Remote monitoring and control method based on intelligent engineering management system

Abstract

The invention discloses a remote monitoring and control method based on an intelligent engineering management system, and relates to the technical field of the Internet of Things, and the method comprises the steps: carrying out the data collection of an engineering object, carrying out the weight adjustment of multi-modal data through a dynamic weight distribution algorithm, and carrying out the fusion, and obtaining an engineering state feature vector; constructing a prediction model by using historical data and the engineering state feature vector, and calculating and analyzing dynamic characteristics through phase-space reconstruction and a Lyapunov exponent to obtain a chaos prediction model; performing anomaly prediction on the engineering object, calculating a Lyapunov index and identifying an abnormal mode to obtain an anomaly prediction result; performing fault-tolerant analysis on the abnormal prediction result and the network state by using an AI model, predicting a fault risk, switching a control mode, and obtaining a fault-tolerant operation state; and analyzing the monitoring data by using an AI algorithm to generate a control instruction. According to the method, the data acquisition quality is adaptively optimized, so that the reliability under complex working conditions is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a remote monitoring and control method based on an intelligent engineering management system. Background Art

[0002] Existing engineering monitoring systems rely on fixed rules or single indicators to allocate weights for multimodal sensors, which makes it difficult to adapt to the dynamic fluctuations of sensor performance under complex working conditions. For example, when the signal quality of a vibration sensor drops sharply in a strong interference scenario, the static weight allocation mechanism still uses the preset value, resulting in distortion in the dynamic fusion of multimodal data.

[0003] When the equipment operating conditions change suddenly, the static parameter setting will miss the key dynamic characteristics, causing a sharp drop in prediction accuracy. The existing independent parameter calculation strategy is inefficient and does not work in depth with the prediction model, making it difficult to effectively identify early weak fault signals in industrial scenarios, resulting in delayed fault warning. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a remote monitoring and control method based on an intelligent engineering management system to solve the problems of real-time monitoring failure and fault warning lag caused by dynamic fusion distortion of multimodal data and parameter rigidity.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a remote monitoring and control method based on an intelligent engineering management system, which includes collecting data on an engineering object, and fusing multi-modal data by adjusting weights using a dynamic weight allocation algorithm to obtain an engineering state feature vector; The prediction model is constructed using historical data and engineering state feature vectors, and the chaos prediction model is obtained by analyzing the dynamic characteristics through phase space reconstruction and Lyapunov index calculation. Perform anomaly prediction on engineering objects, calculate Lyapunov exponents and identify anomaly patterns to obtain anomaly prediction results; Use AI models to perform fault-tolerant analysis on abnormal prediction results and network status, predict failure risks, and switch control modes to obtain a fault-tolerant operating state; Use 5G network to transmit and process engineering status feature vectors, prediction results, and fault-tolerant operation status, upload them to the cloud server, and obtain monitoring data; Use AI algorithms to analyze monitoring data and obtain control instructions; Edge devices are used to execute and process control instructions, adjust the status of engineering objects, collect feedback data and upload it to the cloud to obtain execution results.

[0007] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system described in the present invention, the dynamic weight allocation algorithm is: Perform real-time weight reduction operations based on sensor data quality assessment results; Combine the real-time operating parameters of the equipment to dynamically upgrade the weight of key physical quantity sensors; The weight-adjusted multimodal data are concatenated into an engineering status feature vector according to a preset dimension.

[0008] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system of the present invention, the specific steps of obtaining the chaos prediction model are as follows: Generate multi-dimensional trajectories of the device's dynamic characteristics based on phase space reconstruction; Determine the state of engineering chaos through Lyapunov exponent calculation; Dynamically switch the prediction model structure according to the chaotic state.

[0009] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system of the present invention, the specific steps of obtaining the engineering state feature vector are as follows: Extracting frequency domain feature subvectors of vibration signals; Extract time-domain statistical quantum vectors from temperature, pressure and humidity data; Extract the encoding sub-vector from the equipment operating condition parameters.

