Automatic computer control method and system based on artificial intelligence

Abstract

The invention relates to the technical field of computer control, and discloses an automatic computer control method and system based on artificial intelligence. The method comprises the following steps: firstly, acquiring multi-source heterogeneous sensor data, including environmental parameters, equipment running states and operation instruction logs; then, an environment state evolution model is constructed, an equipment health degree evaluation index is generated, and an operation instruction log is analyzed; and then, inputting the results into a multi-modal decision model to generate a dynamic control instruction set, constructing a control strategy topological graph through an adaptive optimization algorithm, and outputting an equipment regulation and control scheme. The system correspondingly comprises a multi-source data acquisition module, an environment modeling module, a health assessment module and the like. Multi-source data can be effectively processed, the health condition of equipment can be accurately evaluated, a control instruction and a regulation and control scheme can be intelligently generated, the intelligent level of computer control is improved, and the reliability and the operation efficiency of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer control technology, and particularly to an automated computer control method and system based on artificial intelligence. Background Art

[0002] In today's digital age, the application scenarios of computer systems have become increasingly complex and diverse. From the automated control in industrial production, to the large-scale equipment operation and maintenance in data centers, and then to the environmental and equipment management in intelligent buildings, unprecedented challenges have been posed to the accuracy, efficiency, and intelligence of computer control. Traditional computer control methods have been difficult to meet these growing demands and have exposed many drawbacks.

[0003] Early computer control mainly relied on manual operation. Operators needed to monitor the device status in real time and issue control instructions based on experience. This method not only consumed a large amount of manpower, but was also extremely vulnerable to human factors such as fatigue and negligence, resulting in low control accuracy and slow response speed, and was difficult to cope with complex and changeable working scenarios. For example, in a large data center, manually inspecting the server status and manually adjusting the device parameters was inefficient and could not detect potential fault hazards in a timely manner. Once a problem occurred, it might cause serious consequences such as data loss and service interruption.

[0004] With the development of technology, rule-based automated control methods have emerged. These methods achieve device control by presetting a series of fixed rules, which improve the control efficiency and accuracy to a certain extent. However, their flexibility and adaptability are poor, and they are unable to cope with complex dynamic environments and emergencies. Taking an industrial automated production line as an example, when factors such as temperature and humidity in the production environment change, or when there are slight performance fluctuations in the equipment, the rule-based control system is difficult to make timely and reasonable adjustments, which may lead to a decline in product quality, a reduction in production efficiency, and even cause equipment failures.

[0005] In addition, existing computer control methods have deficiencies in processing multi-source heterogeneous data. During the operation of a computer system, a large amount of environmental parameters, device operation status data, and operation instruction logs from different sensors are generated. These data have diverse formats and complex structures. Traditional methods often cannot fully integrate and utilize these data, resulting in a serious information island phenomenon and being unable to provide comprehensive and accurate bases for control decisions. For example, in an intelligent building, the temperature and humidity data collected by environmental sensors are independent of the energy consumption data of device operation, and cannot be analyzed collaboratively, so it is impossible to achieve optimal energy management and intelligent control of devices.

[0006] In terms of equipment health management, traditional methods usually rely on regular manual inspections and simple threshold alarm mechanisms. This approach cannot evaluate the health status of equipment in real time and accurately, making it difficult to predict potential failures. Often, repairs are carried out only after equipment failures occur, resulting in production interruptions and economic losses. For example, when there are slight signs of failure in a server hard drive, if not detected and measures taken in a timely manner, it may lead to complete damage of the hard drive and data loss. Summary of the Invention

[0007] The purpose of the present invention is to provide an automated computer control method and system based on artificial intelligence to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: An automated computer control method based on artificial intelligence, the method includes: Obtain a multi-source heterogeneous sensor data set; the sensor data includes environmental parameters, equipment operation status, and operation instruction logs; Based on the environmental parameters, construct an environmental state evolution model through a dynamic space modeling algorithm, and the evolution model includes a parameter correlation matrix and an abnormal fluctuation prediction threshold; According to the equipment operation status, generate an equipment health assessment index through a frequency domain feature extraction algorithm, and the assessment index includes a failure probability value and a performance degradation rate; Perform time series correlation analysis on the operation instruction logs to generate an operation logic chain and an instruction conflict detection result; Input the environmental state evolution model, equipment health assessment index, and operation logic chain into a multi-modal decision-making model to generate a dynamic control instruction set; Based on the dynamic control instruction set, construct a control strategy topology diagram through an adaptive optimization algorithm, and output an equipment regulation plan; the nodes of the control strategy topology diagram represent control action units, and the edges represent action priorities and execution condition constraints.

