system channel includes: Voltage sensors, current sensors, and power sensors are deployed in the power supply channel to collect power status data; Vibration sensors, acoustic sensors, and temperature sensors are installed in the surgical instrument channel to collect status data of mechanical components; Flow sensors, temperature sensors, and pressure sensors are deployed in the cooling system channels to collect cooling efficiency data; The power status data, mechanical component status data, and cooling efficiency data are preprocessed, including signal noise reduction, time synchronization, and data standardization.

[0011] Preferably, the mechanism for constructing heterogeneous data representation based on matrix manifolds includes: Construct a power channel feature matrix, a device channel feature matrix, and a cooling system feature matrix. The dimensions of the feature matrix include a time dimension, a channel index dimension, and a feature index dimension. Design a nonlinear mapping function to project the feature matrix onto a shared feature space; Construct a geodesic distance metric to quantify the similarity of data points in the feature space; Multi-scale feature extraction is achieved, extracting time-domain features, frequency-domain features, and joint time-frequency-domain features respectively.

[0012] Preferably, the multi-channel fusion algorithm for establishing adaptive feature weight optimization includes: Construct an information gain matrix to quantify the contribution of each channel's data to system state prediction; Establish a feature-fault correlation matrix to quantify the ability of features to identify different types of faults; Identify the operating modes of the surgical power system and design specific weight allocation strategies for different operating modes; Achieve a multi-level fusion mechanism, including feature-level fusion, decision-level fusion, and spatiotemporal scale fusion.

[0013] Preferably, the aging degradation model of the surgical power system is a data-driven degradation model, including: Design multidimensional health indicators to comprehensively assess the system's status; Construct a degradation trend matrix to capture the aging patterns of components; To perform nonlinear degradation path analysis and identify triggering conditions that accelerate degradation; Establish a life prediction model based on degradation trends to dynamically estimate the remaining life of equipment.

[0014] Preferably, the aging-based predictive maintenance scheme for the surgical power system includes: Construct a risk-based maintenance timing decision framework; Design a cost-benefit balance model to optimize maintenance economy; Establish a maintenance solution library for different failure modes; To enable maintenance window identification and determine the optimal maintenance time based on equipment usage plans; Generate a maintenance plan that includes maintenance recommendations, optimal maintenance time, maintenance cycle, and failure probability.

[0015] As a preferred option, further steps include maintenance effectiveness evaluation and continuous system optimization. Establish a pre- and post-maintenance comparative analysis framework to evaluate the maintenance effectiveness; Design knowledge base update rules and accumulate maintenance experience; Construct a model correction mechanism based on maintenance results to continuously optimize prediction accuracy; Implement a case study learning framework to extract best practices from historical maintenance.

[0016] Preferably, the heterogeneous data representation mechanism based on matrix manifold further includes: Build a data uncertainty model to address sensor noise and data loss issues; Design a data completion strategy based on manifold structure to infer missing values ​​using local geometric information; Construct a low-rank matrix factorization model to capture and reconstruct the intrinsic structure of the data; To achieve feature quality assessment and enhancement, including feature reliability assessment, feature enhancement, and feature compensation.

[0017] A digital twin-based surgical power system prediction and maintenance system includes: The digital twin modeling module is used to digitally model the surgical power system, establish a physical model of the surgical power system, and form a digital twin system through sensor information mapping. The multi-source data acquisition module is used to acquire real-time status information of the surgical power system from the power channel, surgical instrument channel and cooling system channel; The matrix manifold representation module is used to construct a heterogeneous data representation mechanism, mapping multi-source channel data to a unified feature space; The adaptive fusion module is used to establish a multi-channel fusion algorithm with optimized feature weights to dynamically fuse feature representations. A degradation model construction module is used to construct an aging degradation model of the surgical power system based on fusion features; The health status assessment module is used to input real-time status information into the degradation model and output the system aging degree. The predictive maintenance decision module is used to generate predictive maintenance plans based on the system's aging status; and The closed-loop optimization module is used to evaluate the maintenance effect and continuously optimize the prediction model and maintenance strategy.

[0018] The main beneficial effects of this invention include: 1. By establishing a digital twin model of the surgical power system, precise mapping and two-way interaction between physical equipment and digital models can be achieved, providing comprehensive data support for health status assessment. For surgical power systems with multiple operating modes (forward cutting, reverse cutting, and reciprocating cutting), the digital twin model can comprehensively simulate the system behavior under different operating modes, ensuring accurate assessment of the system status under various operating conditions (such as different speed ranges and different load conditions).

[0019] 2. A heterogeneous data representation mechanism based on matrix manifold theory solves the challenge of fusing heterogeneous data from multiple sources, such as power supply, equipment, and cooling systems, improving data utilization efficiency and representation quality. This is particularly important for complex equipment containing multiple subsystems, including the main unit, handle, foot switch, and flushing system pump, as it effectively integrates monitoring data from different subsystems.

