Equipment state real-time monitoring and optimizing method and system based on industrial internet

Abstract

The invention belongs to the technical field of industrial internet monitoring, and relates to an industrial internet-based equipment state real-time monitoring and optimizing method and system, and the method comprises the steps: S1, arranging a sensor layer for collecting equipment operation data; s2, arranging an edge calculation layer for processing data acquired by the sensor layer; and S3, arranging a cloud platform layer for storing, visualizing and optimizing the data. The technical problems of monitoring lag, low optimization efficiency and data isolated island in the prior art are effectively solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of industrial internet monitoring, and relates to a method and system for real-time monitoring and optimization of the state of equipment, in particular, a method and system for real-time monitoring and optimization of the state of equipment based on industrial internet. BACKGROUND

[0002] The existing equipment state monitoring and fault diagnosis method mainly relies on simple data acquisition and basic signal processing technology, which is difficult to meet the intelligent monitoring needs of modern industrial equipment under complex working conditions. The existing technology has obvious deficiencies in data acquisition accuracy, real-time performance and anti-interference performance, and lacks effective fusion and analysis capabilities for multi-source heterogeneous data.

[0003] To this end, the invention patent with publication number CN120067772B discloses an intelligent decision management method for energy-saving equipment state monitoring and fault diagnosis, which comprises the following steps: integrating a signal conditioning module, a gain amplification unit, a digital-to-analog conversion unit, and an operation processor into an edge computing platform, connecting an environmental parameter acquisition module and an energy consumption acquisition module to a data acquisition channel of the digital-to-analog conversion unit, filtering and amplifying the collected signals based on the signal conditioning module, adjusting the gain coefficient of the acquisition channel through the gain amplification unit, converting the conditioned signals into digital quantities using the digital-to-analog conversion unit, and transmitting the digital quantities to a multi-level data cache through a bus; inputting the monitoring data in the multi-level data cache into a time sequence feature extraction model, performing principal component analysis on the extracted time sequence features to obtain a feature vector, building a working condition judgment model based on a support vector machine, inputting the feature vector into the working condition judgment model for classification, establishing a corresponding relationship between the classification result and a preset sampling frequency threshold, selecting the sampling frequency parameter corresponding to the current working condition according to the corresponding relationship, inputting the adjusted sampling data into a data fusion model, inputting the fused data into a collaborative diagnosis system composed of a rule model, a machine learning model, and a deep network model, embedding a feature mapping library constructed by equipment component parameters, fault types, and fault modes into the collaborative diagnosis system, correcting the diagnosis result based on the feature mapping library, and inputting the corrected diagnosis result into a decision system; inputting the equipment state indicators output by the decision system and the operation records, maintenance records, and fault records in the equipment historical database into a trend modeling unit, decomposing the state indicators into trend items, periodic items, and random items in the trend modeling unit, establishing a device degradation model based on a Weibull distribution, combining the device degradation model with a long short-term memory network to build a residual life prediction model, taking the state indicator decomposition items, degradation parameters, and time scale as input variables in the residual life prediction model, combining fault evolution analysis to build a risk warning function, combining the risk warning function with the residual life prediction model to establish a hierarchical warning system, generating a maintenance scheme according to the hierarchical warning system, inputting the maintenance scheme into a maintenance decision unit to generate an equipment maintenance plan, and distributing maintenance tasks through a remote communication module based on the equipment maintenance plan.

[0004] The above technical solution has achieved high-reliability acquisition and processing of industrial field signals to some extent. However, there are still the following technical problems: The monitoring data processing is lagging, and the existing technology relies on cloud centralized processing of data, which has network delay and cannot meet the real-time warning demand (such as the abnormal vibration response time of high-speed rotating equipment needs to be <100ms); The optimization strategy mode is fixed, and the existing technology usually alarms based on preset threshold values, lacks deep analysis of equipment operation data, and cannot dynamically adjust process parameters (such as motor power output) according to working conditions.

[0005] Data islands are formed, different manufacturer device protocols are not unified, data interoperability is poor, and it is difficult to build a global optimization model across devices.

[0006] This is the deficiency of the prior art.

[0007] Therefore, the present application provides a method and system for real-time monitoring and optimization of device status based on industrial internet to solve the above-mentioned defects in the prior art, which is very necessary. SUMMARY

[0008] The purpose of the present application is to provide a method and system for real-time monitoring and optimization of device status based on industrial internet to solve the above-mentioned technical problems.

[0009] To achieve the above-mentioned purpose, the present application provides the following technical solution: A method for real-time monitoring and optimization of device status based on industrial internet, comprising the following steps: Step S1: arranging a sensor layer for collecting device operation data; The sensor layer includes a plurality of types of Internet of Things sensors, which collect key parameters of the device to be tested in real time, and the deployment position of the Internet of Things sensors is determined according to the structural characteristics and key monitoring points of the device to be tested; Step S2: arranging an edge computing layer for processing data collected by the sensor layer; The edge computing layer includes an edge computing gateway, which performs preprocessing, anomaly detection processing and protocol conversion processing on the data collected by the sensor layer; Step S3: arranging a cloud platform layer for storing, visualizing and optimizing data; The cloud platform layer includes a digital twin module and an optimization engine module, the digital twin module is used to construct a device asset digital twin model, and the optimization engine module generates an optimization strategy for device process parameters through a machine learning algorithm.

[0010] As a preferred, in the step S1: The sensor layer is built-in with an edge-side preprocessing unit, which performs preliminary filtering processing on the original collected data, and the collected data contains timestamp data, which is synchronized with the sensor collected data through a hardware clock synchronization protocol.

[0011] The step S1 achieves the following technical effects: The edge-side preliminary filtering unit built-in in the sensor layer can directly filter invalid noise such as electromagnetic interference and mechanical vibration clutter, greatly improve the signal-to-noise ratio of the original data, reduce the invalid processing amount of the subsequent edge computing layer, and reduce the processing pressure from the data source.

