Multi-dimensional intelligent inspection multi-device cooperative control method

Abstract

The invention relates to the technical field of intelligent inspection control, and discloses a multi-dimensional intelligent inspection multi-device cooperative control method. The method comprises the following steps: synchronously acquiring a real-time operation parameter set of a plurality of monitored devices, wherein the real-time operation parameter set comprises a vibration spectrum, a temperature distribution map and a current waveform; performing multi-source heterogeneous data fusion processing on the real-time operation parameter set to generate a multi-dimensional feature vector; based on a preset collaborative optimization target, performing dynamic association analysis on the multi-dimensional feature vector to generate an equipment operation feature vector; and performing a matching decision according to the equipment operation feature vector and a preset cooperative constraint condition, and outputting a cooperative control instruction set. According to the method, through multi-dimensional data acquisition, fusion processing and dynamic association analysis, comprehensive perception and cooperative control of the multi-device operation state are realized, the coordination and reliability of device group operation are improved, and the method is suitable for multi-device cooperative operation and maintenance in a complex scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent inspection control, in particular to a multi-dimensional intelligent inspection multi-device cooperative control method. BACKGROUND

[0002] In the fields of industrial production, power operation and maintenance, and intelligent manufacturing, the stable operation of devices is directly related to production efficiency and system safety. With the continuous improvement of industrial automation, the number of devices deployed in various production scenarios continues to increase, and the types of devices are increasingly diverse, from traditional mechanical transmission devices to precision electronic instruments, from high-voltage power devices to automated production lines, forming a complex multi-device cooperative operation system.

[0003] Currently, the inspection and control of these devices mostly adopt a single-device independent monitoring mode. For example, the monitoring of vibration parameters usually relies on a separate vibration sensor, temperature monitoring relies on an infrared temperature measuring device, and current monitoring is through a current transformer and other devices. The monitoring data of various types are transmitted to different processing systems, making it difficult to realize centralized integration and correlation analysis of the data. This decentralized monitoring method has obvious limitations: the operating states of different devices are related to each other, and the abnormality of a single parameter is often a manifestation of multi-device cooperative failure, which is difficult to capture through independent monitoring; the control instructions output by each monitoring system are usually only for a single device, lacking global coordination consideration, which may lead to action conflicts between devices and affect overall operation efficiency.

[0004] The processing of multi-source monitoring data in the prior art mostly stays at the simple summary level, failing to fully exploit the internal relationship between different types of data, resulting in low utilization rate of data value. In complex working conditions, the operating state of a device is affected by multiple factors, and the analysis result of a single data dimension cannot fully reflect the real state of the device, which may easily lead to misjudgment or missed judgment, thereby affecting the accuracy of control decisions. With the increasing intelligence of devices, the real-time and precision requirements of multi-device cooperative control are increasing, and the existing technology has been difficult to meet the efficient operation and maintenance needs in complex scenarios. SUMMARY

[0005] The purpose of the present application is to provide a multi-dimensional intelligent inspection multi-device cooperative control method to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a multi-dimensional intelligent inspection multi-device cooperative control method, which comprises:

[0007] S1: synchronously collecting a real-time operating parameter set of a plurality of monitored devices, the real-time operating parameter set comprising a vibration spectrum, a temperature distribution map, and a current waveform;

[0008] S2: performing multi-source heterogeneous data fusion processing on the real-time operation parameter set to generate a multi-dimensional feature vector;

[0009] S3: performing dynamic correlation analysis on the multi-dimensional feature vector based on a preset collaborative optimization target to generate a device operation feature vector;

[0010] S4: performing matching decision according to the device operation feature vector and a preset collaborative constraint condition to output a collaborative control instruction set.

[0011] Preferably, the multi-source heterogeneous data fusion processing on the real-time operation parameter set comprises:

[0012] Independent component extraction is performed on the vibration spectrum by using a blind source separation algorithm to separate the device body vibration feature and the environmental interference component;

[0013] The temperature distribution map is regionally segmented, and a thermal anomaly coefficient of each partition is calculated by using a heat conduction gradient model;

[0014] The harmonic distortion feature and the transient response component of the current waveform are analyzed by time-frequency transformation;

[0015] The device body vibration feature, the thermal anomaly coefficient, the harmonic distortion feature and the transient response component are combined into a multi-dimensional feature vector.

