CN-122027668-A - Industrial Internet of things equipment monitoring system and method based on self-adaptive sampling

CN122027668ACN 122027668 ACN122027668 ACN 122027668ACN-122027668-A

Abstract

The invention discloses an industrial Internet of things equipment monitoring system and method based on self-adaptive sampling, and relates to the technical field of industrial Internet of things, wherein the method comprises the steps of collecting and preprocessing equipment data by an edge node, and calculating abnormal characteristic indexes; when the index exceeds a local threshold, generating an event trigger signal and pushing the event trigger signal to a management platform, synchronously collecting the edge side resource state, calculating and determining the self-adaptive sampling frequency of the next period, and receiving multi-node information by the management platform to perform global strategy optimization and issuing an adjustment instruction. The system comprises a management platform, a sensing network platform and a perception control platform, wherein the perception control platform comprises an edge data acquisition and preprocessing module, the sensing network platform comprises an event triggering and feature extraction module, a dynamic sampling frequency adjustment model and a resource perception module, and the management platform comprises a central policy cooperation and decision module. The invention can realize the optimal balance of the utilization efficiency of the system resources while improving the monitoring instantaneity and the abnormal capturing capability.

Inventors

SHAO ZEHUA
LI YONG
HE LEI
WEI YONG

Assignees

成都秦川物联网科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (9)