[0012] At the same time, the hardware clock synchronization protocol and timestamp ensure the time consistency of cross-sensor data, reduce the time deviation of multi-dimensional data, and solve the state analysis distortion problem caused by time asynchronization in the prior art.

[0013] Preferably, the step S2 specifically comprises: Step S21: a data preprocessing step, in which: Two-stage filtering processing is performed on the data, the first stage of filtering processing adopts Kalman filtering algorithm to process the data, and dynamically eliminates measurement noise; the second stage of filtering processing adopts a Butterworth low-pass filter to process the data, and filters out high-frequency interference; A feature extraction algorithm is used to extract feature values from the preprocessed data; Step S22: a data anomaly detection processing step, in which: A rapid detection layer and an intelligent detection layer are arranged, the rapid detection layer compares the preprocessed data with a preset threshold in real time; the intelligent detection layer runs a lightweight isolation forest algorithm model, and performs deep analysis on the extracted feature values to identify implicit anomalies; Step S23: a protocol conversion processing step, in which: A multi-protocol analysis engine is built in the edge computing gateway, and analysis rules are preset for different types of protocols, the analysis rules comprising: For general industrial protocols, the physical quantity data is directly analyzed by reading the function code, data address and check bit in the protocol frame structure; For manufacturer private protocols, based on the communication specification disclosed by the protocol, the binary data in the device specific data area is analyzed and converted into decimal or floating point type device running parameters; For service-oriented protocols, the real-time state data of the device is obtained by calling the node attributes in the information model, and the data type is analyzed; The analyzed protocol data of various types is uniformly mapped to a preset standardized data model, the standardized data model comprising: device unique identifier, data acquisition timestamp, data type, data value, data quality code (such as "valid" "invalid"); The standardized data model is encapsulated into a message body through the MQTT protocol.

[0014] The technical effects obtained by step S2 are as follows: First, improve data quality and ensure the basis of subsequent analysis accuracy, through the synergistic effect of two-stage filtering - the first stage of Kalman filtering dynamically eliminates random noise, thereby improving the signal-to-noise ratio of the data, and the second stage of Butterworth low-pass filter specifically filters out high-frequency interference, retains the key feature frequency of device operation, and effectively avoids the interference of noise on feature extraction.

[0015] Second, reduce the burden of edge side preprocessing, improve the real-time performance of the system, filter and feature extraction are completed in the edge computing layer, avoid network bandwidth occupation caused by direct uploading of raw data to the cloud, reduce the storage and computing pressure of the cloud.

[0016] Third, while improving the accuracy and response speed of anomaly detection, the fast detection layer realizes fast response based on a preset threshold, ensuring that urgent anomalies can be immediately captured, and the capable detection layer deeply excavates implicit abnormal situations through a lightweight isolation forest algorithm, improving the identification accuracy of weak fault harmonics in vibration spectrum, and realizing early warning of early equipment failure, two-stage detection cooperates to reduce the false negative rate and false positive rate to, reduces the risk of equipment damage caused by false negatives and the loss of unplanned downtime caused by false positives.

[0017] Fourth, break the limitation of data island and realize full interconnection of heterogeneous devices, the multi-protocol analysis engine supports conversion of multiple types of protocols, covers mainstream manufacturer devices, greatly improves data interconnection rate, and standardizes data model to unify data format, so that running data of different manufacturer devices can be directly associated and analyzed on the cloud platform, providing data basis for cross-device collaborative optimization. At the same time, after conversion to MQTT protocol, the bandwidth occupied by data transmission is reduced, and key data can be ensured not to be lost. And the standardized message body is more convenient for the cloud platform to quickly analyze and reduce cloud data processing delay.

[0018] As preferred, the feature extraction algorithm is used in step S21 to extract feature values from the preprocessed data, specifically including: a step of extracting time domain features of vibration sensor data in the sensor layer and a step of extracting frequency domain features of vibration sensor data in the sensor layer; The step of extracting time domain features of vibration sensor data in the sensor layer includes: The effective value of vibration is calculated by the following formula:

[0019] In the formula, RMS is the effective value of the feature, x i is the amplitude value of the i-th sampling point, and N is the number of sampling points in the sliding window; during feature extraction, the window sliding step is matched with the sensor sampling frequency; The step of extracting frequency domain features of vibration sensor data in the sensor layer includes: The time domain signal is subjected to fast Fourier transform to convert the signal from time domain to frequency domain, and the spectral gravity center is extracted by the following formula:

[0020] f K is the K-th frequency component, A kTo correspond to the amplitude.

[0021] The technical effect of the feature extraction algorithm used in step S21 to extract feature values from the preprocessed data is: The feature extraction algorithm can accurately extract key information reflecting the device state from the original sensor data while meeting the real-time constraints of the edge computing layer; reduce the time length of single-device multi-dimensional feature extraction, providing high-quality input for subsequent anomaly detection; the combination of frequency domain and time domain features allows early identification of device failures, providing sufficient time windows for predictive maintenance.

[0022] As a preferred, in the step S3: Based on the three-dimensional size parameters and physical properties uploaded by the edge computing layer, a 1:1 scale virtual model is constructed using a three-dimensional modeling tool, and by accessing device nameplate data and real-time feature data, a "physical property-operation state-life cycle" three-dimensional label is bound to the virtual model, forming a complete digital twin; The optimization engine module generates device-level and production line-level optimization strategies based on historical data and real-time data through machine learning and optimization algorithms.

[0023] The technical effect achieved by step S3 is: The digital twin model realizes the "virtual-real mapping" of the device state, and the management personnel can intuitively master the running state of the whole production line through the cloud interface; the cloud platform layer not only receives real-time data from the edge computing layer, but also feeds back to the edge side through deep analysis and intelligent decision-making, forming a closed loop of "edge real-time response-cloud global optimization", effectively solving the technical problems of "monitoring lag, low optimization efficiency, and data island" in the prior art.