[0009] Preferably, the construction of the environmental state evolution model through the dynamic space modeling algorithm includes: Perform noise filtering on the environmental parameters to generate a standardized environmental data sequence; Based on a preset environmental benchmark parameter library, generate parameter correlation weights through covariance matrix analysis; Use a time series prediction algorithm to calculate the change trend of environmental parameters in the future period, and combine with the abnormal fluctuation prediction threshold to generate a parameter deviation degree; Encode the parameter correlation weight, parameter deviation degree, and abnormal fluctuation prediction threshold into an environmental state evolution model.

[0010] Preferably, the generation of the equipment health assessment index through the frequency domain feature extraction algorithm includes: Perform wavelet transform processing on the operating state of the device to extract multi-scale frequency domain energy distribution characteristics; Construct a fault feature library based on historical fault data of the device, and calculate the similarity between the current frequency domain characteristics and the fault feature library through a matching algorithm; Generate a fault probability value according to the similarity and a preset degradation rate threshold; Combine the fault probability value with the performance degradation rate to form an equipment health assessment index.

[0011] Preferably, the multi-modal decision-making model includes a data fusion module and an instruction generation module. The data fusion module includes: Normalize the parameter deviation degree in the environmental state evolution model to obtain a first fusion vector; Perform discrete mapping on the fault probability value in the equipment health assessment index to generate a second fusion vector; Perform logical priority sorting on the instruction conflict detection results in the operation logic chain to generate a third fusion vector; Merge the first fusion vector, the second fusion vector, and the third fusion vector into a multi-modal decision input sequence through a feature splicing layer.

[0012] Preferably, the instruction generation module includes: Align the multi-modal decision input sequence in the spatio-temporal dimension to generate an instruction correlation tensor; Extract the temporal dependence features in the sequence through a gated recurrent unit to generate a candidate set of dynamic control instructions; Perform weighted fusion on the instruction correlation tensor and the candidate set of dynamic control instructions to generate a candidate instruction scoring matrix; Retain the instructions with scores higher than a preset value through a threshold screening algorithm and output a set of dynamic control instructions.

[0013] Preferably, constructing a control strategy topology diagram through an adaptive optimization algorithm includes: Initialize the node attributes according to the control action unit, and generate an edge weight matrix based on the execution condition constraints; Use the set of dynamic control instructions as the node input state, and the edge weight matrix consists of action priorities and execution condition constraints; Iteratively update the action value function of each node through a dynamic programming algorithm and adjust the edge weight matrix; Generate an optimal control strategy sequence covering all nodes according to the adjusted edge weight matrix.

[0014] Preferably, the method for constructing the environmental reference parameter library includes: Collect reference parameter samples under various typical environmental scenarios, and extract parameter correlation features and fluctuation range thresholds; Perform principal component analysis on the reference parameter samples to generate dimensionality-reduced feature vectors; Classify the dimensionality-reduced feature vectors according to the environmental categories and associate them with the parameter fluctuation prediction models; Store the classified feature vectors as an environmental reference parameter library and regularly update the models based on real-time collected data.

[0015] Preferably, the optimization method of the fault feature library includes: Calculate the initial fault mode granularity and the minimum fault sample density according to the distribution of the device historical fault data; Divide the fault mode categories through a clustering algorithm and calculate the feature similarity between different categories; Dynamically adjust the fault mode granularity and the minimum fault sample density according to the similarity to optimize the calculation accuracy of the fault probability value.