[0016] Preferably, the generation of the device operation feature vector comprises:

[0017] A device operation state space model is constructed, and the multi-dimensional feature vector is mapped to a coordinate point in the device operation state space;

[0018] The deviation degree of each coordinate point from a preset reference operation trajectory is calculated by using a grey correlation analysis algorithm;

[0019] The device operation feature vector is generated according to the deviation degree.

[0020] Preferably, the dynamic correlation analysis on the multi-dimensional feature vector based on the preset collaborative optimization target comprises:

[0021] A multi-device collaborative correlation matrix is established, and the row vectors of the collaborative correlation matrix correspond to the device operation feature vectors of the devices;

[0022] A core correlation mode of the collaborative correlation matrix is extracted by using a tensor decomposition algorithm;

[0023] The dynamic coupling weight between the devices is calculated according to the core correlation mode.

[0024] Preferably, the matching decision according to the device operation feature vector and the preset collaborative constraint condition comprises:

[0025] inputting the dynamic coupling weight into a constraint satisfaction model, the constraint satisfaction model including a load balancing threshold and an energy efficiency optimization boundary;

[0026] generating a load redistribution instruction when the device operation feature vector exceeds the load balancing threshold;

[0027] generating an operation parameter adjustment instruction when the device operation feature vector reaches the energy efficiency optimization boundary.

[0028] Preferably, the output set of collaborative control instructions further includes:

[0029] screening a set of key monitoring devices, the screening of the set of key monitoring devices being performed in the following manner:

[0030] calculating an entropy value index of the device operation feature vector of each device;

[0031] sorting according to the product of the dynamic coupling weight and the entropy value index, and selecting the top N devices in the sorting result as the set of key monitoring devices.

[0032] Preferably, the output set of collaborative control instructions includes:

[0033] performing sensitivity analysis on the device operation feature vector of the set of key monitoring devices to identify a dominant control dimension;

[0034] performing priority weighting on the load redistribution instruction and the operation parameter adjustment instruction according to the dominant control dimension to generate a set of collaborative control instructions.

[0035] Preferably, the method further includes:

[0036] constructing a collaborative control effect feedback loop, the execution of the collaborative control effect feedback loop being performed in the following manner:

[0037] after implementing the set of collaborative control instructions, re-collecting a set of feedback operation parameters of the monitored devices;

[0038] comparing the set of feedback operation parameters with a preset optimization target baseline in terms of residual error;

[0039] when the residual error exceeds a convergence tolerance, updating the collaborative constraint condition and triggering the matching decision of S4.

[0040] Preferably, the updating of the collaborative constraint condition and the triggering of the matching decision of S4 include:

[0041] extracting a high-frequency fluctuation component and a steady-state deviation component from the residual error;

[0042] separating a system inherent deviation from random interference through an adaptive filtering algorithm;

[0043] correcting the load balancing threshold and the energy efficiency optimization boundary according to the system inherent deviation.

[0044] Preferably, the output cooperative control instruction set is followed by:

[0045] Based on the device topology relationship network, the propagation path of the cooperative control instruction set is parsed;

[0046] A distributed consistency protocol is used to synchronize the instruction execution timing of multiple devices;

[0047] During the instruction execution process, the change gradient of the multi-dimensional feature vector is monitored in real time, and when the change gradient exceeds the dynamic response threshold, the instruction execution is interrupted and returned to S1.

[0048] Compared with the prior art, the beneficial effects of the present application are:

[0049] By synchronously collecting real-time running parameter sets of multiple monitored devices, including vibration spectrum, temperature distribution spectrum, current waveform and other types of key data, comprehensive perception of the device running state is achieved. This multi-dimensional data collection method breaks through the limitations of traditional single parameter monitoring and can more comprehensively reflect the actual running state of the device, providing rich basic information for subsequent analysis and control.

[0050] In the data processing link, a multi-source heterogeneous data fusion processing is used to generate a multi-dimensional feature vector, effectively integrating data information of different types and different sources, eliminating redundancy and conflicts between data, and mining potential associations between various parameters. This fusion processing not only improves the quality and consistency of the data, but also converts scattered information into a feature vector with comprehensive value, laying a foundation for accurate analysis of the device running state.

[0051] Based on the preset cooperative optimization target, a dynamic correlation analysis is performed on the multi-dimensional feature vector to generate a device running feature vector. This process fully considers the mutual influence and cooperative relationship between multiple devices and can grasp the running situation of the device group from a global perspective. Dynamic correlation analysis allows the system to adjust the analysis strategy in real time according to the changes in the device running state, timely capture the dynamic correlation between devices, and avoid the lag or deviation that may occur in static analysis.