[0024] In addition, the present application also provides a device state real-time monitoring and optimization system based on industrial internet, comprising: A sensor layer module for collecting device operation data; The sensor layer includes various types of Internet of Things sensors, which collect key parameters of the device to be tested in real time, and determine the deployment position of the Internet of Things sensors according to the structural characteristics and key monitoring points of the device to be tested; An edge computing layer module for processing data collected by the sensor layer; The edge computing layer includes an edge computing gateway, which performs pre-processing, anomaly detection processing and protocol conversion processing on the data collected by the sensor layer; A cloud platform layer module for storing, visualizing and optimizing data; The cloud platform layer comprises a digital twin module and an optimization engine module, the digital twin module is used for constructing a device asset digital twin model, and the optimization engine module generates an optimization strategy of a device process parameter through a machine learning algorithm.

[0025] As preferred, in the sensor layer module: The sensor layer built-in edge side pre-processing unit performs preliminary filtering processing on the original collected data, and the collected data contains timestamp data, which is consistent with the sensor collected data in time through a hardware clock synchronization protocol.

[0026] The sensor layer module has the following technical effects: The sensor layer built-in edge side preliminary filtering unit can directly filter invalid noise such as electromagnetic interference and mechanical vibration clutter, greatly improves the signal-to-noise ratio of the original data, reduces the invalid processing amount of the subsequent edge computing layer, and reduces the processing pressure from the data source.

[0027] At the same time, the hardware clock synchronization protocol and the timestamp ensure the time consistency of cross-sensor data, reduce the time deviation of multi-dimensional data, and solve the state analysis distortion problem caused by different time in the prior art.

[0028] As preferred, the edge computing layer module specifically comprises: A data preprocessing submodule, in which: Two-stage filtering processing is performed on the data, the first-stage filtering processing adopts a Kalman filtering algorithm to process the data and dynamically eliminates measurement noise, and the second-stage filtering processing adopts a Butterworth low-pass filter to process the data and filters out high-frequency interference; A feature extraction algorithm is used to extract feature values from the preprocessed data; A data anomaly detection processing submodule, in which: A rapid detection layer and an intelligent detection layer are arranged, the rapid detection layer compares a preset threshold value with the preprocessed data in real time, and the intelligent detection layer runs a lightweight isolation forest algorithm model, deeply analyzes the extracted feature values, and identifies implicit anomalies; A protocol conversion processing submodule, in which: A multi-protocol analysis engine is built in the edge computing gateway, preset analysis rules are set for different types of protocols, and the analysis rules comprise: For a general industrial protocol, physical quantity data is directly analyzed by reading function codes, data addresses and check bits in a protocol frame structure; For a manufacturer's private protocol, binary data in a device specific data area is analyzed based on a communication specification disclosed by the protocol, and is converted into a device running parameter in a decimal or floating point type; For the service-oriented protocol, the real-time state data of the device is obtained by calling the node attribute in the information model thereof, and the data type is parsed; The parsed protocol data of various types is uniformly mapped to a preset standardized data model, and the standardized data model includes: a device unique identifier, a data collection timestamp, a data type, a data value, and a data quality code; The standardized data model is encapsulated into a message body through the MQTT protocol.

[0029] The technical effects achieved by the edge computing layer module are as follows: First, improve data quality and ensure the accuracy of subsequent analysis, through the synergistic effect of two-stage filtering: the first stage Kalman filter dynamically eliminates random noise, thereby improving the signal-to-noise ratio of the data, and the second stage Butterworth low-pass filter specifically filters out high-frequency interference and retains the key feature frequency of device operation, effectively avoiding the interference of noise on feature extraction.

[0030] Second, reduce the burden of edge-side preprocessing and improve the real-time performance of the system, the filtering and feature extraction are completed in the edge computing layer, avoiding the network bandwidth occupation caused by directly uploading the original data to the cloud, and reducing the storage and computing pressure of the cloud.

[0031] Third, improve the accuracy and response speed of anomaly detection, the fast detection layer realizes fast response based on a preset threshold, ensures that emergency anomalies can be immediately captured, and the capable detection layer deeply excavates implicit abnormal situations through a lightweight isolation forest algorithm, improves the recognition accuracy of weak fault harmonics in vibration spectrum, realizes early warning of early device failure, and reduces the false negative rate and false positive rate to reduce the risk of device damage caused by false negatives and the loss caused by false positives.

[0032] Fourth, break the limitation of data silos and realize full interconnection of heterogeneous devices, the multi-protocol analysis engine supports conversion of multiple types of protocols, covers mainstream manufacturer devices, greatly improves the data interconnection rate, and the standardized data model unifies the data format, so that the operation data of devices of different manufacturers can be directly associated and analyzed on the cloud platform, providing a data basis for cross-device collaborative optimization. At the same time, after conversion to the MQTT protocol, the bandwidth occupied by data transmission is reduced, and key data can be ensured not to be lost. The standardized message body is more convenient for the cloud platform to quickly analyze and reduce the delay of cloud data processing.

[0033] As a preferred, the data preprocessing submodule adopts a feature extraction algorithm to extract feature values from the preprocessed data, specifically including: a time domain feature extraction unit for vibration sensor data in the sensor layer and a frequency domain feature extraction unit for vibration sensor data in the sensor layer; The time domain feature extraction unit for the vibration sensor data in the sensor layer comprises: The vibration effective value is calculated by the following formula:

[0034] In the formula, RMS is the effective value of the feature, x i is the amplitude of the i-th sampling point, and N is the number of sampling points in the sliding window; during feature extraction, the window sliding step is matched with the sensor sampling frequency; The frequency domain feature extraction unit for the vibration sensor data in the sensor layer comprises: The time domain signal is subjected to fast Fourier transform to convert the signal from the time domain to the frequency domain, and the spectral gravity center is extracted by the following formula:

[0035] f K is the K-th frequency component, and A k is the corresponding amplitude.

[0036] The technical effect of the feature extraction algorithm for extracting feature values from the preprocessed data in the data preprocessing submodule is: The feature extraction algorithm can accurately extract key information reflecting the device state from the original sensor data while meeting the real-time constraints of the edge computing layer; the length of single-device multi-dimensional feature extraction is reduced, providing high-quality input for subsequent anomaly detection; the combination of frequency domain and time domain features makes the identification of early device failures advance, providing sufficient time window for predictive maintenance.

