CN-121436975-B - Intelligent digital management method and system for equipment assets

CN121436975BCN 121436975 BCN121436975 BCN 121436975BCN-121436975-B

Abstract

The invention discloses an intelligent digital management method and system for equipment assets, and relates to the technical field of equipment management, wherein the method comprises the steps of collecting multi-mode time sequence data streams sent by physical equipment by using a sensor cluster and converting the multi-mode time sequence data streams into high-dimensional feature vectors; the method comprises the steps of utilizing a high-dimensional feature vector to deduce internal state variables of equipment at the current moment, generating an equipment health state vector after combination, analyzing the equipment health state vector in time sequence arrangement, generating a health degree degradation track, inputting the health degree degradation track to a decision maker, generating an optimal maintenance action sequence for the equipment, executing the generated maintenance action sequence, and continuously optimizing the decision maker according to the executed equipment state. The invention breaks the traditional equipment management information island, can reduce unplanned shutdown, reduce maintenance cost, improve equipment utilization rate and management efficiency, and realize the transition from passive maintenance to predictive management.

Inventors

SU XIHONG
HE YUMENG

Assignees

北京北方科诚信息技术股份有限公司
北京中科数字巨人科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20251216

Claims (6)

1. An intelligent digital management method for equipment assets, comprising: Step1, collecting a multi-mode time sequence data stream sent by physical equipment by using a sensor cluster, and converting the multi-mode time sequence data stream into a high-dimensional feature vector; The sensor cluster at least comprises a vibration sensor, a temperature sensor, an acoustic sensor and a control unit, wherein the vibration sensor is used for collecting vibration frequency spectrums of equipment in three axial directions X, Y, Z; Step2, deducing an internal state variable of the equipment at the current moment by using the high-dimensional feature vector, and generating an equipment health state vector after combination, wherein the method specifically comprises the following substeps: The multi-mode time sequence data of the health equipment is processed into a high-dimensional feature vector set, the obtained high-dimensional feature vector is reduced to a low-dimensional state space by a self-encoder, and health state data points are converged into a compact area in the space, which is called the reference health attractor; calculating the deviation energy level vector between the short-term state track formed by the current high-dimensional feature vector and the reference healthy attractor, namely, the high-dimensional feature vector at the current moment Projecting to the low-dimensional state space to obtain projection points By introducing a sliding window mechanism, a projection point sequence in the latest time period can be obtained I.e. short-term state trajectory of the device, said deviation energy level vector The calculation formula of (2) is expressed as: Wherein Is the high-dimensional feature vector projection point at the current moment, Is any point in the reference healthy attractor M, Is a short-term state trajectory I, i+1th, i-1 th track point, i takes on the value 2~n-1, n is The number of trace points in the track, For restoring force factors, the natural restoring force of the equipment is quantified, calibrated through historical data, In the standard healthy attractor M The closest point; Firstly, respectively normalizing each component in the calculated deviation energy level vector, then introducing an adaptive weight mechanism, endowing corresponding weight coefficients according to the relative importance degree of each component, and finally combining the weighted deviation energy level components into the equipment health state vector at the moment; Step3, analyzing the equipment health state vector under the time sequence arrangement to generate a health degree degradation track, wherein the method specifically comprises the following substeps: self-adapting enhancement of the device health status vector based on the context data of the working condition environment of the device, i.e. for each time step t Taking the materials back and forth The effective time steps form local neighborhood Extracting context data of the working condition environment of the equipment under the corresponding time step, converting the context data into vectors, and adding the vectors into local neighborhood, namely Then bringing the data in the local neighborhood into the formula: In (3) obtaining an enhanced health status vector Wherein Is the sensitivity adjustment parameter, j is the time step index in the local neighborhood, m is the total number of samples in the local neighborhood, The device health state vector and the working condition environment vector at the jth time step in the local neighborhood are respectively, Is the working condition environment vector at the current time step t, Is the maximum difference value of the global working condition, Is a cosine similarity function; And calculating degradation factors of each dimension of the enhanced equipment health state vector, wherein the degradation factor calculation formula is expressed as follows: Wherein A degradation factor representing the d-th dimension of the device health status vector at time step t, 、 The d-th dimension component of the device health status vector at time step t and time step t-1 respectively, Is the time interval of time steps t and t-1, Is a time series sequence of the d-th dimensional component of the device health status vector, Is a time sequence of the working condition environment vectors, Is a covariance operation function, which is a function of the covariance operation, Is a standard deviation operation function, and is characterized by that, Is the dimension index in the summation process, D is the total dimension of the device health status vector, The meaning of each parameter of the dimension is consistent with that of each parameter of the d dimension; according to the degradation factors of each dimension, calculating the health degree of the equipment at each time step t, and finishing the health degree into an original health degree sequence The calculation formula is expressed as: Wherein Respectively adjustable scale parameters and nonlinear coefficients, Is the maximum