Remote operation and maintenance management method for laser equipment based on cloud collaboration

Abstract

The invention discloses a laser equipment remote operation and maintenance management method based on cloud collaboration, and the method comprises the following steps: S1, constructing an operation data set of remote operation and maintenance management, and uploading the operation data set to a cloud platform; s2, inputting the pre-processed equipment state characteristics into a particle generation network, and constructing an initial remote operation and maintenance strategy set; s3, performing local disturbance optimization on the initial remote operation and maintenance strategy set by applying a fireworks algorithm, and constructing a candidate strategy set; s4, constructing a final remote operation and maintenance strategy through the initial remote operation and maintenance strategy and the remote operation and maintenance strategy enhanced by the fireworks algorithm; s5, issuing the final remote operation and maintenance strategy to an edge node through the cloud platform, and driving fault response, maintenance scheduling and maintenance task execution; and S6, updating the particle generation network and fireworks algorithm parameters. The fireworks algorithm, the particle generation network model and the industrial equipment operation and maintenance data feedback technology are combined, and laser equipment remote operation and maintenance management based on cloud collaboration is achieved.

Description

Technical Field

[0001] This invention relates to the field of remote operation and maintenance optimization technology, and in particular to a remote operation and maintenance management method for laser equipment based on cloud collaboration. Background Technology

[0002] With the continuous improvement of the intelligence and informatization level of industrial manufacturing equipment, especially the widespread application of high-precision laser industrial equipment such as lasers, laser cutting machines, and laser marking equipment in precision machining, microstructure manufacturing, and automated operations in complex scenarios, equipment operation and maintenance management is gradually shifting from traditional manual inspection and periodic maintenance to intelligent diagnosis, predictive maintenance, and remote operation and maintenance. Currently, to reduce operation and maintenance costs, improve equipment availability, and ensure production continuity, more and more manufacturing enterprises, equipment manufacturers, and service providers are beginning to build cloud-based remote operation and maintenance systems. By collecting equipment operating parameters and remotely analyzing equipment health status, these systems enable early detection of potential faults and the delivery of handling suggestions.

[0003] In existing technologies, common remote operation and maintenance solutions mainly rely on SCADA, MES, or industrial gateway platforms to achieve data uploading, remote monitoring, and event response. These systems often rely on static rules or experience models to determine equipment status, resulting in significant problems such as weak ability to identify complex fault modes, poor model generalization ability, and a lack of elaboration in response mechanisms. Furthermore, while some current equipment management platforms integrate edge computing nodes and simple model deployment capabilities, their policy generation typically depends on preset templates or offline training results, making it difficult to generate real-time adaptive remote operation and maintenance policies in the face of continuously changing equipment operating states, fluctuating environmental conditions, and dynamic adjustments to resource scheduling pressures.

[0004] On the other hand, existing research attempts to introduce optimization techniques such as evolutionary algorithms and reinforcement learning into the field of intelligent policy generation to optimize the planning and scheduling of remote operation and maintenance processes. However, these studies generally suffer from the following shortcomings: First, the optimization algorithms are highly dependent on the initial individual or sample structure, resulting in poor policy diversity, limited solution space exploration capabilities, and a tendency to get trapped in local optima. Second, the model's response mechanism to feedback data is relatively weak, lacking the ability to adaptively update based on feedback, making it difficult to form a complete closed loop of operation and maintenance policy evolution. Third, the cloud-edge collaboration mechanism is poorly designed, with edge nodes typically serving only as execution terminals or data acquisition nodes, failing to participate in local policy decision-making and fine-tuning optimization, resulting in significant response delays and insufficient system flexibility and robustness.

[0005] In remote policy generation methods based on evolutionary computation, individual initialization typically employs random or small-sample expansion, leading to unstable initial solution quality and a rapid decline in population diversity during early optimization phases, thus weakening global search capabilities. Furthermore, traditional fireworks algorithms, due to their reliance on fitness ranking for explosion center selection, exhibit a pattern of "superior individuals clustering and perturbing, while inferior individuals do not participate in evolution," resulting in population convergence, premature convergence, and insufficient exploration of novel regions. This is particularly problematic in remote operation and maintenance scenarios, where strategies require strong adaptability and diversity to match heterogeneous equipment conditions; such algorithmic structures struggle to meet the demands of high-frequency decision-making.

[0006] Furthermore, traditional strategy optimization systems often treat model training, strategy generation, strategy execution, and feedback analysis as separate modules, lacking a fusion mechanism for unified model control and the ability to dynamically adjust fusion parameters under feedback. This results in large fluctuations in strategy execution performance, ineffective feedback data in informing the model, and a lack of iterative update mechanisms for remote operation and maintenance strategies. Currently, in most systems, particle generation networks or generative models are primarily used in image processing and natural language processing, lacking dedicated structural designs for industrial intelligent operation and maintenance, and failing to structurally integrate with swarm intelligence optimization algorithms. This hinders the engineering implementation of generative model-based strategy optimization in operation and maintenance scenarios.