[0016] Preferably, the iterative update method of the dynamic programming algorithm includes: Define the state transition function between nodes as a weighted combination of the action priority and the execution condition constraints; Initialize the action value function of each node to zero and the starting state reward value to a preset initial value; Calculate the maximum cumulative reward value of each node based on the previous nodes through a recurrence equation and record the optimal action path; Reverse-derive a complete control strategy sequence according to the optimal action path.

[0017] Preferably, the present invention further includes an automated computer control system based on artificial intelligence, and the system includes: Multi-source data acquisition module: used to obtain a multi-source heterogeneous sensor data set, and the sensor data includes environmental parameters, device operation status and operation instruction logs; wherein, the environmental parameters include temperature, humidity and light intensity, and the device operation status includes current fluctuation characteristics and mechanical vibration frequency; Environmental modeling module: configured to construct an environmental state evolution model based on the environmental parameters through a dynamic space modeling algorithm, and the evolution model includes a parameter correlation matrix and an abnormal fluctuation prediction threshold; Health assessment module: used to generate device health assessment indicators through a frequency domain feature extraction algorithm according to the device operation status, and the assessment indicators include a fault probability value and a performance degradation rate; Instruction analysis module: perform temporal correlation analysis on the operation instruction logs to generate an operation logic chain and an instruction conflict detection result; Decision-making generation module: Input the environmental state evolution model, equipment health assessment metrics, and operation logic chain into the multi-modal decision-making model to generate a dynamic control instruction set; the multi-modal decision-making model includes a data fusion module and an instruction generation module; Policy output module: Based on the dynamic control instruction set, construct a control policy topology map through an adaptive optimization algorithm and output an equipment regulation plan; the nodes of the control policy topology map represent control action units, and the edges represent action priorities and execution condition constraints.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: The automated computer control method and system based on artificial intelligence proposed by the present invention have significant beneficial effects in many aspects. At the level of data processing and analysis, the invention breaks through the limitation of insufficient data utilization in traditional control methods by obtaining a multi-source heterogeneous sensor data set covering various key information such as environmental parameters, equipment operating status, and operation instruction logs. Using the dynamic space modeling algorithm to deeply mine environmental parameters, the constructed environmental state evolution model can accurately grasp the environmental change trend. For example, in an intelligent factory, it can predict in advance the possible impact on production equipment according to the changes in environmental parameters such as workshop temperature and humidity, and light intensity in real time, provide a suitable environmental guarantee for equipment operation, and avoid equipment failures or production quality degradation caused by environmental factors.

[0019] For the assessment of equipment health, the frequency-domain feature extraction algorithm plays an important role. It generates assessment metrics including failure probability values and performance degradation rates based on the equipment operating status, changing the traditional passive mode that relies on regular inspections and simple threshold alarms. Taking server equipment as an example, by continuously monitoring operation status data such as current fluctuation characteristics and mechanical vibration frequencies, the probability of equipment failure and the speed of performance degradation can be accurately calculated, enabling maintenance personnel to formulate maintenance plans in advance, repair or replace components before the equipment is about to fail, effectively reducing equipment downtime, improving the reliability and stability of the system, and reducing maintenance costs.

[0020] The temporal correlation analysis of operation instruction logs is also of great significance. The generated operation logic chain and instruction conflict detection results ensure the accurate execution of operation instructions. In complex computer system operations, system errors or abnormal behaviors caused by instruction conflicts are avoided. For example, in a large database management system with multiple users operating simultaneously, the operation logic can be clearly sorted out, potential instruction conflicts can be discovered and resolved in a timely manner, ensuring data integrity and the normal operation of the system.

[0021] The multi-modal decision-making model is one of the core advantages of the present invention. It organically integrates the environmental state evolution model, the equipment health assessment index, and the operation logic chain, and the generated dynamic control instruction set is more scientific and reasonable. This model can comprehensively consider various factors and make accurate decisions in different scenarios. Taking an intelligent data center as an example, when the environmental temperature rises and the health of some server devices deteriorates, the model will comprehensively analyze this information and generate targeted control instructions, such as adjusting the air-conditioning cooling power and optimizing the server load distribution, to achieve efficient energy-saving management of the data center and reliable operation of the equipment.