[0010] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system described in the present invention, when the sensor weight value is continuously lower than the activation threshold, it automatically switches to the redundant sensor data channel and resets the initial parameters of the dynamic weight allocation algorithm.

[0011] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system described in the present invention, the specific steps of the abnormal prediction process are as follows: Conduct multi-dimensional residual analysis on the predicted and actual values ​​of the engineering status characteristic vector; Generate abnormal trigger signals based on dynamic threshold judgment mechanism; Multi-dimensional feature extraction and risk level classification of abnormal patterns are performed through pre-trained classification models.

[0012] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system described in the present invention, the specific steps of the fault-tolerant analysis are as follows: Establish a joint probability mapping relationship between anomaly type, network delay and device load rate; Dynamically select control instruction generation nodes according to real-time joint probability values; When the edge node is selected to generate control instructions, the device status data is uploaded to the cloud synchronously for model parameter correction.

[0013] As a preferred solution of the remote monitoring and control method based on the intelligent engineering management system described in the present invention, wherein: according to the real-time bandwidth fluctuation of the 5G network, the full transmission mode of the engineering state feature vector and the differential compression transmission mode are dynamically switched; Construct a three-level data transmission queue, in which abnormal warning data is forced to occupy the transmission channel; A control instruction set within a safety tolerance range is pre-stored in the edge node, and the control instruction set is executed according to a preset priority when the network is interrupted.

[0014] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the remote monitoring and control method based on the intelligent engineering management system as described in the first aspect of the present invention is implemented.

[0015] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the remote monitoring and control method based on an intelligent engineering management system as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: through the dynamic weight fusion mechanism of multimodal sensors, the quality of data acquisition is adaptively optimized, and the reliability of system state perception under complex working conditions is significantly improved; the prediction model construction method based on chaotic dynamics theory breaks through the structural limitations of traditional prediction models, realizes dynamic characteristic driven model adaptive switching, and greatly improves the accuracy and adaptability of system state prediction; multi-dimensional joint probabilistic fault-tolerant control technology integrates abnormal pattern recognition and network state evaluation, establishes an edge-cloud collaborative decision-making mechanism, and effectively enhances the control robustness of the system under abnormal working conditions; the intelligent network transmission strategy ensures the real-time and integrity of key information transmission in a complex network environment through dynamic bandwidth allocation and multi-level data priority management; the closed-loop feedback optimization system iteratively updates the control model parameters in real time, continuously improves the system operation stability and self-optimization ability, and significantly reduces the risk of misjudgment. The synergistic effect of various technical links realizes the leapfrog performance improvement of the engineering management system in the dimensions of data perception, state prediction, risk disposal and remote control. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a schematic diagram of the remote monitoring and control method based on the intelligent engineering management system in Example 1.

[0019] Figure 2 This is a flow chart of multimodal data fusion in Example 1.

[0020] Figure 3 A flow chart for constructing the chaos prediction model in Example 1.

[0021] Figure 4 This is a flow chart of closed-loop control and fault-tolerant processing in Example 1. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1, reference Figure 1~Figure 4 , which is the first embodiment of the present invention, provides a remote monitoring and control method based on an intelligent engineering management system, comprising the following steps: S1. Collect data on the engineering object, and use a dynamic weight allocation algorithm to adjust the weights of multi-modal data to fuse them, and obtain the engineering state feature vector; Specifically, four types of sensors are deployed at the key physical nodes of the project object, including: XYZ-100A vibration sensor is installed at the motor bearing, with a range of ±50g and a resolution of 0.01g; 0-10MPa range pressure sensor is configured at the hydraulic pump outlet, and the non-linear error is <0.1%FS; PT100 temperature sensor is arranged inside the distribution cabinet, with an accuracy of ±0.5℃; and humidity sensor with an accuracy of ±2%RH is set at the pipeline connection.

[0026] The 5G module is used to synchronize the clocks of multiple sensors, ensuring that the timestamp alignment error does not exceed 1 millisecond. Each sensor collects data at a set frequency, with the vibration sensor sampling frequency of 100 Hz, the temperature sensor of 1 Hz, and the pressure sensor of 10 Hz.