[0052] According to the device running feature vector and the preset cooperative constraint condition, a matching decision is made and a cooperative control instruction set is output, ensuring the globality and cooperativeness of the control instruction. By matching the cooperative constraint condition, the output control instruction can take into account the running requirements of each device, avoiding the imbalance of the overall running caused by single device control, and realizing the coordinated action between multiple devices. This cooperative control method can make the device group maintain efficient and stable running state under complex working conditions, reduce the efficiency loss caused by improper cooperation between devices, and improve the running coordination and reliability of the entire system, suitable for complex scenarios of multiple device cooperative work. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the multi-dimensional intelligent inspection and multi-device collaborative control method described in this invention.

[0054] Figure 2 A flowchart for multi-source heterogeneous data fusion processing;

[0055] Figure 3 This is a flowchart for dynamic correlation analysis;

[0056] Figure 4 A flowchart for selecting a set of key monitoring devices;

[0057] Figure 5 A flowchart for updating collaborative constraints. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figure 1 This invention provides a multi-dimensional intelligent inspection and multi-device collaborative control method, the method comprising:

[0060] By simultaneously acquiring real-time operating parameter sets from multiple monitored devices, including vibration spectra, temperature distribution maps, and current waveforms, comprehensive monitoring of device operating status is achieved. The real-time operating parameter sets undergo multi-source heterogeneous data fusion processing to generate multi-dimensional feature vectors, which characterize the overall operating status of the devices. Based on preset collaborative optimization objectives, dynamic correlation analysis is performed on the multi-dimensional feature vectors to extract the correlation relationships between devices and generate device operating feature vectors. By matching the device operating feature vectors with preset collaborative constraints, a collaborative control command set is output to achieve optimized operation of multiple devices. This method effectively improves the stability and energy efficiency of collaborative device operation and is applicable to fields such as industrial automation and energy management.

[0061] Example 1: See Figure 2 This involves the fusion and processing of multi-source heterogeneous data and the generation of equipment operation feature vectors. The realization of this stage relies on the precise analysis of vibration spectra, temperature distribution maps, and current waveforms, combined with data fusion technology to construct multi-dimensional feature vectors that comprehensively reflect the equipment's operating status.

[0062] The analysis of vibration spectrum employs a blind source separation algorithm, which can effectively distinguish the device body vibration characteristics from environmental interference components. The vibration signals generated by the device during operation usually contain multiple components, including the inherent vibration of mechanical parts, external environmental noise, and possible abnormal vibration patterns. The blind source separation algorithm uses independent component analysis technology to decompose the mixed signal into several independent components, thereby extracting vibration characteristics related only to the device's own operating state. This process avoids the interference of environmental noise on vibration data analysis, making subsequent state evaluation more accurate.

[0063] The processing of temperature distribution map adopts region segmentation technology to divide the temperature distribution on the device surface into multiple local regions. The temperature data of each region is calculated through a heat conduction gradient model to analyze the trend and abnormalities of temperature changes. The heat conduction gradient model can identify local overheating or uneven temperature distribution, and then calculate the thermal anomaly coefficient of each partition. This coefficient reflects the operating state of the device in terms of thermodynamics and can be used to determine whether there are poor heat dissipation, local overheating, or other temperature-related potential problems.

[0064] The analysis of current waveform is achieved through time-frequency transform technology, which can capture both time-domain and frequency-domain characteristics of the current signal. Current waveform usually contains fundamental component, harmonic component, and transient response component. Time-frequency transform can effectively separate these components and extract harmonic distortion features and transient response components. Harmonic distortion features reflect the nonlinear characteristics of electrical systems, while transient response components can be used to judge the dynamic performance of electrical equipment. These features together constitute the key parameters of current waveform, providing important basis for subsequent device state evaluation.

[0065] After the feature extraction of vibration spectrum, temperature distribution map, and current waveform, these features are combined into a multi-dimensional feature vector. The multi-dimensional feature vector is a comprehensive data representation that covers information in multiple dimensions such as mechanical vibration, thermodynamic characteristics, and electrical performance.

[0066] The generation of device operating feature vector is based on the device operating state space model. This model maps the multi-dimensional feature vector to an abstract state space, where each coordinate point represents the operating state of the device at a certain time. The construction of the state space relies on historical normal operation data, and the baseline operating trajectory of the device under optimal working conditions is determined through machine learning or statistical analysis methods. The baseline operating trajectory represents the typical state change law of the device under normal conditions and is an important reference for judging whether the current operating state deviates from the normal range.