[0020] 3. An adaptive feature weight optimization multi-channel fusion algorithm is adopted to dynamically adjust the feature importance under different working modes, thereby improving the accuracy and robustness of system state prediction. In particular, for equipment state prediction under three different working modes (forward cutting, reverse cutting, and reciprocating cutting), the algorithm can adaptively adjust the prediction strategy according to the characteristics of different modes.

[0021] 4. Construct a data-driven aging degradation model to accurately predict equipment performance degradation trends and provide a scientific basis for maintenance decisions. This is particularly important for maintaining key performance indicators of the surgical power system (such as rotational speed accuracy, radial runout, and axial runout) within specified ranges.

[0022] 5. An intelligent maintenance decision-making mechanism based on risk assessment and cost-benefit analysis optimizes maintenance timing and strategies, reducing maintenance costs while ensuring equipment reliability. This is of significant value to medical institutions in optimizing equipment management, reducing unnecessary maintenance downtime, and extending equipment lifespan.

[0023] 6. Establish a maintenance effectiveness evaluation and knowledge accumulation mechanism to achieve continuous system optimization and adaptive adjustment, thereby improving long-term predictive maintenance effectiveness. By learning from historical maintenance experience, the system can continuously improve its predictive maintenance capabilities for complex surgical power systems.

[0024] In summary, the predictive maintenance method and system provided by this invention can significantly improve the reliability and safety of surgical power systems, ensuring high-precision operation under various working modes (such as the technical requirements of ±3% no-load speed error and ±5% load speed error), extending equipment lifespan, and reducing maintenance costs. This is of great significance for improving the quality of medical services and surgical safety. Attached Figure Description

[0025] Figure 1 This is a flowchart of the predictive maintenance method for surgical power systems based on digital twins, as described in this invention. Detailed Implementation

[0026] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Reference Figure 1 The present invention provides a predictive maintenance method for surgical dynamic systems based on digital twins, which mainly includes the following steps: digitally modeling the surgical dynamic system, collecting multi-source channel data, constructing a matrix manifold representation mechanism, realizing adaptive feature fusion, establishing a degradation model, evaluating the system state, and generating a predictive maintenance scheme. Each step will be described in detail below.

[0028] In a preferred embodiment of the present invention, digital modeling of the surgical power system refers to establishing a physical model of the surgical power system, mapping the physical model of the surgical power system by collecting information from surgical power sensors, and forming a mapped-driven digital twin system of surgical power. Specifically, this step includes the following: First, a static geometric model is established, including the power supply unit, transmission mechanism, and cooling system. This model contains basic information such as the geometric dimensions, topology, and material properties of the surgical power system. For example, the geometric dimensions, electrical connection methods, and material properties of the power supply unit; the bearing structure, gear configuration, and transmission ratio of the transmission mechanism; and the piping layout and radiator parameters of the cooling system.

[0029] Secondly, a multi-physics domain dynamic behavior model is constructed, encompassing the mechanical, thermal, and electrical domains. The mechanical domain model describes the motion characteristics of the transmission components, including kinematic equations, dynamic properties, and vibration response; the thermal domain model simulates the internal temperature field distribution and heat conduction process of the equipment; and the electrical domain model characterizes the electrical performance and response characteristics of the power supply and control system. These multi-physics domain models are interconnected through coupling equations, forming a complete dynamic system description.

[0030] Secondly, design a sensor deployment strategy and establish a real-time data synchronization channel. Based on the key points and potential fault locations of the system, rationally deploy various types of sensors and establish data acquisition and transmission channels to ensure that the physical equipment status can be mapped to the digital model in real time.

[0031] Finally, a real-time data synchronization channel enables bidirectional interaction between the physical system and the digital model, supporting automatic parameter calibration. Specifically, by comparing the differences between sensor measurements and model predictions, model parameters are dynamically adjusted to ensure consistency between the digital twin model and the actual device state. The calibration process can employ the following optimization objective function: , in: To optimize the objective function, let represent the overall error of the model parameters; This is the model parameter vector, containing all the parameters that need to be calibrated; These are the actual measured values ​​from the sensors, representing the actual state quantities of the physical equipment. These are model predictions, representing values ​​based on parameters. The calculated model output; N is the number of data points, representing the number of samples used for calibration; This is the regularization coefficient, typically ranging from 0.01 to -0.1, used to prevent overfitting; The squared L2 norm of the parameter vector represents a penalty term for parameter size. The first term of this function is the mean squared error, which measures the difference between the model's predicted values ​​and the actual measured values; the second term is a regularization term, which prevents overfitting due to excessively large parameters.