[0027] The data preprocessing process is to use sliding average filtering for vibration signals with a window length of 50 sampling points, and median filtering for temperature and humidity data with a time window of 5 seconds. All data are processed by Z-score standardization. The dynamic weight calculation logic is that during the data quality assessment stage, when the sensor signal-to-noise ratio (SNR) is lower than 20dB, the sensor weight value is halved; if the actual sampling frequency of the vibration sensor is lower than 95Hz, its weight is directly reset to zero.

[0028] Condition matching stage: When the equipment load rate exceeds 90%, the weight of the vibration sensor increases to 0.8, and the weight of the temperature and humidity decreases to 0.1; when the load rate is less than 30%, the weight of the pressure sensor increases to 0.5. The weighted sensor data is spliced ​​into a 128-dimensional vector according to a 1-second time window: the first 64 dimensions store vibration spectrum features (FFT peak and energy entropy); the middle 32 dimensions record the time series statistics (mean and variance) of temperature, pressure, and humidity; the last 32 dimensions encode the condition label (load rate, operating mode) Furthermore, when the weight value of a sensor is zero for three consecutive times, the edge gateway automatically triggers an alarm and switches to the backup sensor. For example, when the main vibration sensor fails, the backup vibration sensor is immediately enabled to continue collecting data. The sensor data is connected to the HG-2000 edge gateway through the RS-485 bus, and the built-in FPGA chip is used to complete real-time filtering and weight calculation, and the processing delay is controlled within 2 milliseconds.

[0029] It should be noted that the XYZ-100A model was selected for the dynamic sensor because its 50g range and 0.01g resolution can accurately capture the mechanical vibration characteristics of the equipment. Compared with the traditional NTP protocol, 5G clock synchronization technology optimizes the time alignment error from 10 milliseconds to 1 millisecond. The vibration characteristics occupy 64 dimensions because their spectral characteristics are highly sensitive to the health status of the equipment. The working condition label is allocated 32 dimensions separately to strengthen the guiding role of parameters such as load rate on the weight strategy. NVIDIA Jetson Nano provides 4TFLOPS computing power to support real-time weight updates in a 5-second period, and FPGA hardware acceleration makes the filtering processing efficiency 15 times higher than the CPU solution.

[0030] S2. Use historical data and engineering state feature vectors to construct a prediction model, and analyze dynamic characteristics through phase space reconstruction and Lyapunov index calculation to obtain a chaos prediction model; Specifically, the engineering status feature vectors of the past six months were loaded from the cloud MySQL database, with a total of about 100,000 data items. Data cleaning includes missing value processing using linear interpolation, and the completion mechanism is activated when there are more than three consecutive missing items; abnormal data filtering directly eliminates records during equipment downtime and maintenance.

[0031] The method for obtaining reconstruction parameters is as follows: the embedding dimension is automatically optimized using the mutual information method and is finally set to 10 dimensions; the delay time is calculated based on the autocorrelation function and is fixed at 50 milliseconds.

[0032] The reconstruction process includes arranging the 128-dimensional engineering state feature vectors in time series to form an [N×128] matrix; embedding each feature dimension with time delay to generate a trajectory. The expression of the trajectory is as follows: in, is the embedding dimension, is the delay time, is the number of time points of the original data; All dimensional trajectories are merged to form a global phase space data set. The Wolf algorithm is used to calculate the maximum Lyapunov exponent λ, and the chaotic state judgment threshold is λ>0.05. The index value is recalculated every 30 minutes, and the prediction mode is switched dynamically.

[0033] The network structure includes the number of neurons in the input layer corresponding to the phase space dimension (10×128); the hidden layer adopts a bidirectional LSTM structure with 64 nodes and a tanh activation function; the output layer fully connected structure generates a 128-dimensional prediction vector.