[0067] The grey relational analysis algorithm is used to calculate the deviation degree of the current device operating state from the benchmark operating trajectory. This algorithm quantifies the abnormality level of the device operating state by comparing the dynamic trends of multi-dimensional feature vectors. The advantage of grey relational analysis lies in its ability to handle data correlation problems with incomplete information, making it suitable for situations where data may be missing or noisy in industrial environments. The calculation results of the deviation degree are used to generate a device operating feature vector, which not only contains quantitative indicators of the current state of the device, but also reflects the degree of difference from the ideal operating state.

[0068] The generation process of the device operating feature vector is a dynamic and adaptive analysis process. As the device operating time progresses, the benchmark operating trajectory can be updated based on the latest normal operating data to adapt to possible slow changes such as aging or wear and tear. This dynamic adjustment mechanism enables the state evaluation model to maintain high accuracy, effectively identifying abnormal states even in long-term operation.

[0069] Embodiment 2: see Figure 3 , focusing on two key aspects of dynamic correlation analysis and matching decision. Dynamic correlation analysis aims to reveal the potential coupling relationship between devices, while matching decision generates specific collaborative control instructions based on the analysis results. The implementation of this stage relies on the construction of the multi-device collaborative correlation matrix, the application of tensor decomposition algorithm and the matching logic of the constraint satisfaction model.

[0070] The basis of dynamic correlation analysis is the construction of the multi-device collaborative correlation matrix. The row vectors of this matrix are composed of the operating feature vectors of each device, and each element of the matrix reflects the mutual relationship between different devices in the operating state. The construction of the collaborative correlation matrix enables the state correlation between devices to be presented in the form of structured data, facilitating subsequent mathematical analysis and pattern extraction. The tensor decomposition algorithm is used to extract the core correlation pattern from the collaborative correlation matrix, which can handle high-order data structures and mine common features in the operating state of the device. The extraction process of the core correlation pattern is realized through high-order singular value decomposition, which decomposes the original matrix into several low-rank components, representing the most significant coupling relationship between devices. The identification of the core correlation pattern enables the system to understand the mutual influence mechanism between devices, providing a theoretical basis for collaborative control.

[0071] The calculation of dynamic coupling weights is based on the extracted core correlation pattern. The influence of the operating state of each device on other devices is quantified by dynamic coupling weights, and the size of the weight value reflects the importance of the device in collaborative operation. The calculation of dynamic coupling weights considers factors such as the similarity of device operating features, time sequence correlation and physical connection relationship, ensuring the rationality of weight allocation. These weight values will serve as important input parameters for subsequent matching decisions, guiding the system in formulating control strategies in terms of load balancing and energy efficiency optimization.

[0072] The matching decision process employs a constraint satisfaction model that incorporates both load balancing thresholds and energy efficiency optimization boundaries as collaborative constraints. Load balancing thresholds define the allowable range of load differences between devices, beyond which some devices may operate under excessive load; energy efficiency optimization boundaries define the reasonable interval for device operating parameters, beyond which energy waste or performance degradation may occur. The dynamic coupling weights and device operating feature vectors are jointly input into the constraint satisfaction model, and matching analysis is performed through fuzzy logic rules. The advantage of fuzzy logic lies in its ability to handle the non-precise problems commonly found in industrial environments, such as parameter fluctuations and measurement errors, making the decision results more robust.

[0073] When the device operating feature vector exceeds the load balancing threshold, the system identifies the problem of uneven load distribution. This situation is usually manifested as the operating parameters of some devices being significantly higher than those of other devices, such as consistently high current values or abnormal temperature rises. In response to this situation, the system generates load redistribution instructions to redistribute the load among devices by adjusting their working modes or task assignments. The specific content of the load redistribution instructions may include adjusting the working frequency of devices, changing the task execution order, or enabling standby devices, among others, which aim to restore the load balancing state of the system.

[0074] When the device operating feature vector touches the energy efficiency optimization boundary, the system detects the situation where the operating parameters deviate from the optimal interval. This situation may manifest as increased energy consumption, decreased efficiency, or performance fluctuations of the device. In response to this situation, the system generates operating parameter adjustment instructions to fine-tune the control parameters of the device to return it to the high-efficiency operating interval. The operating parameter adjustment instructions may involve optimization of key parameters such as voltage, frequency, and speed, based on the characteristic curve and historical operating data of the device, ensuring the reasonableness and effectiveness of parameter changes.