1. The industrial Internet of things equipment monitoring system based on the self-adaptive sampling is characterized by comprising a management platform, a sensing network platform and a perception control platform which are sequentially connected in a communication mode; the perception control platform comprises: The edge data acquisition and preprocessing module is deployed at an edge computing node of an industrial field and is used for acquiring original operation parameters of target industrial equipment in real time at an initial sampling frequency, filtering, denoising and normalizing the original operation parameters to generate a normalized equipment state time sequence data stream; The sensing network platform comprises: The event triggering and feature extraction module is deployed at the edge computing node and is used for receiving the standardized equipment state time sequence data stream in real time, calculating an abnormal feature index of a current data window based on a preset lightweight abnormal detection algorithm, generating an event triggering signal containing the abnormal feature index and a corresponding timestamp when the abnormal feature index exceeds a preset local triggering threshold value, and pushing the event triggering signal and an associated original data fragment to the management platform in real time through a streaming data channel; The dynamic sampling frequency adjustment model is deployed at the edge computing node and is used for receiving the abnormal characteristic index, the equipment history state data, the edge side network bandwidth utilization rate, the CPU load and the memory occupancy rate which are monitored in real time by the resource sensing module, and dynamically computing and outputting the recommended sampling frequency of the next sampling period according to a preset optimization objective function; The resource sensing module is deployed on the edge computing node and is used for periodically collecting and quantifying the computing resource load state of the edge computing node and the network communication link quality between the edge computing node and the management platform to generate a resource state vector; The management platform comprises: the central policy coordination and decision module is deployed in a cloud or factory level data center and is used for receiving event trigger signals and resource status reports from a plurality of edge nodes, performing global situation analysis and policy optimization, generating global sampling policy adjustment instructions and transmitting the global sampling policy adjustment instructions to the corresponding edge nodes so as to coordinate resource competition and monitoring priority among multiple devices.
2. The adaptive sampling-based industrial internet of things device monitoring system according to claim 1, wherein the lightweight anomaly detection algorithm adopts a composite detection mechanism based on statistical process control and a moving window, and obtains a mean value and a standard deviation baseline of key operation parameters through historical data training in a steady state operation stage of the device, maintains a sliding data window with a length of N in a real-time monitoring stage, calculates a Z-score statistic of data in the window or a control limit based on an exponentially weighted moving average, wherein the anomaly characteristic index is defined as a deviation degree of a current window statistic relative to the steady state baseline, and the calculation formula is as follows: Wherein, the Is the first The parameter values of the individual sampling points, And The steady state mean and standard deviation baseline for the parameter respectively, Is an abnormal characteristic index at the time t.
3. The adaptive sampling-based industrial internet of things device monitoring system of claim 1, wherein the dynamic sampling frequency adjustment model is a multiple-input single-output regression model, and the input feature vector comprises an abnormal feature index at the current time Trend of change of the index over past M time windows, device history run mode encoding, and network bandwidth utilization provided by resource awareness module CPU load And memory occupancy rate The output of the model recommends a sampling frequency for the next period Relative to the fundamental frequency Is of the adjustment coefficient of (2) I.e. Wherein Is constrained to a range of values The expression of the optimization objective function is as follows: Wherein, the And In order to trade-off the coefficients, The zero is prevented from being divided for a very small constant, 、、 Is a resource weight.
4. The adaptive sampling-based industrial internet of things device monitoring system of claim 3, wherein the event triggering and feature extraction module comprises a dual threshold comparator and a data buffer queue, the local triggering threshold comprising a lower pre-warning threshold And a higher emergency threshold And (2) and When abnormal characteristic index Exceeding the limit But is lower than When the module generates event trigger signals of early warning level and caches the current and the later data fragments for a period of time, when Exceeding the limit And when the module immediately generates an event trigger signal of an emergency level, empties the cache queue and pushes the complete abnormal data fragments including the historical cache data with the highest priority.
5. The system of claim 4, wherein the central policy coordination and decision module performs global policy optimization by receiving resource status vectors and event trigger histories of all online edge nodes, constructing a global resource competition graph in terms of factories or production lines, identifying network bottleneck areas or computing resource overload nodes, and adjusting coefficients for sampling frequencies of low priority devices in resource shortage areas based on a preset device critical priority list Applying additional down-regulation penalty, and simultaneously, for the equipment area continuously reporting the high-level event trigger signal, instructing the related edge node to temporarily promote the dynamic sampling frequency adjustment model of the related edge node Upper limit of (2) And coordinates the neighboring nodes to release network bandwidth resources for them.
6. The adaptive sampling-based industrial internet of things device monitoring system of claim 1, wherein the resource awareness module implements resource quantification by invoking an operating system interface and a network probe tool, the resource status vector comprising a network bandwidth utilization CPU load Memory occupancy rate Wherein the network bandwidth utilization The CPU load is estimated by periodically sending test data packets to the management platform and calculating the round trip delay and the packet loss rate Is the duty ratio of CPU busy time in the latest sampling period, memory occupancy ratio Is the ratio of used memory to total memory.
7. The adaptive sampling-based industrial internet of things device monitoring system according to claim 1, wherein the streaming data channel adopts a publish/subscribe mode based on a message queue, an edge node is used as a producer to package event trigger signals and data into messages with specific topics for publication, and the management platform is used as a consumer to subscribe to the topics of all the edge nodes.
8. An industrial Internet of things equipment monitoring method based on adaptive sampling is characterized by comprising the following steps of: S110, an edge computing node collects original operation parameters of target industrial equipment at an initial sampling frequency, and preprocesses the original operation parameters to generate a standardized equipment state time sequence data stream; S120, based on the standardized equipment state time sequence data stream, calculating abnormal characteristic indexes of a current data window in real time by using a lightweight abnormal detection algorithm, and judging whether the abnormal characteristic indexes exceed a local trigger threshold; s130, when the abnormal characteristic index exceeds the local trigger threshold, immediately generating an event trigger signal, and pushing the signal and associated data to a management platform in real time through a streaming channel; s140, synchronously acquiring the network bandwidth utilization rate, the CPU load and the memory occupancy rate at the edge side to form a resource state vector; s150, inputting the abnormal characteristic index, the equipment history state data and the resource state vector into a dynamic sampling frequency adjustment model, and calculating and determining the self-adaptive sampling frequency of the next sampling period; s160, the management platform receives event trigger signals and resource state information from a plurality of edge nodes, performs global policy optimization, and generates and issues sampling policy adjustment instructions to the corresponding edge nodes; and S170, the edge computing node updates the sampling frequency of data acquisition according to the received self-adaptive sampling frequency or the sampling strategy adjustment instruction, and enters the next monitoring period.
9. The method for monitoring industrial internet of things equipment based on adaptive sampling according to claim 8, wherein in step S120, the method for calculating the abnormal characteristic index comprises obtaining a steady-state mean value and a standard deviation baseline of a key operation parameter through historical data training in a steady-state operation stage of the equipment, extracting parameter values of sampling points in a current sliding data window with a length of N in a real-time monitoring stage, respectively calculating a difference value of each parameter value and the steady-state mean value, dividing the difference value by the standard deviation baseline to obtain a standard deviation, squaring and summing the standard deviation of all sampling points in the sliding data window, dividing the summed result by the total number of sampling points of N, and obtaining the abnormal characteristic index representing the deviation degree of current window data relative to the historical steady-state baseline.