allowable value of the d dimension of the equipment health state vector under the equipment health state; inputting an original health degree sequence to a GPR model, and outputting a complete health degree degradation track; step4, inputting a health degree degradation track to a decision maker, and generating an optimal maintenance action sequence for equipment, wherein the training process of the decision maker specifically comprises the following substeps: The method comprises the steps of collecting historical equipment operation data and constructing a training data set, wherein the historical equipment operation data comprises a health degree sequence, maintenance action records and state feedback after maintenance, the data is cut into a plurality of segments according to a maintenance period, and each segment comprises a health degree degradation track, an executed maintenance action sequence and a corresponding health degree change result thereof to form the training data set; The decision maker comprises a strategy network and an evaluation network, wherein the strategy network is used for generating a maintenance action sequence according to an input health degradation track, and the evaluation network is used for predicting health trend after executing the maintenance sequence; Thawing a policy network of the decision maker, calculating target loss of the decision maker in real time by combining an evaluation network, and fine-tuning the decision maker based on a minimum loss principle until the model converges; step5, executing the generated maintenance action sequence, and continuously optimizing the decision maker according to the executed equipment state.
2. The method for intelligent digital management of equipment assets according to claim 1, wherein the multi-modal time series data stream is converted into a high-dimensional feature vector, and the method is specifically divided into the following sub-steps: Performing time alignment and preprocessing on data streams acquired by the sensor clusters; inputting the preprocessed data stream into a pre-trained special feature extraction neural network, and extracting vibration feature vectors, temperature feature vectors and acoustic feature vectors of equipment; And splicing the extracted modal feature vectors to generate a high-dimensional feature vector.
3. The method for intelligent digital management of equipment assets according to claim 1, wherein the decision maker is continuously optimized according to the executed equipment state, and the method is specifically divided into the following sub-steps: Calculating the current health degree of the equipment according to the multi-mode time sequence data stream fed back by the sensor cluster; if the health degree exceeds the predicted health degree, the decision weights of various maintenance actions in the current maintenance action sequence are enhanced; and if the health degree does not reach the predicted health degree, correcting the predicted deviation of the evaluation network.
4. An equipment asset intelligent digital management system is characterized by comprising a data stream conversion module, a health state vector generation module, a health degradation track generation module, a maintenance action sequence generation module and a maintenance action sequence execution module, wherein the data stream conversion module, the health state vector generation module, the health degradation track generation module, the maintenance action sequence generation module and the maintenance action sequence execution module are used for executing the equipment asset intelligent digital management method according to any one of claims 1 to 3; The data stream conversion module is used for collecting the multi-mode time sequence data stream sent by the physical equipment by using the sensor cluster and converting the multi-mode time sequence data stream into a high-dimensional feature vector; The health state vector generation module is used for utilizing the high-dimensional characteristic vector to deduce an internal state variable of the equipment at the current moment, and generating an equipment health state vector after combination; the health degradation track generation module is used for analyzing the health state vector of the equipment in time sequence arrangement and generating a health degree degradation track; The maintenance action sequence generation module is used for inputting the health degree degradation track to the decision maker and generating an optimal maintenance action sequence for the equipment; And the maintenance action sequence execution module is used for executing the generated maintenance action sequence and continuously optimizing the decision maker according to the executed equipment state.
5. The intelligent digital management system of equipment assets according to claim 4, wherein the data stream conversion module comprises a data stream preprocessing sub-module, a feature vector extraction sub-module and a high-dimensional feature vector generation sub-module; The data stream preprocessing sub-module is used for performing time alignment and preprocessing on the data streams acquired by the sensor clusters; the feature vector extraction submodule is used for inputting the preprocessed data stream into a pre-trained special feature extraction neural network and extracting vibration feature vectors, temperature feature vectors and acoustic feature vectors of equipment; and the high-dimensional feature vector generation sub-module is used for splicing the extracted modal feature vectors to generate a high-dimensional feature vector.
6. The intelligent digital management system of equipment assets according to claim 4, wherein the health degradation track generation module specifically comprises a feature enhancer module, a health degree sequence generation sub-module and a health degree degradation track perfection sub-module; The characteristic enhancer module is used for adaptively enhancing the health state vector of the equipment based on the context data of the working condition environment where the equipment is located; the health degree sequence generation sub-module is used for calculating the health degree of the equipment under each time step t according to the degradation factors of each dimension of the equipment health state vector and finishing the health degree into an original health degree sequence; and the health degree degradation track perfecting submodule is used for inputting the original health degree sequence into the GPR model and outputting a complete health degree degradation track.