[0075] The output of the matching decision is a set of collaborative control instructions, which includes a combination of load redistribution instructions and operating parameter adjustment instructions. The generation process of the instruction set takes into account the dynamic coupling relationship between devices, ensuring that control measures not only optimize individual devices but also consider the collaborative effect of the overall system. The execution order and priority of the instruction set are dynamically adjusted according to the real-time state of the devices, prioritizing control requirements that have the greatest impact on the system. This dynamic priority mechanism enables the system to quickly respond to changes in complex operating environments and maintain the stability of multi-device collaborative operation.

[0076] Example 3: see Figure 4, the screening of the set of key monitoring devices and the generation of the set of collaborative control instructions constitute the core link. The focus of work in this stage is to identify the key devices that have the greatest impact on the overall operation of the system from among the numerous monitored devices, and to generate accurate control instructions based on the state characteristics of these devices. The system quantifies the uncertainty of the device state by calculating the entropy value index of the device operation feature vector, and devices with higher entropy values usually exhibit more complex changes in operating state. The calculation of the entropy value index uses the following formula:

[0077]

[0078] wherein: represents the entropy value index of the i-th device, represents the probability distribution of the i-th device in the j-th feature dimension, and n represents the total number of feature vector dimensions. This formula quantifies the complexity of the device operating state through information entropy theory, and a higher entropy value indicates greater uncertainty in the device state, which requires closer attention.

[0079] The product of the dynamic coupling weight and the entropy value index is used for the priority ranking of the device. The dynamic coupling weight reflects the influence of the device in the collaborative operation network, while the entropy value index represents the fluctuation degree of the device's own state. The product of the two comprehensively considers the influence potential of the device on the system and the abnormality degree of its own state, providing a quantitative basis for the screening of key devices. The system ranks all devices in descending order according to the product calculation results, and selects the top-ranked devices to form the set of key monitoring devices. This set represents devices that have important influence on the overall operation of the system and may face operating state abnormalities, and is the object that the control strategy needs to prioritize.

[0080] Sensitivity analysis is carried out on the basis of the set of key monitoring devices, with the aim of identifying the dominant control dimensions. The analysis process examines the sensitivity of the device operation feature vector to each input parameter, and determines the most influential control variable by calculating the ratio of the change in the feature vector to the change in the parameter. Parameter dimensions with higher sensitivity usually correspond to key influencing factors of the device operating state, and control measures targeting these dimensions often produce more significant effects. The system determines the dominant control dimensions by comparing the sensitivity values of different parameter dimensions, and these dimensions will become the main basis for the subsequent development of control instructions.

[0081] The generation process of the collaborative control instruction set prioritizes the load redistribution instructions and the operating parameter adjustment instructions. The prioritization is mainly based on two aspects: one is the sorting position of the equipment in the key monitoring set, the higher the sorting position, the higher the priority of the control instruction; the other is the sorting result of the control dimension in the sensitivity analysis, the control instruction for the high sensitivity dimension obtains a higher weight. This double weighting mechanism ensures that control resources can be allocated to the most needed and most effective link. The final collaborative control instruction set is an ordered set containing multiple control measures, each instruction indicating the target equipment, control parameter, adjustment direction, and execution priority.

[0082] The generation of the instruction set also considers the collaborative relationship between the equipment. Due to the dynamic coupling effect between the key monitoring equipment, the control measures for a certain equipment may have a chain effect on other equipment. The system evaluates this potential impact when generating instructions and makes necessary adjustments to the instruction parameters to avoid conflicts between control measures. This coordination mechanism ensures the overall consistency of multi-equipment control, enabling the system to implement local adjustments while maintaining global optimization.

[0083] During implementation, the system continuously monitors the running state changes of the key monitoring equipment set. When a significant change in equipment state is found, the system recalculates the entropy value index and the dynamic coupling weight, and updates the composition of the key monitoring equipment set if necessary. This dynamic adjustment mechanism enables the system to adapt to changes in the operating environment and always maintain precise control over the most critical equipment. At the same time, the priority of the control instruction is also fine-tuned according to the real-time state, ensuring that system resources are always allocated to the most needed control link.

[0084] Embodiment 4: Refer to Figure 5 The establishment and operation of the collaborative control effect feedback loop constitute the core mechanism of the system's continuous optimization. This feedback loop dynamically adjusts the collaborative constraint conditions by monitoring the changes in equipment state after the execution of control instructions, achieving self-improvement of the control strategy. Taking a collaborative control system consisting of five industrial pump machines as an example, the implementation process of this embodiment is described in detail.