[0034] The Adam optimizer is used, the initial learning rate is 0.001, the batch size is fixed at 32, the early stopping strategy monitoring window is 5 rounds, the loss function uses the mean square error, and the verification accuracy requires MSE < 0.01 Furthermore, when entering a stable state (λ≤0.05), it automatically switches to a lightweight ARIMA model, reducing computing resource consumption by 60%. The phase space reconstruction operation is accelerated through FPGA, and the single analysis time is compressed to less than 10 milliseconds, which is 8 times faster than the CPU solution.

[0035] It should be noted that the mutual information method was chosen to determine the embedding dimension because it can effectively capture the correlation of nonlinear features. Setting a 30-minute update cycle is a typical time scale for considering changes in equipment operating conditions. The bidirectional LSTM structure can simultaneously capture the dependencies between the previous and next time series, and the prediction accuracy is improved by 12% compared to the unidirectional structure. The use of Tesla V100 GPU training reduces the training time of 200 rounds of models to 90 minutes. When the Lyapunov exponent is calculated abnormally, the system automatically rolls back the most recent valid value and triggers an alarm to ensure the continuity of the prediction mode switching and avoid interruption of control instructions.

[0036] S3, perform abnormal prediction on the engineering object, calculate the Lyapunov index and identify the abnormal pattern to obtain the abnormal prediction result; Specifically, historical data is used to pre-train the LSTM model: the input layer is set to a structure that can receive 128-dimensional feature vectors for 6000 consecutive time steps; the data set is divided according to the equipment operation status, and the Dropout rate is preferably 0.2 to prevent overfitting; a bidirectional LSTM structure (32 units each in the forward and backward directions) is used as the hidden layer, and the Dropout rate is preferably 0.2 to prevent overfitting; the TimeDistributed fully connected layer and the linear activation function are combined as the output layer, and the weighted mse function is used as the training loss function.

[0037] The current 128-dimensional engineering status feature vector is input into the pre-trained LSTM model, and the predicted value for the next 30 seconds is output; the actual collected data is synchronized through the 5G module to ensure that the timestamp error does not exceed 10 milliseconds; residual calculation refers to calculating the absolute difference between the predicted value and the actual value item by item by dimension to generate a 128-dimensional residual vector. The baseline standard deviation is calculated based on normal data in the past 24 hours. The initial threshold interval is set to mean ±3 times the standard deviation (μ±3σ); the sliding window updates the standard deviation every 10 minutes, the window length is 300 seconds, and the step length is 60 seconds The abnormal trigger logic requires that the residual of any dimension exceeds the abnormal threshold three times in a row.

[0038] Feature extraction includes energy concentration, which refers to the proportion of residual energy in the first 10 dimensions of vibration, and the abnormal threshold is set to 60%; the temperature-pressure correlation is calculated by the Pearson coefficient, and the absolute value of the threshold is 0.3; the residual mutation gradient statistical change rate, the threshold is set to 5σ / second; the classification model uses a pre-trained SVM, the kernel function is RBF, and the regularization parameter C=1.0.

[0039] Risk levels are divided into three levels: high priority (sensor failure) triggers switching of redundant sensors within 5 seconds; medium priority (equipment overload) executes power limit instructions within 30 seconds; low priority (environmental interference) pushes alarm information within 60 seconds.

[0040] Furthermore, when the SVM classification confidence is lower than 90%, the manual review process is automatically triggered. The review data and classification results are written to the cloud log in real time, and suspicious records are marked. Environmental interference anomalies require synchronous verification of the consistency of the temperature and humidity sensors. When the data difference exceeds 15%, the secondary confirmation process is triggered. If the verification fails, it is downgraded to an unclassified anomaly.

[0041] It should be noted that the 3σ principle was chosen because it can cover 99.7% of normally distributed data, and the false alarm rate is controlled within 0.3%. The dynamic adjustment range is limited to 2σ-4σ to adapt to the drift characteristics of equipment conditions. SVM uses RBF kernel function to effectively handle nonlinear classification problems, and the 95% accuracy of the test set meets the needs of industrial scenarios. The 200MB memory usage of the edge computing node (Jetson XavierNX) ensures real-time operation of the model. The backup decision tree model is used as a disaster recovery solution and automatically switches when the SVM has low confidence for five consecutive times. The data closed-loop mechanism updates the model incrementally every week to continuously optimize the classification accuracy.