[0022] Finally, the control strategy topology diagram constructed based on the adaptive optimization algorithm and the output device regulation plan provide intuitive and effective guidance for device control. The control strategy topology diagram uses nodes to represent control action units and edges to represent action priorities and execution condition constraints, making the control process clearer and more orderly. In an industrial automation production line, according to the production tasks and equipment status, the optimal device regulation plan can be quickly generated to achieve the automation and intelligence of the production process, improving production efficiency and product quality. Brief Description of the Drawings

[0023] Figure 1 is the working principle diagram of the automated computer control method based on artificial intelligence according to the present invention; Figure 2 is the flowchart of constructing the environmental state evolution model by the dynamic space modeling algorithm; Figure 3 is the flowchart of the data fusion module of the multi-modal decision-making model; Figure 4 is the flowchart of constructing the control strategy topology diagram by the adaptive optimization algorithm. Detailed Embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] Please refer to Figures 1 - 4 , the present invention relates to an automated computer control method based on artificial intelligence, and its specific implementation process is as follows: Collect data using multiple sensors. These data cover environmental parameters, equipment operating status, and operation instruction logs. Among them, environmental parameters include temperature, humidity, light intensity, etc., equipment operating status data such as current fluctuation characteristics, mechanical vibration frequency, etc., and operation instruction logs record past operation instruction information. These multi-source heterogeneous data provide basic data support for subsequent analysis and decision-making.

[0026] Based on the obtained environmental parameters, use the dynamic space modeling algorithm to construct an environmental state evolution model. This model includes a parameter correlation matrix and an abnormal fluctuation prediction threshold. By analyzing the correlation relationship between environmental parameters and predicting possible abnormal fluctuations of environmental parameters, it is possible to better understand the change trend of the environmental state and provide an environmental basis for subsequent decision-making.

[0027] Based on the equipment operating status data, use the frequency domain feature extraction algorithm to generate equipment health assessment indicators. These indicators include the failure probability value and the performance degradation rate, which can intuitively reflect the current health status of the equipment, predict possible equipment failures in advance, and provide important references for equipment maintenance and management.

[0028] Perform time-series correlation analysis on the operation instruction logs to generate an operation logic chain and an instruction conflict detection result. The operation logic chain shows the logical sequence relationship between operation instructions, while the instruction conflict detection result can timely detect possible instruction conflict problems and ensure the correct execution of operation instructions.

[0029] Input the environmental state evolution model, equipment health assessment indicators, and operation logic chain into the multi-modal decision model to generate a dynamic control instruction set through this model. The multi-modal decision model comprehensively considers various factors, making the generated control instructions more scientific and reasonable.

[0030] Based on the dynamic control instruction set, use the adaptive optimization algorithm to construct a control strategy topology graph. In this topology graph, nodes represent control action units, and edges represent action priorities and execution condition constraints. Finally, output the equipment regulation plan according to the control strategy topology graph to achieve precise and effective control of the equipment.

[0031] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: In actual application scenarios, building an environmental state evolution model is a crucial step. During the building process, it is necessary to first process the obtained environmental parameters. Perform noise filtering on the environmental parameters, adopt a suitable filtering algorithm to remove the noise interference in the environmental parameters, and generate a standardized environmental data sequence to ensure the accuracy and reliability of the data. For example, when abnormal jumps occur in the collected temperature data, these abnormal values can be removed through the filtering algorithm, enabling the data to more truly reflect the changes in the environmental temperature.

[0032] Based on a preset environmental benchmark parameter library, generate parameter correlation weights through covariance matrix analysis. The environmental benchmark parameter library stores benchmark parameter samples and related features under various typical environmental scenarios. Suppose there are benchmark parameter samples of temperature, humidity, and light intensity in the environmental benchmark parameter library. Covariance matrix analysis can calculate the degree of correlation between these parameters, and then obtain the parameter correlation weights. Let there be environmental parameters, which are respectively , and the covariance matrix represents the covariance between the -th parameter and the -th parameter. The calculation formula is:

[0033] where is the number of samples, is the value of the -th parameter in the -th sample, is the sample mean of the -th parameter. According to the covariance matrix , the parameter correlation weight can be further calculated to measure the degree of tight correlation between parameters.