[0032] This invention collects real-time status information from three key channels of the surgical power system: the power supply channel, the surgical instrument channel, and the cooling system channel. Specific implementation details are as follows: Voltage, current, and power sensors are deployed in the power supply channel to collect power status data. The voltage sensor has a sampling frequency of 1kHz, a range of 0-250V, and an accuracy of ±0.1%; the current sensor also has a sampling frequency of 1kHz, a range of 0-30A, and an accuracy of ±0.2%; the power sensor calculates instantaneous power and cumulative power consumption from the voltage and current signals.

[0033] Vibration sensors, acoustic sensors, and temperature sensors are arranged in the surgical instrument channel to collect status data of mechanical components. The vibration sensor has a sampling frequency of 10 kHz to capture high-frequency mechanical fault characteristics; the acoustic sensor has a frequency range of 20 Hz-20 kHz to monitor abnormal noise; and the temperature sensor has a sampling frequency of 1 Hz to monitor temperature changes at bearings and joints.

[0034] Flow sensors, temperature sensors, and pressure sensors are deployed in the cooling system channels to collect cooling efficiency data. The flow sensor monitors the coolant flow rate at a sampling frequency of 10Hz; the temperature sensor monitors the temperature difference between the coolant inlet and outlet at a sampling frequency of 1Hz; and the pressure sensor monitors system pressure fluctuations at a sampling frequency of 100Hz.

[0035] The collected power status data, mechanical component status data, and cooling efficiency data are preprocessed, including signal denoising, time synchronization, and data standardization. Signal denoising employs wavelet transform to remove high-frequency noise while preserving signal characteristics; time synchronization uses a unified timestamp mechanism to ensure data from different sampling frequencies are aligned; and data standardization uses Z-score normalization to convert data from different physical quantities to a uniform scale. , in: The data is standardized and dimensionless. This is raw data; the units depend on the specific physical quantity. This is the mean of the data series, in units of 1 and 2. same; For standard deviation, units are... The same. This standardization method is suitable for approximately normally distributed data. For non-normally distributed data, quantile standardization can be used. After standardization, the data of different physical quantities are converted into dimensionless data with a mean of 0 and a standard deviation of 1, which facilitates subsequent processing and comparison.

[0036] After acquiring multi-source channel data, this invention constructs a heterogeneous data representation mechanism based on a matrix manifold, mapping heterogeneous data from different channels to a unified feature space. The specific implementation of this mechanism is as follows: First, construct the feature matrices for the three channels. Power channel feature matrix. This represents the nth feature data of channel index i at time t; instrument channel feature matrix. and cooling system channel feature matrix They have the same structure. The typical dimensions of these feature matrices are: time dimension. (Representing 1000 time points), channel index dimension (Indicates the number of sub-channels per channel), feature index dimension (This represents the number of features in each sub-channel).

[0037] Secondly, a nonlinear mapping function is designed to project the feature matrix onto a shared feature space. This mapping function can be expressed as: , in: The input vector is the reshaped feature representation with the same dimension as the number of neurons in the output layer; X is the original feature matrix; W is the weight matrix with the same dimension as the number of neurons in the output layer. The input vector dimension; b is the bias vector, and its dimension is equal to the number of neurons in the output layer. For non-linear activation functions, the ReLU function is preferred in this embodiment: Where x is the input value, when Output x when The output is 0. The mapping process is performed separately for each of the three channels to obtain the mapped feature representation. , and This nonlinear mapping can capture complex patterns and nonlinear relationships between features in the data.

[0038] Then, a geodesic distance metric is constructed to quantify the similarity of data points in the feature space. Geodesic distance considers the nonlinear structure of the feature space and reflects the intrinsic relationships of the data better than Euclidean distance. For two points in the feature space... and The geodesic distance can be calculated using the following steps: 1) Construct a k-nearest neighbor graph (k is usually 5-10), and connect each point to its k nearest neighbors; 2) Calculate the Euclidean distance between adjacent points as the edge weight; 3) Use Dijkstra's algorithm to calculate the shortest path length between any two points on the map as the geodesic distance.

[0039] Finally, multi-scale feature extraction is implemented, extracting time-domain features, frequency-domain features, and joint time-frequency-domain features respectively. Time-domain features include statistical moments (mean, standard deviation, skewness, kurtosis, etc.) and trend features; frequency-domain features are extracted using Fast Fourier Transform (FFT) to obtain spectral features; and joint time-frequency-domain features are obtained using Continuous Wavelet Transform (CWT). Specifically, for the signal... Its continuous wavelet transform can be expressed as: , in: Here, represents the wavelet coefficients, indicating the time-frequency characteristics of the signal at scale a and translation b; a is the scale parameter, controlling the width of the wavelet, which is inversely proportional to the frequency; b is the translation parameter, controlling the position of the wavelet, corresponding to time. The input signal is a one-dimensional data that varies over time. is the conjugate of the wavelet basis functions, describing the wavelet shape used to analyze the signal; t is the time variable; the integration operation represents the inner product of the signal and the scaled and translated wavelet basis functions, used to measure the energy of the signal at a specific time and frequency. In practical applications, the Morlet wavelet is preferred as the basis function, and its expression is: , in: The Morlet wavelet function is a complex-valued function; t is the time variable. Pi; As a normalization factor, ensure that the wavelet energy is 1; The Gaussian envelope controls the temporal locality of the wavelet; Let i be a complex exponential function, where i is the imaginary unit. The center frequency, typically 5-6, determines the frequency domain resolution of the wavelet. The Morlet wavelet combines the characteristics of a Gaussian window and a complex exponential function, possessing both good time-domain locality and a well-defined center frequency, making it particularly suitable for analyzing transient features in vibration signals.