Description

Intelligent digital management method and system for equipment assets Technical Field The invention relates to the technical field of equipment management, in particular to an equipment asset intelligent digital management method and system. Background In modern industrial production, infrastructure operations, and daily operations in large institutions (e.g., hospitals, universities), equipment assets are used as core production elements, and the management efficiency is directly related to the operation cost, production safety, and economic benefits. Traditional equipment asset management modes mainly rely on manual ledgers, periodic inspection and planned maintenance, and the limitations of the traditional equipment asset management modes are increasingly prominent, and mainly appear in the following aspects: First, the information island phenomenon is serious, and the data value is difficult to mine. Equipment asset information of an enterprise is typically distributed among different management departments and systems, such as purchasing information in an ERP system, maintenance records in a CMMS system, and real-time operation data in a SCADA system. The lack of efficient data communication and integration between these systems forms individual "islands of information". The manager has difficulty in obtaining a complete and unified view of the whole life cycle of the device, and cannot perform global data analysis and optimization decision. Second, maintenance strategies are passively lagged and unplanned downtime is costly. The traditional maintenance modes are mostly post-maintenance (Breakdown Maintenance) or fixed planned preventive maintenance (PREVENTIVE MAINTENANCE). The former usually intervenes after the equipment fails, resulting in unplanned downtime, resulting in significant production loss and safety risks, while the latter, based on a fixed time or operating period, may produce "overserving", wasting resources, and may fail to discover potential failures in time, resulting in insufficient maintenance. Neither of these modes can accurately predict the health of the device. Thirdly, the management process is highly dependent on manual work, low in efficiency and prone to error. From asset inventory to status records, from worksheets to maintenance reports, a large number of jobs rely on field inspection by management personnel and paper or Excel form records. The mode has low working efficiency and high labor cost, and is extremely easy to cause data recording errors, information transfer delay or omission due to human factors, so that the accuracy and the instantaneity of management data are greatly reduced. Fourth, asset status is agnostic and full lifecycle management is difficult. For equipment assets with huge quantity and wide distribution, a manager is difficult to accurately grasp dynamic information such as geographic positions, current working states, utilization rates, performance attenuation conditions and the like in real time. The phenomenon of idle and waste of the assets is common, optimal allocation, updating or scrapping decisions cannot be made based on real-time data, and the aim of preserving and increasing the value of the assets is fulfilled. With the maturation and popularization of new generation information technologies such as internet of things (IoT), big data, artificial Intelligence (AI), cloud computing and the like, technical possibilities are provided for solving the problems. Attempts have been made in the industry to utilize these technologies for device management, such as by adding sensors to collect device operational data, or using RFID, two-dimensional codes for asset identification. However, the existing scheme is often simply used for digitizing the traditional management flow, and cannot really realize the intelligent. The method has the problems that the data acquisition dimension is single, fusion analysis of multi-source data is lacking, the data analysis capability is weak, data visualization and simple alarm are mostly only realized, the deep learning and predictive analysis capability is lacking, the system architecture is stiff, and complex and changeable management scenes are difficult to flexibly expand and adapt. Therefore, the field urgently needs an intelligent digital management method for equipment assets, which is deeply integrated with advanced information technology, can break information islands, realize data driving decisions and has self-learning and self-adapting capabilities, so as to fundamentally improve the reliability, utilization rate and management efficiency of the equipment assets and reduce the cost of the whole life cycle. Disclosure of Invention The invention provides an intelligent digital management method for equipment assets, which comprises the following steps: Step1, collecting a multi-mode time sequence data stream sent by physical equipment by using a sensor cluster, and converting the multi-mode time seque