[0034] Adopt a time series prediction algorithm to calculate the change trend of environmental parameters in the future period, and generate a parameter deviation degree in combination with an abnormal fluctuation prediction threshold. The time series prediction algorithm can predict future parameter values based on historical environmental parameter data. Taking the ARIMA model as an example, it can model and predict time series data. Suppose the predicted temperature value at a certain future moment is , and the abnormal fluctuation prediction threshold is , and the current actual temperature value is . Then the parameter deviation degree can be calculated through the formula . The parameter deviation degree can intuitively reflect the degree of difference between the current environmental parameters and the predicted values.

[0035] Encode the parameter correlation weight, parameter deviation degree, and abnormal fluctuation prediction threshold into the environmental state evolution model. Through a specific encoding method, integrate this important information into the environmental state evolution model, enabling the model to comprehensively and accurately reflect the change law of the environmental state and providing reliable environmental information support for subsequent decision-making.

[0036] Example 2: Generating equipment health assessment indicators is of great significance for equipment maintenance and management. During the generation process, first perform wavelet transform processing on the equipment operating state. Wavelet transform is a time-frequency analysis method that can analyze signals at different scales and extract multi-scale frequency domain energy distribution characteristics. For example, for the current fluctuation signal of the equipment, through wavelet transform, it can be decomposed into components in different frequency bands, and each component contains the characteristic information of the equipment under different operating states. Let the equipment operating state signal be , and after wavelet transform, we get , where represents the scale parameter, which controls the stretching of the wavelet function; represents the translation parameter, which controls the position of the wavelet function on the time axis. Different and values correspond to frequency domain energy distribution characteristics at different scales and positions.

[0037] Construct a fault feature library based on the equipment historical fault data, and calculate the similarity between the current frequency domain feature and the fault feature library through a matching algorithm. The fault feature library stores the frequency domain feature information under various historical fault conditions. Assume there are fault modes in the fault feature library, and the corresponding frequency domain feature vectors for each fault mode are , , and the frequency domain feature vector of the current equipment operating state is . The cosine similarity algorithm can be used to calculate the similarity , and the formula is:

[0038] where represents the dot product operation of vectors, represents the norm of the vector. The similarity reflects the similarity degree between the current frequency domain feature and the feature of the rd fault mode.

[0039] Generate a fault probability value based on the similarity and the preset degradation rate threshold. Let the preset degradation rate threshold be . When the similarity exceeds a certain threshold, combined with the performance degradation rate of the equipment, the formula The calculated failure probability value represents the probability of the th type of failure occurring in the device. The failure probability value can help managers understand in advance the risk level of possible failures in the device.

[0040] Combining the failure probability value with the performance degradation rate forms an equipment health assessment index. By comprehensively considering the failure probability value and the performance degradation rate, the health status of the device can be evaluated more comprehensively. For example, when the failure probability value is high and the performance degradation rate is fast, it indicates that the health status of the device is poor and maintenance or replacement is required in a timely manner. The generated equipment health assessment index provides an important decision-making basis for the management and maintenance of the device.

[0041] Example 3: The multimodal decision-making model plays a core decision-making role in the entire automated computer control method, and the data fusion module is an important part of it. In the data fusion module, first, the parameter deviation degree in the environmental state evolution model is normalized. The parameter deviation degree reflects the change of environmental parameters, but the deviation degrees of different parameters may have different magnitudes and ranges. For the convenience of subsequent fusion and analysis, normalization processing is required. Assume the parameter deviation degree is , and it is normalized through the formula , where and are the minimum and maximum values of the parameter deviation degree over a period of time, respectively. The normalized parameter deviation degree is converted to the interval [0,1] to obtain the first fusion vector .

[0042] Perform a discretization mapping on the failure probability value in the equipment health assessment index to generate a second fusion vector. The failure probability value is a continuous numerical value. For better fusion with other information, discretization processing is required. For example, the failure probability value is divided into several levels, such as low, medium, and high. Let the failure probability value be . When , it is mapped to 0 (representing the low failure probability level); when , it is mapped to 1 (representing the medium failure probability level); when , it is mapped to 2 (representing the high failure probability level). Through this discretization mapping, the failure probability value is converted into discrete numerical values to form the second fusion vector .