[0037] As a preferred, in the cloud platform layer module: Based on the three-dimensional size parameters and physical properties uploaded by the edge computing layer, a 1:1 scale virtual model is constructed using a three-dimensional modeling tool, and by accessing the device nameplate data and real-time feature data, a "physical property-operation state-life cycle" three-dimensional label is bound to the virtual model, forming a complete digital twin; The optimization engine module generates device-level and production line-level optimization strategies based on historical data and real-time data through machine learning and optimization algorithms.

[0038] The technical effect achieved by the cloud platform layer module is: The digital twin model realizes the "virtual-real mapping" of the device state, and the management personnel can intuitively master the running state of the whole production line through the cloud interface; the cloud platform layer not only undertakes the real-time data of the edge computing layer, but also feeds back to the edge side through deep analysis and intelligent decision-making, forming a closed loop of "edge real-time response-cloud overall optimization", which effectively solves the technical problems of "monitoring lag, low optimization efficiency, and data island" in the prior art.

[0039] The beneficial effects of the present application are that, by the technical scheme of the present application, the following are achieved: First, real-time collection and edge-side preprocessing of device state data, through rapid processing and analysis of data in the edge computing layer, the amount of data uploaded to the cloud is reduced, the cloud computing pressure and network transmission delay are reduced, and device abnormalities can be identified and warned in real time. Effectively solve the technical problems of monitoring data processing lag in the prior art.

[0040] Second, by constructing a device digital twin model and fusing a machine learning algorithm, accurate mapping and deep analysis of the device running state are realized, and an optimized control strategy is dynamically generated according to the real-time working condition of the device, thereby improving the adaptive ability and operating efficiency of the production line. Effectively solve the technical problem of fixed optimization strategy mode in the prior art.

[0041] Third, multi-protocol compatibility, realizing standardized conversion and centralized management of device data from different manufacturers, breaking down the barriers of heterogeneous device data, and supporting cross-device and cross-production line collaborative optimization. Effectively solve the technical problem of data silos in the prior art.

[0042] As can be seen, compared with the prior art, the present application has outstanding substantial characteristics and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0044] Figure 1 is a flowchart of a device state real-time monitoring and optimization method based on industrial internet provided by the present application.

[0045] Figure 2 is a principle block diagram of a device state real-time monitoring and optimization system based on industrial internet provided by the present application.

[0046] Among them, 1-sensor layer module, 2-edge computing layer module, 3-cloud platform layer module. DETAILED DESCRIPTION

[0047] The present application will be described in detail below in conjunction with the drawings and through specific embodiments. The following embodiments are an explanation of the present application, and the present application is not limited to the following embodiments.

[0048] Embodiment 1: As Figure 1As shown, the method for real-time monitoring and optimization of equipment state based on industrial internet provided by the embodiment comprises the following steps: Step S1: arranging a sensor layer for collecting equipment operation data; The sensor layer comprises a plurality of types of Internet of Things sensors, which collect key parameters of the equipment to be measured in real time, and the deployment positions of the Internet of Things sensors are determined according to the structural characteristics and key monitoring points of the equipment to be measured; Step S2: arranging an edge computing layer for processing the data collected by the sensor layer; The edge computing layer comprises an edge computing gateway, which performs preprocessing, anomaly detection processing and protocol conversion processing on the data collected by the sensor layer; Step S3: arranging a cloud platform layer for storing, visualizing and optimizing the data; The cloud platform layer comprises a digital twin module and an optimization engine module, the digital twin module is used to construct a digital twin model of the equipment asset, and the optimization engine module generates an optimization strategy for the process parameters of the equipment through a machine learning algorithm.

[0049] In the step S1: The sensor layer is internally provided with an edge-side preprocessing unit, which performs preliminary filtering processing on the original collected data, and the collected data contains timestamp data, which is kept consistent with the time of the sensor collected data through a hardware clock synchronization protocol.

[0050] The technical effects of step S1 include: The edge-side preliminary filtering unit internally provided in the sensor layer can directly filter invalid noise such as electromagnetic interference and mechanical vibration clutter, greatly improve the signal-to-noise ratio of the original data, reduce the invalid processing amount of the subsequent edge computing layer, and reduce the processing pressure from the data source.

[0051] At the same time, the hardware clock synchronization protocol and the timestamp ensure the time consistency of cross-sensor data, reduce the time deviation of multi-dimensional data (such as vibration, temperature, current), and solve the state analysis distortion problem (such as misjudging that "vibration peak value is irrelevant to temperature peak value") caused by different time in the prior art.

[0052] The step S2 specifically comprises: Step S21: data preprocessing step, in which: Two-stage filtering processing is performed on the data, the first stage of filtering processing adopts Kalman filtering algorithm to process the data, and dynamically eliminates measurement noise (such as mechanical vibration interference and electromagnetic noise); the second stage of filtering processing adopts a Butterworth low-pass filter to process the data, and filters out high-frequency interference; A feature extraction algorithm is used to extract feature values from the preprocessed data; Step S22: a step of data anomaly detection processing, in which: The rapid detection layer and the intelligent detection layer are arranged, the rapid detection layer compares preset threshold values (such as vibration amplitude > 0.1g, temperature > 80℃) with preprocessed data in real time, and the intelligent detection layer runs a lightweight isolation forest algorithm model to perform deep analysis on the extracted characteristic values and identify implicit anomalies (such as weak fault harmonics appearing in a vibration frequency spectrum). Step S23: a step of protocol conversion processing, in which: The edge computing gateway is internally provided with a multi-protocol analysis engine, preset analysis rules are provided for different types of protocols, and the analysis rules include: For general industrial protocols (such as Modbus RTU / TCP), temperature, current and other physical quantity data are directly analyzed by reading function codes (such as 03H reading a holding register) in a protocol frame structure, data addresses and check bits (CRC check). For manufacturer private protocols (such as Siemens S7 protocol), based on the communication specifications disclosed by the protocol (such as the PDU data unit structure of S7COMM protocol), binary data in a device-specific data area (such as an input / output image area, a timer / counter area) is analyzed to convert the binary data into decimal or floating-point device operating parameters. For OPCUA and other service-oriented protocols, real-time state data (such as vibration amplitude, motor speed) of a device is obtained by calling a node (Node) attribute in an information model thereof, and data types (such as Boolean type, double-precision floating-point type) are analyzed. Various types of protocol data after analysis are uniformly mapped to a preset standardized data model, and the standardized data model includes: a device unique identifier (such as “machine tool-Fanuc-001”), a data acquisition timestamp (accurate to milliseconds), a data type (such as “vibration amplitude” “temperature”), a data value (such as “0.08g” “65℃”), and a data quality code (such as “valid” “invalid”). The standardized data model is encapsulated into a message body through MQTT protocol.