[0040] Multi-scale feature extraction can yield a comprehensive feature representation that reflects the equipment status, laying the foundation for subsequent fusion and analysis.

[0041] This invention establishes a multi-channel fusion algorithm with adaptive feature weight optimization to achieve dynamic fusion of features from different channels. The specific implementation is as follows: First, an information gain matrix M is constructed to quantify the contribution of each channel's data to system state prediction. For features... Its information gain can be expressed as: , in: Features Information gain, measured in bits, measures the amount of information provided by a feature. The entropy of system state Y measures the uncertainty of the state; Known features The conditional entropy of Y under given conditions represents the conditional entropy of Y given features. The remaining uncertainty after the state. The formula for calculating entropy is: , in: Let Y be the entropy of the random variable Y, expressed in bits; Y is the set of system states; y is a specific state within the set of states. Let y be the probability distribution of state y, with a range of values. ; This is a logarithmic function to the base 2. This formula calculates the average uncertainty of the state distribution; the more uniform the probability distribution, the greater the entropy. Based on the information gain calculation results, an information gain matrix M is constructed, whose elements... This represents the information gain value of the j-th feature in the i-th channel.

[0042] Secondly, a feature-fault correlation matrix is ​​established to quantify the ability of features to identify different types of faults. For features... and fault type The degree of correlation can be calculated using mutual information: , in: Features and fault types Mutual information, measured in bits, measures the statistical correlation between two entities. This is a joint probability distribution, representing the probability that the characteristic value and the fault type occur simultaneously; and These represent the marginal probability distributions, indicating the independent occurrence probabilities of the feature value and the fault type. Mutual information measures the amount of information shared between two variables; a higher value indicates a stronger ability of the feature to identify that type of fault.

[0043] Secondly, identify the operating modes of the surgical power system and design specific weight allocation strategies for different modes. Common operating modes include no-load (0%–20% load), light-load (20%–50% load), medium-load (50%–80% load), and heavy-load (80%–100% load). The weight allocation strategy differs for each operating mode. For example, under heavy-load conditions, the weight of the mechanical and cooling system channels should be relatively increased, as these components are more prone to problems under high loads; while under no-load conditions, the weight of the power supply channels can be appropriately increased to monitor electrical performance in standby mode.

[0044] Finally, a multi-level fusion mechanism is implemented, including feature-level fusion, decision-level fusion, and spatiotemporal scale fusion. Feature-level fusion adopts a weighted summation method: , Where: F is the fusion feature, with the same dimension as the individual feature vector; The i-th original feature vector; For the corresponding weights, the range of values ​​is... and ; n represents the total number of features. This fusion method is performed in the feature space, directly integrating information from different sources. Decision-level fusion integrates the results of multiple prediction models, using methods such as voting, averaging, or stacking to obtain the final prediction. Spatiotemporal scale fusion integrates prediction results from different time scales (short-term, medium-term, long-term) and spatial scales (component-level, system-level) to form a comprehensive system state representation.

[0045] This invention constructs an aging degradation model of a surgical power system based on fusion features. Preferably, this model is a data-driven degradation model, and the specific implementation is as follows: First, a multi-dimensional health index is designed to comprehensively assess the system status. These indexes include performance indicators (such as power output stability and response time), reliability indicators (such as interval between failures and failure rate), and efficiency indicators (such as energy consumption level and thermal efficiency). Each index is assigned a different weight, and the combined results are used to calculate the system health index (HI). , Where: HI is the system health index, dimensionless, with a range of values. A higher value indicates a better health condition; Let be the i-th standardized health indicator, dimensionless, with a range of values. ; For the corresponding weights, the range of values ​​is... and ; This represents the total number of indicators. In practical applications, the weights of the indicators can be determined based on expert experience or historical data analysis. For example, power stability has a weight of 0.3, response time has a weight of 0.25, failure rate has a weight of 0.25, and energy consumption level has a weight of 0.2.