[0007] Therefore, how to provide a cloud-based collaborative remote operation and maintenance management method for laser equipment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0008] One objective of this invention is to propose a cloud-based collaborative remote operation and maintenance management method for laser equipment. This invention combines advanced technologies such as the Fireworks algorithm, the Particle Generator Network model, and industrial equipment operation and maintenance data feedback. By constructing a policy generation mechanism based on solution space learning and a local perturbation optimization algorithm, it achieves remote intelligent operation and maintenance of laser equipment under complex operating conditions. The system completes policy modeling and optimization on a cloud platform and, combined with edge node execution and feedback, constructs a closed-loop policy response system. The Particle Generator Network is responsible for generating structured, high-quality initial policies, the Fireworks algorithm performs local accuracy compensation optimization on the policies, and the feedback module continuously collects task execution results and drives model parameter updates, ensuring that the policy system has continuous learning and adaptive capabilities. This demonstrates significant effectiveness in improving equipment availability, reducing maintenance costs, and shortening fault response time.

[0009] A remote operation and maintenance management method for laser equipment based on cloud collaboration according to an embodiment of the present invention includes the following steps: S1. Collect the operating parameters of the laser equipment, construct a remote operation and maintenance management dataset, preprocess the dataset, and upload it to the cloud platform. S2. Input the preprocessed equipment status features into the particle generation network, learn the solution space distribution based on historical high-quality operation and maintenance strategies, and construct an initial set of remote operation and maintenance strategies to overcome the limitations of the fireworks algorithm, which relies on initial individuals and is prone to population convergence. S3. Generate multiple sparks based on the initial set of remote operation and maintenance strategies, each spark representing a strategy. Apply the fireworks algorithm for local perturbation optimization to construct a candidate strategy set. S4. On the cloud platform, use the fireworks algorithm to compensate for the insufficient local accuracy in the initial remote operation and maintenance strategies generated by the particle generation network. Construct the final remote operation and maintenance strategy by combining the initial remote operation and maintenance strategy with the remote operation and maintenance strategy enhanced by the fireworks algorithm. S5. Deploy the final remote operation and maintenance strategy to edge nodes through the cloud platform to drive fault response, maintenance scheduling, and maintenance task execution. S6. Collect operation and maintenance feedback results and synchronize them to the cloud platform to update the parameters of the particle generation network and the fireworks algorithm, thereby achieving continuous optimization of the remote operation and maintenance strategy.

[0010] Optionally, the operating parameters specifically include laser power, current, voltage, cooling temperature, coolant flow rate, ambient temperature, and vibration signal intensity.

[0011] Optionally, S1 specifically includes: The operating parameters are sorted and indexed by device number, sampling time, and parameter fields to build a remote operation and maintenance management operating dataset. The operating dataset is preprocessed and periodically packaged and uploaded to the cloud.

[0012] Optionally, S2 specifically includes: S21. Receive the running dataset uploaded to the cloud platform, perform statistical analysis and feature extraction on the running dataset, extract key behavioral parameters, state change trends, fluctuation range and interrelationships during equipment operation, and construct a unified dimension equipment state feature vector. S22. The device state feature vector is nonlinearly reduced in dimensionality using the encoding module of the particle generation network. Historical high-quality operation and maintenance strategies stored on the cloud platform are used to train the particle generation network generation module, learn the solution space distribution, and construct an initial set of remote operation and maintenance strategies. ; in, This is the initial set of remote operation and maintenance strategies. For the first A generated remote operation and maintenance policy, For strategy The One portion, The first of the historical mean strategy One portion, For strategy The first corresponding potential space The squares of the variables, where n is the dimension of the policy vector and d is the dimension of the latent space. As a regularization factor, To set the constraint threshold for the boundary of the control strategy set, To construct regularization control terms in the policy generation process, For the generated first A remote operation and maintenance strategy This is the feature vector index in the remote operation and maintenance strategy vector. The nth hidden space vector in the particle generation network One dimension, For the first The strategy in the first The squared deviation from the historical strategy mean in each dimension.

[0013] Optionally, S3 specifically includes: S31. Take each strategy in the initial remote operation and maintenance strategy set as the initial individual of the fireworks algorithm, and set the perturbation range and the upper limit of the number of optimization iterations for each initial individual. ; S32. In each iteration, local perturbation optimization is performed within the perturbation range for each initial policy, generating a spark solution set and constructing a candidate policy set. The optimization objective satisfied by the candidate policy set is: ; in, For the first One candidate strategy, For perturbation solutions, For the first A generated remote operation and maintenance policy, To use strategy The radius of the local perturbation search range centered on the target. For Centered on, with radius The solution space region, For candidate strategies One portion, The first of the historical mean strategy Quantity, For the first The weighted coefficients for each strategy dimension As a diversity regulator, This is the index for remote operations and maintenance dimensions, where n is the dimension of the remote operations and maintenance policy vector. Dimension index of the policy vector This is the feature vector index in the remote operation and maintenance strategy vector. The first candidate strategy Each component The first in the historical mean strategy Each component The square of the difference between them To take the difference between the pair of remote operation and maintenance dimension vectors with the largest difference; S33. Evaluate the fitness of all candidate strategies, select the individual with the best fitness in each round and add it to the update set, and iterate until the maximum number of rounds T is reached.