[0053] Step S2 achieves the following technical effects: First, improve data quality and ensure the accuracy basis for subsequent analysis, through the synergistic effect of two-stage filtering: the first stage Kalman filter dynamically eliminates random noise such as mechanical vibration interference and electromagnetic noise, thereby improving the signal-to-noise ratio of the data, and the second stage Butterworth low-pass filter specifically filters high-frequency interference (such as vibration clutter above 1kHz), retains the key characteristic frequency of device operation, and effectively avoids noise interference on feature extraction.

[0054] Second, reduce the burden of edge side preprocessing, improve the real-time performance of the system, filter and feature extraction are completed in the edge computing layer, avoid network bandwidth occupation caused by uploading raw data directly to the cloud (data volume is reduced by more than 60%), reduce the storage and computing pressure of the cloud.

[0055] Third, while improving the accuracy and response speed of anomaly detection, the fast detection layer realizes fast response based on a preset threshold, ensuring that emergency anomalies (such as device overheating) can be immediately captured, and the detection layer can detect hidden abnormal situations through a lightweight isolation forest algorithm, improving the identification accuracy of weak fault harmonics in vibration spectrum (such as the characteristic frequency of early bearing cracks), and realizing early warning of early equipment failure, two-stage detection cooperates to reduce the false negative rate and false positive rate to, reduces the risk of equipment damage caused by false negatives and the loss of unplanned downtime caused by false positives.

[0056] Fourth, break the limitation of data silos and realize full interconnection of heterogeneous devices, the multi-protocol analysis engine supports conversion of multiple types of protocols (such as Siemens S7, Mitsubishi MC, Modbus, OPCUA), covering mainstream manufacturer devices, greatly improving data interconnection rate, and standardized data model unifying data format (including device identification, timestamp, data type, etc. Core fields), so that the running data of different manufacturer devices (such as the position parameters of ABB robots and the speed parameters of Fanuc machine tools) can be directly associated and analyzed on the cloud platform, providing data basis for cross-device collaborative optimization (such as production line load balancing scheduling). At the same time, after conversion to MQTT protocol, the bandwidth occupied by data transmission is reduced, and key data (such as device status at abnormal time) can be ensured not to be lost. And the standardized message body (JSON format) is more convenient for the cloud platform to quickly parse, reducing the cloud data processing delay.

[0057] The feature extraction algorithm is used to extract feature values from the preprocessed data in step S21, which specifically includes the steps of extracting time domain features of vibration sensor data in the sensor layer and extracting frequency domain features of vibration sensor data in the sensor layer. The step of extracting time domain features of vibration sensor data in the sensor layer includes: The effective value of vibration is calculated by the following formula:

[0058] In the formula, RMS is the effective value of the feature, x i is the amplitude value of the i-th sampling point, and N is the number of sampling points in the sliding window; during feature extraction, the window sliding step is matched with the sensor sampling frequency; The step of extracting frequency domain features of vibration sensor data in the sensor layer includes: The time domain signal is subjected to fast Fourier transform to convert the signal from time domain to frequency domain, and the spectral centroid is extracted by the following:

[0059] f K For the Kth frequency component, A k is the corresponding amplitude.

[0060] The technical effect of the feature extraction algorithm used in this step S21 to extract the feature values from the preprocessed data is: The feature extraction algorithm can accurately extract key information reflecting the device state from the original sensor data while meeting the real-time constraints of the edge computing layer; reduce the time length of single-device multi-dimensional feature extraction, providing high-quality input for subsequent anomaly detection (isolation forest algorithm); the combination of frequency domain and time domain features enables early identification of device failures (such as bearing micro-cracks), providing sufficient time windows for predictive maintenance.

[0061] In the step S3: Based on the device three-dimensional size parameters (such as machine spindle diameter, motor power) and physical properties (such as material thermal conductivity, bearing rated speed) uploaded by the edge computing layer, a 1:1 scale virtual model is constructed using Unity or other three-dimensional modeling tools. By accessing device nameplate data (such as manufacturer, model, date of manufacture) and real-time feature data (such as vibration effective value, temperature gradient), the virtual model is bound with "physical properties - running state - life cycle" three-dimensional labels, forming a complete digital twin; The optimization engine module generates device-level and production line-level optimization strategies based on historical data and real-time data through machine learning and optimization algorithms.

[0062] The technical effect achieved by this step S3 is: The digital twin model realizes the "virtual-real mapping" of device state, and management personnel can intuitively master the running state of the entire production line through the cloud interface; the cloud platform layer not only receives real-time data from the edge computing layer, but also feeds back to the edge side through deep analysis and intelligent decision-making, forming a closed loop of "edge real-time response - cloud global optimization", effectively solving the technical problems of "monitoring lag, low optimization efficiency, and data island" in the prior art.