[0042] S4. Use AI models to perform fault-tolerant analysis on abnormal prediction results and network status, predict failure risks, and switch control modes to obtain a fault-tolerant operating state. Specifically, the input data includes abnormal classification results from S3 (sensor failure / equipment overload / environmental interference); real-time network status data (5G delay, bandwidth occupancy, packet loss rate).

[0043] The Bayesian network construction includes node definitions covering variables such as anomaly type, network delay, and equipment load rate; the conditional probability table is generated based on historical fault data statistics; and the risk level division adopts probability thresholds (low risk <30%, medium risk 30-70%, and high risk >70%).

[0044] Differentiated control is performed according to the risk level. In high-risk mode, edge local PID control is enabled and the motor speed is adjusted by ±5%. In medium-risk mode, cloud collaborative computing is started and the maximum power of the device is limited to 10%. In low-risk mode, the existing control instructions are maintained without adjustment.

[0045] The device status data after the control command is executed is collected every 500 milliseconds. The collected device status parameters include key indicators such as speed, temperature, and pressure. The feedback data is transmitted back to the cloud through the 5G network to update the Bayesian network Furthermore, when the Bayesian network outputs low confidence results three times in a row, it automatically switches to a rule-based risk assessment model; the backup model uses a decision tree algorithm, and the accuracy is maintained above 85%. The Bayesian network is incrementally trained every week, and the model update is triggered when the amount of new data exceeds 1,000; the historical data rolling window retains the records of the last three months.

[0046] It should be noted that the Bayesian network was chosen because it is good at dealing with multi-source uncertainty problems. The initial prior probability setting (sensor failure 0.3 / equipment overload 0.5 / environmental interference 0.2) is based on the historical statistics of equipment failures. The 5G delay threshold of 50ms is determined based on the real-time requirements of industrial control, and the motor speed adjustment range of ±5% is verified by dynamic simulation to ensure equipment safety. The edge computing node (Jetson Xavier NX) has a memory usage of 300MB to achieve real-time risk assessment; the mode switching response speed of 200ms meets the ISO 13849-1 safety standard requirements.

[0047] S5. Use 5G network to transmit and process engineering status feature vectors, prediction results and fault-tolerant operation status, upload them to the cloud server, and obtain monitoring data; Specifically, the engineering status feature vector, anomaly prediction results, and fault-tolerant operation status data are read from the local storage of the edge device; the engineering status feature vector format is a 64-dimensional float32 type array; the anomaly prediction results and fault-tolerant operation status are encapsulated as JSON objects.

[0048] Data compression processing includes using the Gzip algorithm to compress the merged data, with the compression rate controlled at 60%-70%; adding a 4-byte CRC32 checksum to verify data integrity.

[0049] Establishing a network connection includes accessing the 5G base station through the Quectel RM500Q module; obtaining a dynamic IP address and setting an APN access point; initializing the MQTT client to connect to the cloud proxy server.

[0050] Transmission reliability assurance includes automatic switching to the 4G backup channel when the signal strength is lower than -100dBm; initiating an exponential backoff retransmission strategy when no ACK confirmation is received, with a maximum of 5 retries.

[0051] After receiving the data, a CRC32 check is performed to verify the data integrity, with a check failure rate of <0.1%; Gzip decompression is used to restore the original JSON format; the data is stored in a MongoDB time series database with a write delay of <200ms; the digital twin model is updated to synchronize the latest status of the device.

[0052] The cloud generates a monitoring report every 15 minutes, and the characteristic vector trend chart shows the data fluctuations in the last hour; abnormal warning information is pushed to the administrator in real time via SMS / email; the monitoring report retention period is set to 90 days.