[0046] Secondly, a degradation trend matrix is ​​constructed to capture the aging patterns of components. The dimensions of the degradation trend matrix D are... , where p is the number of components and q is the time step. Matrix elements This represents the health index of the i-th component at time step j, with a value range of... By analyzing the time evolution characteristics of this matrix, the degradation patterns and key inflection points of each component can be identified.

[0047] Then, nonlinear degradation path analysis is performed to identify the triggering conditions that accelerate degradation. Traditional linear degradation models often fail to accurately characterize the complex degradation process of actual devices; therefore, this invention employs a nonlinear model to describe the degradation path: , in: The health index at time t is dimensionless and has a range of values. ; The initial health index is usually set to 1; t is the running time in hours. Control the rate of degradation; is a dimensionless shape parameter that controls the shape of the degradation curve. At that time, the degradation exhibits an exponential decay; when At that time, the degradation trend accelerates; when Over time, the degradation rate gradually slows down. Based on historical data, parameter values ​​for different components can be estimated; for example, typical parameters for motor bearings are... , Typical parameters of the cooling system are: , .

[0048] Finally, a life prediction model based on degradation trends is established to dynamically estimate the remaining useful life of the equipment. The formula for calculating the remaining useful life (RUL) is: , Where RUL stands for Remaining Life, in hours, indicating the time the equipment can continue to operate safely. Current time, in hours; For the future Predicted health index; The health index threshold, typically set at 0.6-0.7, indicates the critical state at which equipment requires maintenance. This represents the infimum of the set, i.e., the minimum t-value that satisfies the condition. The formula calculates the time required for the health index to drop to the threshold, providing a time reference for maintenance decisions.

[0049] This invention generates predictive maintenance plans based on system aging assessment results. The specific implementation method is as follows: First, a risk-based maintenance timing decision framework is constructed. This framework considers two key factors: equipment failure risk and maintenance cost. Equipment failure risk The relationship with health index is non-linear: , in: For equipment failure risk, dimensionless, range of values. , representing the probability of equipment failure; k is a proportionality coefficient, usually set to 1, used to adjust the risk range; A health index, dimensionless, with a range of values. n is a risk sensitivity parameter, typically taken as 2-3, to control the non-linear relationship between health index and risk. When... At this point, the risk increases rapidly as the health index declines, which aligns with the actual failure characteristics of equipment. Maintenance decisions are based on the risk-cost balance principle; maintenance recommendations are triggered when the anticipated failure risk cost exceeds the cost of preventative maintenance.

[0050] Secondly, a cost-benefit balance model is designed to optimize maintenance economics. This model comprehensively considers factors such as maintenance costs, downtime losses, spare parts costs, and maintenance effectiveness to calculate the return on investment (ROI) for maintenance. , Where: ROI is the return on investment, expressed as a percentage, representing the proportion of return on maintenance investment; B is the benefit generated by maintenance, expressed in currency (e.g., yuan), including avoided downtime losses, extended equipment life value, etc.; C is the maintenance cost, expressed in currency (e.g., yuan), including labor, spare parts, downtime, and other expenses. When ROI exceeds a set threshold (usually 30%), preventative maintenance is recommended.

[0051] Secondly, a maintenance solution library is established for different failure modes. This library includes diagnostic features and corresponding maintenance measures for common failure types (such as bearing wear, motor overheating, and cooling system blockage). Based on the current state characteristics, the system matches the most similar failure mode and generates targeted maintenance suggestions.

[0052] Then, maintenance window identification is implemented, and the optimal maintenance time is determined by combining the equipment usage plan. The system analyzes the equipment usage schedule, identifies idle windows, and evaluates the match between the window length and maintenance needs, selecting the time period with the least impact on the surgical plan as the recommended maintenance time.

[0053] Finally, a maintenance plan is generated, which includes maintenance recommendations, optimal maintenance time, maintenance cycle, and probability of failure. This plan is presented in a structured manner and includes the following key information: a list of maintenance items and their priorities, estimated maintenance duration, required spare parts and tools, expected maintenance results, and risk warnings.

[0054] To achieve continuous system optimization, this invention also includes steps for maintenance effectiveness evaluation and continuous system optimization: First, establish a pre- and post-maintenance comparative analysis framework to evaluate maintenance effectiveness. Quantify the maintenance effect by comparing health indices, performance parameters, and fault characteristics before and after maintenance. The Maintenance Effect Score (MES) calculation formula is as follows: , Where: MES is the maintenance effectiveness score, which is dimensionless and has a range of values. A higher value indicates better maintenance performance; and These represent the health index before and after maintenance, dimensionless, and their range of values. ; and These are the performance parameters before and after maintenance, with units varying depending on the parameter type; For ideal performance parameters, and and Same units; and These are the fault characteristic intensities before and after maintenance, respectively, and are dimensionless. , and The weighting coefficient is dimensionless and has a range of values. ,and The formula measures the degree of improvement in health index, the degree of recovery in performance parameters, and the degree of reduction in fault characteristics, comprehensively reflecting the maintenance effect.