[0085] After the execution of the control instruction, the system re-collects the operating parameters of each pump machine to form a feedback operating parameter set. These parameters include vibration amplitude, bearing temperature, motor current, and other key indicators, collected at intervals of minutes. The comparison of key parameters of the five pump machines before and after the execution of a control instruction is shown in Table 1.

[0086] Table 1: Comparison of pump control instruction parameters before and after execution.

[0087] Equipment No. Vibration before control (mm / s) Vibration after control (mm / s) Temperature before control (°C) Temperature after control (°C) Current before control (A) Current after control (A) PM-101 4.2 3.8 68 65 42.5 41.2 PM-102 5.1 4.3 72 68 45.3 43.7 PM-103 3.9 3.5 65 63 40.8 39.6 PM-104 4.8 4.1 70 67 43.9 42.4 PM-105 4.5 3.9 69 66 44.1 42.8

[0088] The feedback operating parameter set is compared with the preset optimization target baseline in residual error. The optimization target baseline is a reference standard established according to historical normal operation data, including the ideal value range of each parameter. The system calculates the deviation of the actual value of each parameter from the baseline value to form a residual error data set. Taking the PM-102 device as an example, the baseline value of the vibration parameter is 3.8 mm / s, the actual value after control is 4.3 mm / s, and the vibration residual error is +0.5 mm / s; the baseline value of the temperature parameter is 65°C, the actual value after control is 68°C, and the temperature residual error is +3°C.

[0089] During the residual error analysis process, the system distinguishes between high-frequency fluctuation components and steady-state deviation components. High-frequency fluctuation components mainly reflect the instantaneous changes of parameters, which may be caused by temporary interference; steady-state deviation components reflect the persistent deviation of parameters, which often reflect systematic control deficiencies. In the case of the pump system, the temperature residual error of PM-102 shows a persistent positive deviation, indicating that the heat dissipation problem of the device has not been fundamentally solved and the control strategy needs to be further adjusted.

[0090] The adaptive filtering algorithm is used to separate the system inherent deviation and random interference. This algorithm identifies the deviation components that have regularity and persistence by analyzing the time series characteristics of the residual error. For the pump system, the algorithm finds that the vibration residual errors of PM-101 and PM-102 show periodic fluctuations at certain time periods, which are related to pipeline pressure fluctuations and belong to system inherent deviations; while the current residual error of PM-104 shows random discrete distribution, which is judged as random interference caused by measurement noise.

[0091] When the residual error exceeds the convergence tolerance, the system starts the cooperative constraint condition update mechanism. The convergence tolerance is the allowed deviation range set according to the device characteristics and process requirements. In the case of the pump, the convergence tolerance of the vibration parameter is ±0.3 mm / s, and that of the temperature parameter is ±2°C. Table 1 shows that the parameters of multiple pumps after control still exceed the tolerance range, triggering the constraint condition update process.

[0092] The correction of load balancing threshold and energy efficiency optimization boundary is based on the analysis results of system inherent deviation. For the pump system, it is found that the current load balancing threshold setting does not fully consider the performance differences of each pump, resulting in continuous overload of PM-102. The system adjusts the upper limit of PM-102's load from 46A to 44A according to the actual operation data, and correspondingly improves the load capacity allocation of other pumps. The energy efficiency optimization boundary is adjusted according to the temperature residual error, and the high temperature alarm threshold is reduced from 70°C to 68°C to warn potential heat dissipation problems in advance.

[0093] The updated collaborative constraints are re-input into the matching decision module to generate a new set of control instructions. In the pump system, the new instructions include: reducing the speed setpoint of PM-102, increasing the load sharing ratio of PM-103 and PM-104, starting the auxiliary cooling system, etc. These instructions, based on the consideration of the coupling relationship between devices, specifically address the identified systematic deviations.

[0094] The closed-loop operation of the feedback loop enables the system to continuously track the control effect. In the pump case, after the execution of the second round of control instructions, the system continues to collect data and analyze the residual error. When a new deviation pattern is found in the vibration parameters of PM-101, the system again starts the constraint condition update process, forming a continuous optimization control cycle. This closed-loop mechanism enables the system to adapt to the impact of long-term factors such as device aging and environmental changes, maintaining the effectiveness of the control strategy.