[0053] Furthermore, the transmission strategy is automatically adjusted according to the network conditions. When the bandwidth is sufficient, full data transmission (1 second interval) is enabled; when the bandwidth is tight, it switches to differential transmission mode (5 second interval). The transmission queue is divided according to the importance of the information, and abnormal warning data is set to the highest priority (immediate transmission).

[0054] Feature vector data has a medium priority (allowing 200ms delay), and running log data has a low priority (batch package transmission).

[0055] It should be noted that the MQTT protocol was chosen because of its lightweight characteristics, which is suitable for IoT scenarios and reduces network overhead by 60% compared to the HTTP protocol. The 4-byte CRC32 checksum design minimizes data packet expansion while ensuring a 99.99% error detection rate. The Quectel RM500Q module supports NSA / SA dual-mode 5G and maintains a stable connection within the industrial temperature range of -40°C to +85°C. The edge device storage uses SLC flash memory to ensure that data can be fully preserved for 72 hours in the event of a power outage. The 20ms end-to-end latency requirement of the 5G network drives the design of a dynamic bandwidth adjustment mechanism. The digital twin model is updated every 500ms, and the deviation from the physical device state is controlled within ±2%.

[0056] S6. Use AI algorithms to analyze monitoring data, obtain control instructions, use edge devices to execute and process control instructions, adjust the state of the engineering object, collect feedback data and upload it to the cloud to obtain execution results; Specifically, the LSTM model deployed in the cloud analyzes real-time monitoring data: the input data includes the device status feature vector for the past 10 minutes; the output is the predicted status and optimized control instructions for the next 10 minutes The command types are divided into parameter adjustment commands (such as motor speed ±3%) and operation mode switching commands (such as normal mode / energy-saving mode).

[0057] Control instructions are sent to edge devices through the 5G network, and the end-to-end transmission delay is strictly controlled within 50 milliseconds; actuator types include servo motors, proportional valves and other industrial-grade equipment; Feedback data collection includes equipment status parameters (temperature, vibration, pressure); instruction execution results (actual speed, valve opening); network transmission quality (delay, packet loss rate).

[0058] The cloud evaluates the execution effect of the instructions every 30 seconds. When the effect meets the standard, the current control strategy is continued. When the effect deviation exceeds 5%, a new optimization instruction is generated.

[0059] Furthermore, a real-time scoring system for instruction effects is established, with scoring indicators including energy efficiency, equipment stability, and production quality; a multi-objective optimization algorithm is triggered when the score is lower than 80 points.

[0060] When the 5G network interruption exceeds 200 milliseconds, the edge device automatically enables local cache instructions; if the execution exception occurs three times in total, the emergency shutdown protocol will be triggered.

[0061] This embodiment also provides a computer device, which is suitable for the remote monitoring and control method based on an intelligent engineering management system, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the remote monitoring and control method based on an intelligent engineering management system as proposed in the above embodiment.

[0062] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0063] The present embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the remote monitoring and control method based on the intelligent engineering management system proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0064] In summary, the present invention adopts: the dynamic weight fusion mechanism of multimodal sensors, adaptively optimizes the data acquisition quality, and significantly improves the reliability of system state perception under complex working conditions; the prediction model construction method based on chaotic dynamics theory breaks through the structural limitations of traditional prediction models, realizes dynamic characteristic driven model adaptive switching, and greatly improves the accuracy and adaptability of system state prediction; multi-dimensional joint probability fault-tolerant control technology integrates abnormal pattern recognition and network state evaluation, establishes edge-cloud collaborative decision-making mechanism, and effectively enhances the control robustness of the system under abnormal working conditions; intelligent network transmission strategy ensures the real-time and integrity of key information transmission in complex network environments through dynamic bandwidth allocation and multi-level data priority management; closed-loop feedback optimization system iteratively updates control model parameters in real time, continuously improves system operation stability and self-optimization ability, and significantly reduces the risk of misjudgment. The synergistic effect of various technical links realizes the leapfrog performance improvement of engineering management system in data perception, state prediction, risk disposal and remote control.