[0055] Secondly, design knowledge base update rules to accumulate maintenance experience. The system adds successful maintenance cases (MES>0.8) to the knowledge base, including information such as fault characteristics, maintenance measures, and effectiveness evaluation, to provide a reference for similar situations in the future.

[0056] Furthermore, a model calibration mechanism based on maintenance results is constructed to continuously optimize prediction accuracy. The system compares the problems discovered during actual maintenance with the predicted results, calculates the prediction error, and adjusts the prediction model parameters accordingly. Model updates employ an online learning approach to ensure the model can adapt to dynamic changes in equipment status.

[0057] Finally, a case study learning framework is implemented to extract best practices from historical maintenance. The system analyzes the common characteristics of efficient maintenance cases, extracts maintenance patterns and successful experiences, and forms standardized best practice guidelines to guide future maintenance work.

[0058] The surgical dynamic system prediction and maintenance system based on digital twin provided by this invention includes the following core modules: digital twin modeling module, multi-source data acquisition module, matrix manifold representation module, adaptive fusion module, degradation model construction module, health status assessment module, prediction and maintenance decision module, and closed-loop optimization module.

[0059] The digital twin modeling module is used to digitally model the surgical dynamics system, establish a physical model of the surgical dynamics system, and form a digital twin system through sensor information mapping. This module includes a static modeling unit, a dynamic behavior simulation unit, and a physical-digital synchronization unit.

[0060] The static modeling unit creates the system's geometric model, including precise dimensional information, material properties, and topology. It uses a parametric modeling method, which facilitates subsequent adjustments and updates.

[0061] The dynamic behavior simulation unit is based on multiphysics theory and simulates the dynamic response of the system under different working conditions, including mechanical motion, heat conduction and electrical performance, forming a comprehensive behavioral description.

[0062] The physical-digital synchronization unit enables two-way information exchange between the physical system and the digital model through sensor data and feedback mechanisms, ensuring that the digital model can accurately reflect the real-time status of the physical equipment.

[0063] The multi-source data acquisition module is used to acquire real-time status information of the surgical power system from the power supply channel, surgical instrument channel, and cooling system channel. This module includes a sensor network unit, a data preprocessing unit, and a data storage unit.

[0064] The sensor network unit consists of various sensors distributed at key locations in the system, including voltage, current, temperature, vibration, acoustic, flow, and pressure sensors, which form a unified sensor network through wired or wireless means.

[0065] The data preprocessing unit is responsible for cleaning, denoising, standardizing, and extracting features from the raw sensor data to improve data quality and lay the foundation for subsequent analysis.

[0066] The data storage unit adopts a time-series database architecture, which efficiently stores and manages large amounts of time-series data and supports fast querying and historical data analysis.

[0067] The matrix manifold representation module is used to construct a heterogeneous data representation mechanism, mapping multi-source channel data to a unified feature space. This module includes a feature matrix construction unit, a nonlinear mapping unit, a manifold structure analysis unit, and a multi-scale feature extraction unit.

[0068] The feature matrix construction unit organizes the raw data of each channel into a three-dimensional feature matrix, which facilitates subsequent processing and analysis.

[0069] The nonlinear mapping unit uses a deep neural network to map heterogeneous data to a unified feature space, thus solving the problem of data heterogeneity.

[0070] Based on manifold learning theory, the manifold structure analysis unit analyzes the geometric structure of the feature space, constructs a geodesic distance metric, and more accurately represents the relationships between data.

[0071] The multi-scale feature extraction unit extracts features from multiple perspectives, including the time domain, frequency domain, and time-frequency domain, to comprehensively capture system state information.

[0072] The adaptive fusion module is used to establish a multi-channel fusion algorithm with optimized feature weights, dynamically fusing feature representations. This module includes an information gain analysis unit, a working mode recognition unit, a weight optimization unit, and a multi-level fusion unit.

[0073] The information gain analysis unit calculates the contribution of each feature to the system state prediction and constructs a feature-fault correlation matrix to quantify the value of the features.

[0074] The working mode recognition unit identifies the current working mode (such as no load, light load, medium load, heavy load, etc.) based on system operating parameters, providing a basis for weight adjustment.

[0075] The weight optimization unit dynamically adjusts the weights of each feature based on information gain and working mode to achieve adaptive fusion.

[0076] Multi-level fusion units achieve multi-level fusion at the feature level, decision level, and spatiotemporal scale to form a comprehensive system state representation.

[0077] The degradation model construction module is used to construct an aging degradation model of the surgical dynamic system based on fusion features. This module includes a health indicator design unit, a degradation trend analysis unit, and a lifespan prediction unit.