[0063] As Figure 2 shown, the system for real-time monitoring and optimization of device state based on industrial internet provided by the embodiment includes: A sensor layer module 1, which is used to collect device operation data; The sensor layer includes various types of Internet of Things sensors that collect key parameters of the device to be tested in real time, and determines the deployment position of the Internet of Things sensors according to the structural characteristics and key monitoring points of the device to be tested; an edge computing layer module 2 for processing data collected by the sensor layer; The edge computing layer includes an edge computing gateway that performs preprocessing, anomaly detection processing, and protocol conversion processing on data collected by the sensor layer. a cloud platform layer module 3 for storing, visualizing, and optimizing data; The cloud platform layer includes a digital twin module for constructing a digital twin model of a device asset and an optimization engine module for generating an optimization strategy for device process parameters through a machine learning algorithm.

[0064] In the sensor layer module 1: The sensor layer has an edge-side preprocessing unit built-in to perform preliminary filtering on raw collected data, and the collected data includes timestamp data synchronized with sensor collected data through a hardware clock synchronization protocol.

[0065] The sensor layer module 1 achieves the following technical effects: The edge-side preliminary filtering unit built-in to the sensor layer can directly filter out invalid noise such as electromagnetic interference and mechanical vibration clutter, significantly improving the signal-to-noise ratio of raw data and reducing the amount of invalid processing in the subsequent edge computing layer, thereby reducing processing pressure from the data source.

[0066] At the same time, the hardware clock synchronization protocol and timestamp ensure the time consistency of cross-sensor data, reducing the time deviation of multi-dimensional data (such as vibration, temperature, and current), and solving the state analysis distortion problem (such as misjudging that "vibration peak value is unrelated to temperature peak value") caused by different time synchronization in existing technologies.

[0067] The edge computing layer module 2 specifically includes: A data preprocessing submodule, in which: Two-stage filtering is performed on the data, with the first stage using a Kalman filter algorithm to dynamically eliminate measurement noise (such as mechanical vibration interference and electromagnetic noise), and the second stage using a Butterworth low-pass filter to filter out high-frequency interference. A feature extraction algorithm is used to extract feature values from the preprocessed data. A data anomaly detection processing submodule, in which: A rapid detection layer and an intelligent detection layer are arranged, the rapid detection layer compares pre-set thresholds (such as vibration amplitude > 0.1g and temperature > 80℃) with preprocessed data in real time, and the intelligent detection layer runs a lightweight isolation forest algorithm model to perform deep analysis on extracted feature values and identify implicit anomalies (such as weak fault harmonics appearing in vibration frequency spectrum). Protocol conversion processing submodule, in which: A multi-protocol analysis engine is built in the edge computing gateway, preset analysis rules are set for different types of protocols, and the analysis rules include: For general industrial protocols (such as Modbus RTU / TCP), the function code (such as 03H reading the holding register), data address and check bit (CRC check) in the protocol frame structure are read to directly analyze the physical quantity data such as temperature and current; For manufacturer proprietary protocols (such as Siemens S7 protocol), based on the communication specification disclosed by the protocol (such as the PDU data unit structure of S7COMM protocol), the binary data in the device specific data area (such as the input / output image area, timer / counter area) is converted into decimal or floating point type device running parameters; For OPCUA and other service-oriented protocols, the real-time state data (such as vibration amplitude, motor speed) of the device is obtained by calling the node (Node) attribute in its information model, and the data type (such as Boolean, double-precision floating point) is analyzed; The analyzed protocol data of various types is uniformly mapped to a preset standardized data model, and the standardized data model includes: device unique identifier (such as "machine tool-Fanuc-001"), data acquisition timestamp (accurate to milliseconds), data type (such as "vibration amplitude" "temperature"), data value (such as "0.08g" "65℃"), data quality code (such as "valid" "invalid"); The standardized data model is encapsulated into a message body through the MQTT protocol.

[0068] The technical effects obtained by the edge computing layer module are as follows: First, improve data quality and ensure the accuracy of subsequent analysis, through the synergistic effect of two-stage filtering - the first stage Kalman filter dynamically eliminates mechanical vibration interference, electromagnetic noise and other random noise, thereby improving the signal-to-noise ratio of the data, and the second stage Butterworth low-pass filter specifically filters out high-frequency interference (such as vibration clutter above 1kHz), retains the key feature frequency of device operation, and effectively avoids the interference of noise on feature extraction.

[0069] Second, reduce the burden of edge side preprocessing and improve the real-time performance of the system, filter and feature extraction are completed in the edge computing layer, avoiding the network bandwidth occupation caused by uploading raw data directly to the cloud (data volume is reduced by more than 60%), reducing the cloud storage and computing pressure.

[0070] Third, while improving the accuracy of anomaly detection and response speed, the fast detection layer is based on a preset threshold to achieve fast response, ensuring that emergency anomalies (such as device overheating) can be immediately captured, and the detection layer can detect hidden anomalies through a lightweight isolation forest algorithm to improve the identification accuracy of weak fault harmonics in the vibration spectrum (such as the characteristic frequency of early bearing cracks), and to achieve early warning of device failure, the two-level detection cooperates to reduce the false negative rate and false positive rate to, reduces the risk of device damage caused by false negatives and the loss of unplanned downtime caused by false positives.

[0071] Fourth, breaking the limitations of data silos and achieving full interoperability of heterogeneous devices, the multi-protocol parsing engine supports conversion of multiple types of protocols (such as Siemens S7, Mitsubishi MC, Modbus, OPCUA), covering mainstream manufacturer devices, significantly improving data interoperability, and the standardized data model unifies data formats (including device identification, timestamp, data type, etc. Core fields), enabling the running data of different manufacturer devices (such as the position parameters of ABB robots and the speed parameters of Fanuc machine tools) to be directly associated and analyzed on the cloud platform, providing a data foundation for cross-device collaborative optimization (such as production line load balancing scheduling). At the same time, after conversion to MQTT protocol, the bandwidth occupied by data transmission is reduced, and key data (such as device status at abnormal times) can be ensured not to be lost. And the standardized message body (JSON format) is more convenient for the cloud platform to quickly parse, reducing cloud data processing delay.