[0043] Perform logical priority sorting on the instruction conflict detection results in the operation logic chain to generate a third fusion vector. The instruction conflict detection results may contain various types of conflict information, and it is necessary to perform logical priority sorting on this information. For example, determine the priority according to factors such as the importance of the instruction and the execution time. Assume that there are conflict items in the instruction conflict detection results. Sort them in descending order of priority and convert the sorted results into vector form to generate the third fusion vector .

[0044] Merge the first fusion vector, the second fusion vector, and the third fusion vector into a multi-modal decision input sequence through the feature concatenation layer. The feature concatenation layer concatenates the three fusion vectors in a certain order to form a complete multi-modal decision input sequence. For example, concatenate , , in sequence to obtain the multi-modal decision input sequence . This input sequence synthesizes information from multiple aspects such as the environment, device health, and operation instructions, providing comprehensive data support for the decision-making of the subsequent instruction generation module.

[0045] Example 4: The instruction generation module in the multi-modal decision model is a key part for generating a dynamic control instruction set. In the instruction generation module, first perform spatio-temporal dimension alignment on the multi-modal decision input sequence to generate an instruction correlation tensor. The multi-modal decision input sequence contains different types of information, and its spatio-temporal dimensions may not be consistent, so alignment processing is required. Assume that the multi-modal decision input sequence has a length of in the time dimension and a length of in the feature dimension. Through padding or cropping, etc., convert it into a tensor with unified spatio-temporal dimensions. For example, for a sequence with a shorter time dimension, zero-padding can be performed in the time dimension to make its time length consistent with that of other sequences, thereby generating the instruction correlation tensor , which is convenient for subsequent processing.

[0046] Extract the temporal dependence features in the sequence through a gated recurrent unit to generate a candidate set of dynamic control instructions. The gated recurrent unit (GRU) is a special recurrent neural network structure that can effectively capture the temporal dependence relationship in the time series. Input the instruction correlation tensor into the GRU. The GRU will learn the temporal dependence features in it according to the time order of the input sequence. Let the hidden state of the GRU be , the input be , and calculate the update gate through the formula ​, calculate the reset gate through the formula calculate the reset gate , calculate the candidate hidden state through the formula calculate the candidate hidden state , and finally update the hidden state through the formula update the hidden state , where is the Sigmoid function, represents element-wise multiplication, and are weight matrices. After being processed by the GRU, a set of candidate dynamic control instructions is generated, and the candidate set contains multiple possible control instructions.

[0047] Perform weighted fusion on the instruction correlation tensor and the set of candidate dynamic control instructions to generate a candidate instruction scoring matrix. To evaluate the quality of each candidate instruction, the instruction correlation tensor is weighted and fused with the set of candidate dynamic control instructions . Let the weighted fusion formula of the instruction correlation tensor and the candidate instruction be , where is the weight matrix, is the bias term, represents the score of the th candidate instruction in the th feature dimension. By performing weighted fusion on all candidate instructions, a candidate instruction scoring matrix is generated.

[0048] Retain the instructions with scores higher than the preset value through a threshold screening algorithm, and output the set of dynamic control instructions. Preset a scoring threshold , traverse the candidate instruction scoring matrix , and retain the instructions with scores higher than to form the set of dynamic control instructions . These instructions are the control instructions that are considered most suitable for the current system state after comprehensive evaluation, providing specific operation basis for subsequent device regulation.

[0049] Example 5: Constructing a control strategy topology diagram through an adaptive optimization algorithm and outputting a device regulation plan are key steps to achieve automated computer control. When constructing the control strategy topology diagram, first initialize the node attributes according to the control action units, and generate an edge weight matrix based on the execution condition constraints. The control action units are the basic elements that make up the control strategy topology diagram, and each control action unit corresponds to a node. Set initial attributes for each node, such as the type and initial state of the node. At the same time, determine the edge weight matrix according to the execution condition constraints. Assume that there are one, edge weight matrix The elements in represent the edge weight from node to node . When the execution condition of node is satisfied and can trigger the execution of node , is a non - zero value, and its magnitude can be determined according to factors such as action priority; otherwise .