[0078] The health indicator design unit defines a multi-dimensional health indicator system that covers performance, reliability, and efficiency, comprehensively assessing the health status of the system.

[0079] The degradation trend analysis unit constructs a degradation trend matrix, describes the system degradation path through a nonlinear model, and identifies the triggering conditions that accelerate degradation.

[0080] The life prediction unit estimates the remaining life of the equipment based on degradation trends, providing a time reference for maintenance decisions.

[0081] The health status assessment module is used to input real-time status information into the degradation model and output the system aging level. This module includes a real-time monitoring unit and a status assessment unit.

[0082] The real-time monitoring unit continuously collects system operation data and inputs the data into the degradation model to achieve real-time status updates.

[0083] Based on the output of the degradation model and combined with preset health thresholds, the status assessment unit evaluates the current health status and aging degree of the system and generates a status report.

[0084] The predictive maintenance decision-making module generates predictive maintenance plans based on the system's aging status. This module includes a risk assessment unit, a maintenance timing optimization unit, a maintenance strategy generation unit, and a visualization unit.

[0085] The risk assessment unit calculates the risk of equipment failure and potential losses based on the system's aging level, providing a risk basis for maintenance decisions.

[0086] The maintenance timing optimization unit comprehensively considers risks, costs, and equipment usage plans to determine the optimal maintenance time and minimize the impact of maintenance on normal operations.

[0087] The maintenance strategy generation unit selects or generates targeted maintenance strategies from the maintenance plan library based on fault characteristics and degradation status, including maintenance items, required resources, and expected results.

[0088] The visualization unit presents system status, risk assessment, and maintenance recommendations in an intuitive way, supporting decision-makers to quickly understand and make decisions.

[0089] The closed-loop optimization module is used to evaluate maintenance effectiveness and continuously optimize the predictive model and maintenance strategy. This module includes an effectiveness evaluation unit, a knowledge accumulation unit, and a model update unit.

[0090] The effect evaluation unit compares the changes in system status before and after maintenance, quantifies the maintenance effect, and generates an evaluation report.

[0091] The knowledge accumulation unit collects and organizes maintenance cases, extracts experiences and best practices, enriches the knowledge base, and guides future maintenance.

[0092] The model update unit adjusts the prediction model parameters based on the maintenance results and prediction errors to improve prediction accuracy and achieve continuous optimization and adaptive adjustment of the system.

[0093] Through the coordinated operation of the above modules, the predictive maintenance system of the present invention achieves comprehensive monitoring, accurate assessment and intelligent maintenance decision-making of the surgical power system status, effectively improving equipment reliability, reducing maintenance costs and ensuring surgical safety.

[0094] The present invention provides a method and system for predictive maintenance of surgical power systems based on digital twins. Through steps such as digital twin modeling, multi-source data acquisition, matrix manifold representation, adaptive feature fusion, degradation model construction, health status assessment, and predictive maintenance decision-making, it achieves accurate assessment of the health status of surgical power systems, accurate prediction of performance degradation, and intelligent optimization of maintenance decisions.

[0095] The innovations of this invention are mainly reflected in the heterogeneous data representation mechanism based on matrix manifolds and the multi-channel fusion algorithm with adaptive feature weight optimization. This solves the problem of multi-source heterogeneous data fusion in predictive maintenance of surgical dynamic systems, improving prediction accuracy and maintenance efficiency. Furthermore, this invention establishes a maintenance effect evaluation and knowledge accumulation mechanism, enabling continuous system optimization and adaptive adjustment, forming a complete predictive maintenance closed loop.

[0096] Compared with existing technologies, this invention has the following advantages: it can comprehensively capture the health status information of the surgical power system; accurately predict the trend of equipment performance degradation; optimize maintenance timing and strategies, reducing maintenance costs; and achieve continuous system optimization and knowledge accumulation. This invention is of great significance for improving the reliability and safety of surgical power systems, reducing medical risks, and enhancing the quality of medical services.

[0097] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for predictive maintenance of surgical dynamic systems based on digital twins, characterized in that, include: A digital model of the surgical power system is created, and a physical model of the surgical power system is established. By collecting information from surgical power sensors, the physical model of the surgical power system is mapped to form a mapping-driven digital twin system of surgical power. Real-time status information of the surgical power system is collected from the power supply channel, surgical instrument channel, and cooling system channel; A heterogeneous data representation mechanism based on matrix manifold is constructed to map the real-time status information of the power supply channel, surgical instrument channel and cooling system channel to a unified feature space to form feature representations; An adaptive feature weight optimization multi-channel fusion algorithm is established to dynamically fuse the feature representations to obtain fused features; Based on the aforementioned fusion features, an aging and degradation model of the surgical power system is constructed. The real-time status information of the surgical power system is input into the aging and degradation model of the surgical power system, and the aging degree of the surgical power system is output. A predictive maintenance plan is generated based on the aging degree of the surgical power system.

2. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The digital modeling of the surgical power system includes: Establish a static geometric model that includes the power supply unit, transmission mechanism, and cooling system; Construct a multi-physics domain dynamic behavior model encompassing the mechanical, thermal, and electrical domains; Design a sensor deployment strategy and establish a real-time data synchronization channel; The real-time data synchronization channel enables bidirectional interaction between the physical system and the digital model, supporting automatic parameter calibration.

3. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The real-time status information of the surgical power system collected from the power supply channel, surgical instrument channel, and cooling system channel includes: Voltage sensors, current sensors, and power sensors are deployed in the power supply channel to collect power status data; Vibration sensors, acoustic sensors, and temperature sensors are installed in the surgical instrument channel to collect status data of mechanical components; Flow sensors, temperature sensors, and pressure sensors are deployed in the cooling system channels to collect cooling efficiency data; The power status data, mechanical component status data, and cooling efficiency data are preprocessed, including signal noise reduction, time synchronization, and data standardization.

4. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The construction of a heterogeneous data representation mechanism based on matrix manifolds includes: Construct a power channel feature matrix, a device channel feature matrix, and a cooling system feature matrix. The dimensions of the feature matrix include a time dimension, a channel index dimension, and a feature index dimension. Design a nonlinear mapping function to project the feature matrix onto a shared feature space; Construct a geodesic distance metric to quantify the similarity of data points in the feature space; Multi-scale feature extraction is achieved, extracting time-domain features, frequency-domain features, and joint time-frequency-domain features respectively.

5. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The multi-channel fusion algorithm for establishing adaptive feature weight optimization includes: Construct an information gain matrix to quantify the contribution of each channel's data to system state prediction; Establish a feature-fault correlation matrix to quantify the ability of features to identify different types of faults; Identify the operating modes of the surgical power system and design specific weight allocation strategies for different operating modes; Achieve a multi-level fusion mechanism, including feature-level fusion, decision-level fusion, and spatiotemporal scale fusion.

6. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The aging and degradation model of the surgical power system is a data-driven degradation model, including: Design multidimensional health indicators to comprehensively assess the system's status; Construct a degradation trend matrix to capture the aging patterns of components; To perform nonlinear degradation path analysis and identify triggering conditions that accelerate degradation; Establish a life prediction model based on degradation trends to dynamically estimate the remaining life of equipment.

7. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The aging prediction and maintenance scheme based on the surgical power system includes: Construct a risk-based maintenance timing decision framework; Design a cost-benefit balance model to optimize maintenance economy; Establish a maintenance solution library for different failure modes; To enable maintenance window identification and determine the optimal maintenance time based on equipment usage plans; Generate a maintenance plan that includes maintenance recommendations, optimal maintenance time, maintenance cycle, and failure probability.

8. The method for predictive maintenance of surgical power systems based on digital twins according to claim 7, characterized in that: Further steps include maintenance effectiveness evaluation and continuous system optimization: Establish a pre- and post-maintenance comparative analysis framework to evaluate the maintenance effectiveness; Design knowledge base update rules and accumulate maintenance experience; Construct a model correction mechanism based on maintenance results to continuously optimize prediction accuracy; Implement a case study learning framework to extract best practices from historical maintenance.

9. The method for predictive maintenance of surgical power systems based on digital twins according to claim 1, characterized in that: The heterogeneous data representation mechanism based on matrix manifolds also includes: Build a data uncertainty model to address sensor noise and data loss issues; Design a data completion strategy based on manifold structure to infer missing values ​​using local geometric information; Construct a low-rank matrix factorization model to capture and reconstruct the intrinsic structure of the data; To achieve feature quality assessment and enhancement, including feature reliability assessment, feature enhancement, and feature compensation.

10. A digital twin-based surgical dynamics system predictive maintenance system, used to implement the digital twin-based surgical dynamics system predictive maintenance method according to any one of claims 1-9, characterized in that, include: The digital twin modeling module is used to digitally model the surgical power system, establish a physical model of the surgical power system, and form a digital twin system through sensor information mapping. The multi-source data acquisition module is used to acquire real-time status information of the surgical power system from the power channel, surgical instrument channel and cooling system channel; The matrix manifold representation module is used to construct a heterogeneous data representation mechanism, mapping multi-source channel data to a unified feature space; The adaptive fusion module is used to establish a multi-channel fusion algorithm with optimized feature weights to dynamically fuse feature representations. A degradation model construction module is used to construct an aging degradation model of the surgical power system based on fusion features; The health status assessment module is used to input real-time status information into the degradation model and output the system aging degree. The predictive maintenance decision module is used to generate predictive maintenance plans based on the system's aging status. as well as The closed-loop optimization module is used to evaluate the maintenance effect and continuously optimize the prediction model and maintenance strategy.

Images