[0072] The data preprocessing submodule adopts a feature extraction algorithm to extract feature values from the preprocessed data, specifically including: a time domain feature extraction unit for vibration sensor data in the sensor layer and a frequency domain feature extraction unit for vibration sensor data in the sensor layer; The time domain feature extraction unit for vibration sensor data in the sensor layer includes: The effective value of vibration is calculated by the following formula:

[0073] In the formula, RMS is the effective value of the feature, x i is the amplitude of the i-th sampling point, and N is the number of sampling points in the sliding window; during feature extraction, the window sliding step is matched with the sensor sampling frequency; The frequency domain feature extraction unit for vibration sensor data in the sensor layer includes: The time domain signal is subjected to fast Fourier transform to convert the signal from the time domain to the frequency domain, and the spectral gravity center is extracted by the following formula:

[0074] f K is the K-th frequency component, A kCorresponding to the amplitude.

[0075] The technical effect of the feature extraction algorithm used in the data preprocessing submodule to extract feature values from the preprocessed data is: The feature extraction algorithm can accurately extract key information reflecting the device state from the original sensor data while meeting the real-time constraints of the edge computing layer; reduce the single-device multi-dimensional feature extraction time, providing high-quality input for subsequent anomaly detection (Isolation Forest algorithm); the combination of frequency domain and time domain features allows early identification of device faults (such as bearing micro-cracks), providing sufficient time windows for predictive maintenance.

[0076] In the cloud platform layer module 3: Based on the three-dimensional size parameters (such as machine tool spindle diameter, motor power) and physical properties (such as material thermal conductivity, bearing rated speed) uploaded by the edge computing layer, a 1:1 scale virtual model is constructed using Unity and other three-dimensional modeling tools. By accessing device nameplate data (such as manufacturer, model, and factory date) and real-time feature data (such as vibration effective value and temperature gradient), the virtual model is bound with "physical properties - running state - life cycle" three-dimensional labels, forming a complete digital twin; The optimization engine module generates device-level and production line-level optimization strategies based on historical data and real-time data through machine learning and optimization algorithms.

[0077] The technical effect of the cloud platform layer module 3 is: The digital twin model realizes the "virtual-real mapping" of device state, and management personnel can intuitively grasp the running state of the entire production line through the cloud interface; the cloud platform layer not only receives real-time data from the edge computing layer, but also feeds back to the edge side through deep analysis and intelligent decision-making, forming a closed loop of "edge real-time response - cloud global optimization", effectively solving the technical problems of "monitoring lag, low optimization efficiency, and data island" in the prior art.

[0078] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the method disclosed in the embodiments, since it corresponds to the system disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0079] Those skilled in the art will further appreciate that the functionality of the various examples described herein can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the

[0080] In several embodiments provided in the present application, it should be understood that the disclosed system, system and method can be implemented in other ways. For example, the above-described system embodiments are merely illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division, for example, multiple units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, systems or units, which can be electrical, mechanical or other forms.

[0081] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0082] In addition, the functional modules in each embodiment of the present application can be integrated in one processing unit, or each module can be physically present separately, or two or more modules can be integrated in one unit.

[0083] Similarly, each processing unit in each embodiment of the present application can be integrated in one functional module, or each processing unit can be physically present, or two or more processing units can be integrated in one functional module.

[0084] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0085] Finally, it needs to be explained that in this text, relational terms such as first and second and the like can merely be used to distinguish one entity or action from another, without necessarily requiring or implying any actual such relationship or order between or among the entities or actions. Also, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0086] The above disclosure is only the preferred embodiment of the present application, but the present application is not limited thereto, any non-creative change and several improvements and refinements that any person skilled in the art can think of without departing from the principles of the present application should fall within the protection scope of the present application.

Claims

1. A method for real-time monitoring and optimization of equipment status based on the Industrial Internet, characterized in that, Includes the following steps: Step S1: Deploy the sensor layer to collect equipment operation data; The sensor layer includes various types of IoT sensors, which collect key parameters of the device under test in real time and determine the deployment location of the IoT sensors based on the structural characteristics of the device under test and key monitoring points. Step S2: Deploy an edge computing layer to process the data collected by the sensor layer; The edge computing layer includes an edge computing gateway, which performs preprocessing, anomaly detection, and protocol conversion on the data collected by the sensor layer. Step S3: Deploy the cloud platform layer for data storage, visualization, and optimization; The cloud platform layer includes a digital twin module and an optimization engine module. The digital twin module is used to build digital twin models of equipment assets, while the optimization engine module generates optimization strategies for equipment process parameters through machine learning algorithms.

2. The method for real-time monitoring and optimization of equipment status based on the Industrial Internet according to claim 1, characterized in that, In step S1: The sensor layer has a built-in edge-side preprocessing unit that performs preliminary filtering on the raw acquired data. The acquired data includes timestamp data and is synchronized with the sensor data through a hardware clock synchronization protocol.

3. The method for real-time monitoring and optimization of equipment status based on the Industrial Internet according to claim 2, characterized in that, Step S2 specifically includes: Step S21: Data preprocessing step, in which: The data undergoes two-stage filtering. The first stage uses a Kalman filter algorithm to process the data and dynamically eliminate measurement noise. The second stage uses a Butterworth low-pass filter to process the data and remove high-frequency interference. Feature extraction algorithms are used to extract feature values ​​from the preprocessed data; Step S22: The step of data anomaly detection and processing, in which: A fast detection layer and an intelligent detection layer are deployed. The fast detection layer compares the preset threshold with the preprocessed data in real time. The intelligent detection layer runs a lightweight isolated forest algorithm model to perform in-depth analysis on the extracted feature values ​​and identify hidden anomalies. Step S23: Protocol conversion processing steps, in which: The edge computing gateway has a built-in multi-protocol parsing engine with preset parsing rules for different types of protocols. These parsing rules include: For general industrial protocols, physical quantity data can be directly parsed by reading the function code, data address, and check bit in the protocol frame structure. For vendor-specific protocols, based on the publicly available communication specifications of the protocol, the binary data of a specific data area of ​​the device is parsed and converted into decimal or floating-point device operating parameters; For service-oriented protocols, real-time device status data is obtained by calling node attributes in its information model and parsing the data type; The parsed protocol data is uniformly mapped to a preset standardized data model, which includes: unique device identifier, data acquisition timestamp, data type, data value, and data quality code (such as "valid" or "invalid"). The standardized data model is encapsulated into a message body using the MQTT protocol.