[0095] Example 5: The execution and dynamic monitoring of the collaborative control instruction set constitute the last safeguard link for the safe and stable operation of the system. The focus of this stage is to ensure the orderly propagation and coordinated execution of control instructions in a multi-device system, while monitoring the system response in real time during the execution process to prevent unexpected situations caused by control measures. Taking a distributed water supply pump station system as an example, which includes six main pump units, three standby units, and supporting pipe network pressure monitoring points, the method of example 5 is used to realize multi-pump collaborative control.

[0096] The construction of the device topology relationship network is the first step in the implementation. In the water supply pump station system, this network records the hydraulic connection relationship, electrical linkage relationship, and control signal transmission path between each pump. The main pumps PM1 to PM6 are connected in parallel through the main pipeline, and each pump is equipped with an independent frequency converter; the standby pumps SB1 to SB3 are connected to the main system through the cross-pipeline and can quickly cut in when any main pump fails. Control signals are transmitted through industrial Ethernet, forming a star topology. When the system analyzes the collaborative control instruction set, it will determine the propagation path and execution order of the instructions based on this topology relationship, such as adjusting the downstream pump first and then adjusting the upstream pump to avoid hydraulic impact.

[0097] The application of distributed consistency protocol ensures the synchronization of multi-device instruction execution. In the pump station system, when it is necessary to adjust the speed of multiple pumps simultaneously, the system uses an improved Paxos algorithm to ensure that all pumps complete parameter adjustment within the same control cycle. During the operation of the protocol, the main controller first makes a proposal, and the local controllers of each pump respond to the proposal and feedback the current state, and after multiple rounds of negotiation, an agreed execution plan is reached. This mechanism effectively avoids the problem of control asynchronization caused by network delay or device response difference, preventing water pressure fluctuations caused by inconsistent pump speed adjustment.

[0098] The synchronization of the instruction execution timing takes into account the device response characteristics and process requirements. For a water supply system, pressure stability is the primary control objective. When scheduling the timing of pump speed adjustments, the system prioritizes pumps that have a smaller impact on the network pressure, and then gradually adjusts the key node pumps. At the same time, the adjustment interval of adjacent pumps is controlled within 200 milliseconds, ensuring both smooth pressure transition and adjustment efficiency. This fine timing control makes the multi-pump collaborative adjustment process almost imperceptible to the user's water pressure fluctuations.

[0099] The real-time monitoring system continuously tracks the change gradient of the multi-dimensional feature vector during instruction execution. In the case of a pump station, the main parameters monitored include network pressure gradient, motor current rate of change, bearing vibration acceleration, etc. The system sets dynamic response thresholds for each parameter, which are not fixed values but safety boundaries dynamically calculated based on the current operating conditions. For example, during peak water supply periods, the pressure change gradient threshold will be more stringent than usual to ensure continuous stable water supply.

[0100] The interruption mechanism when the change gradient exceeds the dynamic response threshold is an important safety protection measure for the system. When the current rate of change of a pump is suddenly detected to exceed the threshold, the system will immediately suspend all pump adjustment instructions, maintain the current operating state, and trigger an exception diagnosis process. In the actual operation of the pump station system, this situation may occur when there is a sudden pipe burst in the network or a dramatic change in user water consumption. The interruption mechanism provides a buffer time for the system, preventing device damage or system collapse that may occur if control instructions continue to execute under abnormal conditions.

[0101] The processing flow after interruption includes returning to the real-time operating parameter acquisition phase. After interrupting the instruction execution, the system will immediately start a new round of data acquisition to obtain the latest network pressure, pump status, and other information. These data are used to analyze the cause of the interruption and determine whether it is caused by external interference or a problem with the control strategy. In the case of a pump station, if the analysis finds that the threshold is exceeded due to a sudden change in user water consumption, the system will recalculate the control parameters based on the new water consumption demand; if it is caused by equipment failure, the standby pump switching program will be started.

[0102] The decision to resume system operation is based on a comprehensive analysis of the multi-dimensional feature vector. After identifying the cause of the anomaly and taking appropriate measures, the system will evaluate whether the current state meets the recovery conditions. The evaluation includes multiple dimensions such as network pressure stability, pump load balance, and device health status. Only when all evaluation items are within the safe range will the system regenerate the control instruction set and continue to execute the unfinished adjustment tasks. This cautious recovery strategy minimizes the risk of secondary anomalies.