[0050] Take the dynamic control instruction set as the input state of the node. The edge weight matrix is composed of action priority and execution condition constraints. The dynamic control instruction set provides input information for the construction of the control strategy topology. Each instruction in the dynamic control instruction set is corresponded to the corresponding control action unit as the input state of the node. The edge weight matrix comprehensively considers the action priority and execution condition constraints to ensure that the control actions can be executed in the correct order and conditions when constructing the control strategy.

[0051] Iteratively update the action value function of each node through the dynamic programming algorithm and adjust the edge weight matrix. The dynamic programming algorithm is a method for solving optimal decision - making problems. Define the state transition function between nodes as a weighted combination of action priority and execution condition constraints. Let the state transition function be , where is the action priority from node to node , is the execution condition constraint (1 when the condition is satisfied, 0 when not satisfied), and are weight coefficients. Initialize the action value function of each node to zero, and the starting - state reward value to the preset initial value . Calculate the maximum cumulative reward value of each node based on the previous nodes through the recurrence equation and record the optimal action path, where represents the action value function of node . During the calculation process, continuously adjust the edge weight matrix to optimize the control strategy.

[0052] Generate an optimal control strategy sequence covering all nodes according to the adjusted edge weight matrix. After obtaining the adjusted edge weight matrix through the dynamic programming algorithm, reverse - derive the complete control strategy sequence according to the optimal action path. Starting from the end - point node, gradually back - trace along the recorded optimal action path to the starting - point node to obtain an optimal control strategy sequence covering all nodes. This sequence is the device regulation scheme generated according to the current system state and the dynamic control instruction set, which can achieve efficient and precise control of the device.

[0053] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or apparatus.

[0054] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An automated computer control method based on artificial intelligence, characterized in that, Including: Obtain a multi-source heterogeneous sensor data set; the sensor data includes environmental parameters, device operating status, and operation instruction logs; Based on the environmental parameters, construct an environmental state evolution model through a dynamic space modeling algorithm, and the evolution model includes a parameter correlation matrix and an abnormal fluctuation prediction threshold; According to the device operating status, generate device health assessment indicators through a frequency domain feature extraction algorithm, and the assessment indicators include a failure probability value and a performance degradation rate; Perform time series correlation analysis on the operation instruction logs to generate an operation logic chain and an instruction conflict detection result; Input the environmental state evolution model, device health assessment indicators, and operation logic chain into a multi-modal decision model to generate a dynamic control instruction set; Based on the dynamic control instruction set, construct a control strategy topology map through an adaptive optimization algorithm and output a device regulation plan; the nodes of the control strategy topology map represent control action units, and the edges represent action priority and execution condition constraints.

2. The automated computer control method based on artificial intelligence according to claim 1, characterized in that, The construction of the environmental state evolution model through the dynamic space modeling algorithm includes: Perform noise filtering processing on the environmental parameters to generate a standardized environmental data sequence; Based on a preset environmental benchmark parameter library, generate parameter correlation weights through covariance matrix analysis; Use a time series prediction algorithm to calculate the change trend of environmental parameters in the future period, and generate a parameter deviation degree in combination with the abnormal fluctuation prediction threshold; Encode the parameter correlation weight, parameter deviation degree, and abnormal fluctuation prediction threshold into an environmental state evolution model.

3. An automated computer control method based on artificial intelligence according to claim 1, characterized in that, The generation of device health assessment indicators through the frequency domain feature extraction algorithm includes: Perform wavelet transform processing on the device operating status to extract multi-scale frequency domain energy distribution features; Construct a fault feature library based on historical device fault data, and calculate the similarity between the current frequency domain features and the fault feature library through a matching algorithm; Generate a failure probability value according to the similarity and a preset degradation rate threshold; Combine the failure probability value with the performance degradation rate to form a device health assessment indicator.

4. An automated computer control method based on artificial intelligence according to claim 1, characterized in that The multi-modal decision model includes a data fusion module and an instruction generation module, and the data fusion module includes: Normalize the parameter deviation degree in the environmental state evolution model to obtain a first fusion vector; Perform discrete mapping on the failure probability value in the device health assessment indicator to generate a second fusion vector; Perform logical priority sorting on the instruction conflict detection result in the operation logic chain to generate a third fusion vector; Combine the first fusion vector, the second fusion vector, and the third fusion vector into a multi-modal decision input sequence through a feature splicing layer.