4. The method for real-time monitoring and optimization of equipment status based on the Industrial Internet according to claim 3, characterized in that, In step S21, a feature extraction algorithm is used to extract feature values ​​from the preprocessed data. Specifically, this includes the steps of extracting the time-domain features of the vibration sensor data in the sensor layer and the steps of extracting the frequency-domain features of the vibration sensor data in the sensor layer. The steps for extracting the time-domain features of vibration sensor data in the sensor layer include: The effective value of vibration is calculated using the following formula: In the formula, RMS is the effective value of the characteristic, x i Let N be the amplitude of the i-th sampling point, and N be the number of sampling points within the sliding window; during feature extraction, the window sliding step size is matched with the sensor sampling frequency; The steps for extracting the frequency domain features of vibration sensor data in the sensor layer are as follows: Performing a Fast Fourier Transform on the time-domain signal transforms it from the time domain to the frequency domain, and the spectral centroid is extracted as follows: f K For the Kth frequency component, A k This corresponds to the amplitude.

5. The method for real-time monitoring and optimization of equipment status based on the Industrial Internet according to claim 4, characterized in that, In step S3: Based on the device's 3D size parameters and physical attributes uploaded from the edge computing layer, a 1:1 scale virtual model is constructed using 3D modeling tools. By accessing the device nameplate data and real-time feature data, 3D labels of "physical attributes - operating status - life cycle" are bound to the virtual model to form a complete digital twin. The optimization engine module generates equipment-level and production line-level optimization strategies based on historical and real-time data through machine learning and optimization algorithms.

6. A system for real-time monitoring and optimization of equipment status based on the Industrial Internet, characterized in that, include: Sensor layer module, which is used to collect equipment operation data; The sensor layer includes various types of IoT sensors, which collect key parameters of the device under test in real time and determine the deployment location of the IoT sensors based on the structural characteristics of the device under test and key monitoring points. Edge computing layer module, which is used to process the data collected by the sensor layer; The edge computing layer includes an edge computing gateway, which performs preprocessing, anomaly detection, and protocol conversion on the data collected by the sensor layer. The cloud platform layer module is used for data storage, visualization, and optimization. The cloud platform layer includes a digital twin module and an optimization engine module. The digital twin module is used to build digital twin models of equipment assets, while the optimization engine module generates optimization strategies for equipment process parameters through machine learning algorithms.

7. A system for real-time monitoring and optimization of equipment status based on the Industrial Internet, as described in claim 6, is characterized in that... In the aforementioned sensor layer module: The sensor layer has a built-in edge-side preprocessing unit that performs preliminary filtering on the raw acquired data. The acquired data includes timestamp data and is synchronized with the sensor data through a hardware clock synchronization protocol.

8. A system for real-time monitoring and optimization of equipment status based on the Industrial Internet, as described in claim 7, is characterized in that... The edge computing layer module specifically includes: The data preprocessing submodule contains: The data undergoes two-stage filtering. The first stage uses a Kalman filter algorithm to process the data and dynamically eliminate measurement noise. The second stage uses a Butterworth low-pass filter to process the data and remove high-frequency interference. Feature extraction algorithms are used to extract feature values ​​from the preprocessed data; The data anomaly detection and processing submodule contains: A fast detection layer and an intelligent detection layer are deployed. The fast detection layer compares the preset threshold with the preprocessed data in real time. The intelligent detection layer runs a lightweight isolated forest algorithm model to perform in-depth analysis on the extracted feature values ​​and identify hidden anomalies. The protocol conversion processing submodule contains: The edge computing gateway has a built-in multi-protocol parsing engine with preset parsing rules for different types of protocols. These parsing rules include: For general industrial protocols, physical quantity data can be directly parsed by reading the function code, data address, and check bit in the protocol frame structure. For vendor-specific protocols, based on the publicly available communication specifications of the protocol, the binary data of a specific data area of ​​the device is parsed and converted into decimal or floating-point device operating parameters; For service-oriented protocols, real-time device status data is obtained by calling node attributes in its information model and parsing the data type; The parsed protocol data is uniformly mapped to a preset standardized data model, which includes: unique device identifier, data acquisition timestamp, data type, data value, and data quality code. The standardized data model is encapsulated into a message body using the MQTT protocol.

9. A system for real-time monitoring and optimization of equipment status based on the Industrial Internet, as described in claim 8, is characterized in that... The data preprocessing submodule employs a feature extraction algorithm to extract feature values ​​from the preprocessed data. Specifically, it includes: a unit for extracting time-domain features of vibration sensor data in the sensor layer and a unit for extracting frequency-domain features of vibration sensor data in the sensor layer. The unit for extracting the time-domain features of vibration sensor data in the sensor layer includes: The effective value of vibration is calculated using the following formula: In the formula, RMS is the effective value of the characteristic, x i Let N be the amplitude of the i-th sampling point, and N be the number of sampling points within the sliding window; during feature extraction, the window sliding step size is matched with the sensor sampling frequency; The unit for extracting the frequency domain characteristics of vibration sensor data in the sensor layer includes: Performing a Fast Fourier Transform on the time-domain signal transforms it from the time domain to the frequency domain, and the spectral centroid is extracted as follows: f K For the Kth frequency component, A k This corresponds to the amplitude.

10. A system for real-time monitoring and optimization of equipment status based on the Industrial Internet, as described in claim 9, is characterized in that... In the aforementioned cloud platform layer module: Based on the device's 3D size parameters and physical attributes uploaded from the edge computing layer, a 1:1 scale virtual model is constructed using 3D modeling tools. By accessing the device nameplate data and real-time feature data, 3D labels of "physical attributes - operating status - life cycle" are bound to the virtual model to form a complete digital twin. The optimization engine module generates equipment-level and production line-level optimization strategies based on historical and real-time data through machine learning and optimization algorithms.

Images