[0103] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it is intended to be limited only by the words recited in the appended claims. It is to be understood that the terms such as first and second, etc., merely are used to differentiate one from another without necessarily implying or requiring any actual relationship or order between them. Moreover, the terms "comprises", "comprising", or any other variations 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 include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0104] While the embodiments of the application have been shown and described herein, it is to be understood that the application is not limited to these embodiments. Rather, many modifications, changes, substitutions, and alterations can be made thereto without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional intelligent inspection and multi-device collaborative control method, characterized in that, include: S1: Simultaneously collect real-time operating parameter sets of multiple monitored devices, including vibration spectrum, temperature distribution spectrum and current waveform; S2: Perform multi-source heterogeneous data fusion processing on the real-time operating parameter set to generate a multi-dimensional feature vector; S3: Based on the preset collaborative optimization objective, perform dynamic correlation analysis on the multidimensional feature vector to generate the device operation feature vector; S4: Make matching decisions based on the equipment operation feature vector and preset collaborative constraints, and output a collaborative control instruction set.

2. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 1, characterized in that, The multi-source heterogeneous data fusion processing of the real-time operating parameter set includes: A blind source separation algorithm is used to extract independent components from the vibration spectrum, separating the vibration characteristics of the equipment body from the environmental interference components. The temperature distribution map is divided into regions, and the thermal anomaly coefficient of each region is calculated using the heat conduction gradient model. The harmonic distortion characteristics and transient response components of the current waveform are analyzed by time-frequency transformation. The vibration characteristics, thermal anomaly coefficient, harmonic distortion characteristics, and transient response components of the device body are combined into a multidimensional feature vector.

3. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 2, characterized in that, The generated device operation feature vector includes: Construct a device operating state space model and map the multidimensional feature vectors to coordinate points in the device operating state space; The deviation of each coordinate point from the preset benchmark trajectory is calculated using the grey relational analysis algorithm. The device operation feature vector is generated based on the deviation.

4. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 3, characterized in that, The dynamic correlation analysis of the multidimensional feature vector based on the preset collaborative optimization objective includes: Establish a multi-device collaborative association matrix, wherein the row vectors of the collaborative association matrix correspond to the device operation feature vectors of each device; The core association patterns of the collaborative association matrix are extracted using the tensor decomposition algorithm; Calculate the dynamic coupling weights between devices based on the core association pattern.

5. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 4, characterized in that, The matching decision based on the device's operating feature vector and preset collaborative constraints includes: The dynamic coupling weights are input into the constraint satisfaction model, which includes a load balancing threshold and an energy efficiency optimization boundary. When the device's operating characteristic vector exceeds the load balancing threshold, a load redistribution instruction is generated. When the device's operating feature vector reaches the energy efficiency optimization boundary, an operating parameter adjustment command is generated.

6. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 5, characterized in that, The output cooperative control instruction set also includes: The selection of key monitoring equipment sets is carried out using the following method: Calculate the entropy index of the equipment operation feature vector of each device; The devices are sorted according to the product of dynamic coupling weight and entropy index, and the top N devices in the sorting result are selected as the set of key monitoring devices.

7. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 6, characterized in that, The output cooperative control instruction set includes: Sensitivity analysis is performed on the equipment operation feature vectors of the key monitoring equipment set to identify the dominant control dimension; Based on the dominant control dimension, load redistribution instructions and operating parameter adjustment instructions are prioritized and weighted to generate a collaborative control instruction set.

8. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 7, characterized in that, The method further includes: A collaborative control effect feedback loop is constructed, and the execution method of the collaborative control effect feedback loop is as follows: After implementing the collaborative control instruction set, the feedback operating parameter set of the monitored equipment is re-collected; The feedback set of operating parameters is compared with the preset optimization target baseline using residuals. When the residual exceeds the convergence tolerance, the collaborative constraints are updated and the matching decision of S4 is triggered.

9. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 8, characterized in that, The updating of collaborative constraints and triggering of the matching decision in S4 includes: Extract the high-frequency fluctuation component and steady-state deviation component from the residual; The system's inherent bias and random disturbances are separated using an adaptive filtering algorithm; The load balancing threshold and energy efficiency optimization boundary are adjusted based on the inherent bias of the system.

10. The multi-dimensional intelligent inspection and multi-device collaborative control method according to claim 1, characterized in that, The output cooperative control instruction set is followed by: The propagation path of the collaborative control instruction set is analyzed based on the device topology network. A distributed consensus protocol is used to synchronize the instruction execution timing of multiple devices. The gradient of the multidimensional feature vector is monitored in real time during instruction execution. When the gradient exceeds the dynamic response threshold, instruction execution is interrupted and S1 is returned.

Images