Description

Industrial Internet of things equipment monitoring system and method based on self-adaptive sampling Technical Field The invention belongs to the technical field of industrial Internet of things, and particularly relates to an industrial Internet of things equipment monitoring system and method based on self-adaptive sampling. Background The industrial Internet of things is used as a core support for intelligent manufacturing and digital transformation, is widely applied to the fields of equipment state monitoring, predictive maintenance, production process optimization and the like, and has the core that accurate sensing and intelligent control on physical equipment are realized through high-timeliness and high-reliability data acquisition and analysis. In the system, the equipment monitoring system plays a key role of capturing operation parameters in real time, identifying abnormal behaviors and triggering a response mechanism, and the performance of the equipment monitoring system directly determines the stability and the operation and maintenance efficiency of the whole industrial system. The industrial Internet of things equipment monitoring method based on the self-adaptive sampling focuses on dynamically adjusting a data acquisition strategy to match the change of the running state of equipment, and aims to optimize the use efficiency of communication and computing resources while guaranteeing the abnormal detection sensitivity. The core goal of the direction is to break through the limitation of the traditional static or periodic sampling mechanism and construct an intelligent monitoring architecture capable of automatically adjusting the sampling behavior according to the real-time working condition. In the prior art, an instruction-driven data acquisition mode is generally adopted, namely, after a sampling instruction is issued periodically or in an event triggering mode by an upper platform, the edge equipment can upload data, so that unavoidable delay exists between the occurrence of abnormality and the perception of a system, meanwhile, the sampling frequency is set by a plurality of fixed rules or manual experience, dynamic optimization cannot be carried out according to the current running state of the equipment, network bandwidth or edge computing load, redundant data transmission and resource waste are caused when the equipment is in a steady state, and key characteristics are omitted due to insufficient sampling when sudden abnormality or abrupt change of working conditions, so that the real-time response capability and the self-adaption level of a monitoring system are severely restricted. The above problems are particularly prominent in complex industrial scenarios with high dynamics and strong interference, and a new monitoring paradigm integrating event triggering mechanism and resource sensing capability is needed to realize low-delay, high-efficiency and strong-adaptive device state sensing. Disclosure of Invention The invention aims to make up the defects of the prior art, provides an industrial Internet of things equipment monitoring system and method based on self-adaptive sampling, and can effectively solve the problems in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: On one hand, an industrial Internet of things equipment monitoring system based on self-adaptive sampling is provided, and the system comprises a management platform, a sensing network platform and a perception control platform which are sequentially in communication connection; the perception control platform comprises: The edge data acquisition and preprocessing module is deployed at an edge computing node of an industrial field and is used for acquiring original operation parameters of target industrial equipment in real time at an initial sampling frequency, filtering, denoising and normalizing the original operation parameters to generate a normalized equipment state time sequence data stream; The sensing network platform comprises: the event triggering and feature extraction module is deployed at the edge computing node, connected with the edge data acquisition and preprocessing module, and used for receiving the standardized equipment state time sequence data stream in real time, calculating an abnormal feature index of a current data window based on a preset lightweight abnormal detection algorithm, generating an event triggering signal containing the abnormal feature index and a corresponding timestamp when the abnormal feature index exceeds a preset local triggering threshold value, and pushing the event triggering signal and an associated original data segment to a management platform in real time through a streaming data channel; The dynamic sampling frequency adjustment model is deployed at the edge computing node, the input end of the dynamic sampling frequency adjustment model is connected with the event triggering and fea