[0065] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A remote monitoring and control method based on an intelligent engineering management system, characterized in that: include, Collect data on engineering objects, and use a dynamic weight allocation algorithm to adjust the weights of multi-modal data for fusion to obtain the engineering status feature vector; The prediction model is constructed using historical data and engineering state feature vectors, and the chaos prediction model is obtained by analyzing the dynamic characteristics through phase space reconstruction and Lyapunov index calculation. Perform anomaly prediction on engineering objects, calculate Lyapunov exponents and identify anomaly patterns to obtain anomaly prediction results; Use AI models to perform fault-tolerant analysis on abnormal prediction results and network status, predict failure risks, and switch control modes to obtain a fault-tolerant operating state; Use 5G network to transmit and process engineering status feature vectors, prediction results, and fault-tolerant operation status, upload them to the cloud server, and obtain monitoring data; Use AI algorithms to analyze monitoring data, obtain control instructions, use edge devices to execute and process the control instructions, adjust the status of the engineering object, collect feedback data and upload it to the cloud to obtain execution results.

2. The remote monitoring and control method based on the intelligent engineering management system according to claim 1, characterized in that: The dynamic weight allocation algorithm is: Perform real-time weight reduction operations based on sensor data quality assessment results; The weights of key physical quantity sensors are dynamically upgraded based on the real-time operating parameters of the equipment.

3. The remote monitoring and control method based on the intelligent engineering management system according to claim 1, characterized in that: The specific steps of obtaining the engineering status feature vector are as follows: Extract the frequency domain characteristic sub-vector of the vibration signal, extract the time domain statistical sub-vector from the temperature, pressure and humidity data, extract the coding sub-vector from the equipment operating parameters, and adjust it according to the dynamic weight allocation algorithm; The weight-adjusted sub-vectors are concatenated into the engineering status feature vector according to the preset dimension.

4. The remote monitoring and control method based on the intelligent engineering management system according to claim 1, characterized in that: The specific steps of obtaining the chaos prediction model are as follows: Generate multi-dimensional trajectories of the device's dynamic characteristics based on phase space reconstruction; Determine the state of engineering chaos through Lyapunov exponent calculation; Dynamically switch the prediction model structure according to the chaotic state.

5. The remote monitoring and control method based on the intelligent engineering management system according to claim 1, characterized in that: The switching control mode means that when the sensor weight value is continuously lower than the activation threshold, it automatically switches to the redundant sensor data channel and resets the initial parameters of the dynamic weight allocation algorithm.

6. The remote monitoring and control method based on the intelligent engineering management system according to claim 5, characterized in that: The specific steps of the abnormal prediction process are as follows: Conduct multi-dimensional residual analysis on the predicted and actual values ​​of the engineering status characteristic vector; Generate an abnormal trigger signal based on a dynamic abnormal threshold determination mechanism; Multi-dimensional feature extraction and risk level classification of abnormal patterns are performed through pre-trained classification models.

7. The remote monitoring and control method based on the intelligent engineering management system according to claim 6, characterized in that: The specific steps of the fault-tolerance analysis are as follows: Establish a joint probability mapping relationship between anomaly type, network delay and device load rate; Dynamically select control instruction generation nodes according to real-time joint probability values; When the edge node is selected to generate control instructions, the device status data is uploaded to the cloud synchronously for model parameter correction.

8. The remote monitoring and control method based on the intelligent engineering management system according to claim 7, characterized in that: According to the real-time bandwidth fluctuation of the 5G network, dynamically switch between the full transmission mode of the engineering status feature vector and the differential compression transmission mode; Build a three-level data transmission queue; A control instruction set within a safety tolerance range is pre-stored in the edge node, and the control instruction set is executed according to a preset priority when the network is interrupted.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the remote monitoring and control method based on the intelligent engineering management system described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the remote monitoring and control method based on an intelligent engineering management system described in any one of claims 1 to 8 are implemented.

Images