5. An automated computer control method based on artificial intelligence according to claim 4, characterized in that, The instruction generation module includes: Perform spatio-temporal dimension alignment on the multi-modal decision input sequence to generate an instruction correlation tensor; Extract the temporal dependence features in the sequence through a gated recurrent unit to generate a dynamic control instruction candidate set; Perform weighted fusion on the instruction correlation tensor and the dynamic control instruction candidate set to generate a candidate instruction scoring matrix; Retain the instructions with scores higher than a preset value through a threshold screening algorithm and output a dynamic control instruction set.

6. An automated computer control method based on artificial intelligence according to claim 1, characterized in that, The construction of the control strategy topology map through the adaptive optimization algorithm includes: Initialize node attributes according to the control action unit, and generate an edge weight matrix based on the execution condition constraints; Use the dynamic control instruction set as the node input state, and the edge weight matrix consists of action priorities and execution condition constraints; Iteratively update the action value function of each node through the dynamic programming algorithm and adjust the edge weight matrix; Generate an optimal control strategy sequence covering all nodes according to the adjusted edge weight matrix.

7. An automated computer control method based on artificial intelligence according to claim 2, characterized in that, The construction method of the environmental benchmark parameter library includes: Collect benchmark parameter samples under various typical environmental scenarios, and extract parameter correlation features and fluctuation range thresholds; Perform principal component analysis on the benchmark parameter samples to generate a dimensionality-reduced feature vector; Classify the dimensionality-reduced feature vectors according to the environmental category and associate with a parameter fluctuation prediction model; Store the classified feature vectors as an environmental benchmark parameter library and update the model regularly based on real-time collected data.

8. An automated computer control method based on artificial intelligence according to claim 3, characterized in that, The optimization method of the fault feature library includes: Calculate the initial fault mode granularity and the minimum fault sample density according to the historical fault data distribution of the device; Divide the fault mode categories through a clustering algorithm and calculate the feature similarity between categories; Dynamically adjust the fault mode granularity and the minimum fault sample density according to the similarity to optimize the calculation accuracy of the fault probability value.

9. An automated computer control method based on artificial intelligence according to claim 6, characterized in that, The iterative update method of the dynamic programming algorithm includes: Define the state transition function between nodes as a weighted combination of action priorities and execution condition constraints; Initialize the action value function of each node to zero, and the starting state reward value to a preset initial value; Calculate the maximum cumulative reward value of each node based on the previous nodes through a recurrence equation and record the optimal action path; Reverse-derive a complete control strategy sequence according to the optimal action path.

10. An automated computer control system based on artificial intelligence, characterized in that, Include: Multi-source data acquisition module: used to obtain a multi-source heterogeneous sensor data set, and the sensor data includes environmental parameters, device operating status, and operation instruction logs; among them, the environmental parameters include temperature, humidity, and light intensity, and the device operating status includes current fluctuation characteristics and mechanical vibration frequency; Environmental modeling module: configured to construct an environmental state evolution model through a dynamic space modeling algorithm based on the environmental parameters, and the evolution model includes a parameter correlation matrix and an abnormal fluctuation prediction threshold; Health assessment module: used to generate a device health assessment index through a frequency domain feature extraction algorithm according to the device operating status, and the assessment index includes a fault probability value and a performance degradation rate; Instruction analysis module: perform temporal correlation analysis on the operation instruction logs to generate an operation logic chain and an instruction conflict detection result; Decision generation module: input the environmental state evolution model, device health assessment index, and operation logic chain into a multi-modal decision model to generate a dynamic control instruction set; the multi-modal decision model includes a data fusion module and an instruction generation module; Strategy output module: used to construct a control strategy topology map through an adaptive optimization algorithm based on the dynamic control instruction set and output a device regulation plan; the nodes of the control strategy topology map represent control action units, and the edges represent action priorities and execution condition constraints.

Images