[0014] Optionally, S4 specifically includes: To address the shortcomings in local accuracy of the initial remote operation and maintenance (O&M) strategy set generated by particle generation networks, a fusion mechanism based on two-layer compensation is adopted for strategy construction. Each strategy in the initial strategy set is indexed and paired with its corresponding candidate strategy optimized by the fireworks algorithm. A fusion strategy is constructed through local correction and global direction offset. The fusion objective not only minimizes the deviation between the final O&M strategy and the candidate and initial remote O&M strategies, but also controls and minimizes the magnitude of the difference between the final O&M strategy and the candidate and initial remote O&M strategies. ; in, This is the final set of remote operation and maintenance strategies. For the first The final remote operation and maintenance strategy generated by the fusion The candidate solution after fusion is the first one in the initial strategy. Components of each dimension For the first One candidate strategy, For the current integration of the first Components of each dimension For the first The fusion weight coefficients of each dimension For the first Compensation adjustment factors in each dimension This is the feature vector index in the remote operation and maintenance strategy vector. For the first in the set of fusion strategies Index of each strategy This is the index for remote operations and maintenance dimensions, where n is the dimension of the remote operations and maintenance policy vector. Dimension index of the policy vector This is the global regularization coefficient. To obtain the difference between the pair of remote operation and maintenance dimension vectors with the largest difference, for The square of the deviation from the fusion target value.

[0015] Optionally, the edge node specifically includes an edge communication module, a local execution control unit, an edge perception and feedback acquisition module, and an edge computing processing unit.

[0016] Optionally, S5 specifically includes: S51. Based on the optimization features of each strategy in the final fusion strategy set, construct a structured scheduling instruction set for edge node execution on the cloud platform. S52. The scheduling instruction set is sent to the corresponding edge nodes through the remote communication module, and the activation status of each instruction is determined based on a multi-factor fusion scoring mechanism. After the instruction is triggered, the edge node drives the execution of remote operation and maintenance actions, including fault response, maintenance scheduling, and maintenance tasks. ; in, For the first The comprehensive scheduling score of each remote scheduling instruction. For candidate policies in the set Index in This is the set of indexes for currently active candidate policies. This is the feature vector index in the remote operation and maintenance strategy vector. For the first The first fusion strategy in the Components in each dimension For the first The importance weight of each integration strategy dimension For resource scheduling factor weight coefficients, For indexing resource items, For the current task, the first The utilization rate of each resource unit. For the first The fusion strategy corresponds to the first edge node. Availability score for each resource unit As a time conflict penalty factor, For remote operation and maintenance dimension indexing, For the first The planned execution time point for a remote operation and maintenance strategy The currently activated number The execution time of remote operation and maintenance policies executed by each edge node is the index number of the remote operation and maintenance scheduling instruction currently being evaluated, n is the dimension of the remote operation and maintenance policy vector, and m is the number of remote operation and maintenance policies that have been scheduled.

[0017] Optionally, the operation and maintenance feedback results specifically include task execution success rate, response latency, task execution time and resource consumption; Optionally, S6 specifically includes: S61. During the execution of remote operation and maintenance tasks, continuously collect operation and maintenance feedback data from edge nodes and periodically synchronize it to the cloud platform; S62. In the cloud platform, update the particle generation network. Based on the changes in response latency and resource consumption levels in the feedback data, correct the spatial sampling distribution of the particle generation network. Utilize the success rate and fitness changes of the fusion strategy in actual execution to adjust the generation module parameters of the particle generation network, including the fusion strategy vector output structure and feature mapping method. S63. On the cloud platform, the key parameters of the fireworks algorithm are updated. Based on the timeliness and local convergence characteristics of the fusion strategy set in the execution of edge nodes, the perturbation range, spark generation quantity and search granularity in the fireworks algorithm are updated.

[0018] The beneficial effects of this invention are: The present invention proposes a cloud-based remote operation and maintenance management method for laser equipment that combines particle generation network optimization of fireworks algorithm. It constructs an intelligent operation and maintenance system based on cloud-edge collaboration mechanism, focusing on remote status monitoring, strategy generation, dynamic optimization and closed-loop execution of laser equipment throughout its entire life cycle. It achieves significant breakthroughs and improvements on the basis of existing technologies and has outstanding engineering application value and promotion prospects.

[0019] By introducing a particle generation network to model the potential mapping relationship between the device's operating state and the policy space, the system can fully learn and express the distribution characteristics of historical high-quality operation and maintenance policies in the solution space, effectively overcoming the limitations of traditional optimization methods that rely on random sampling in the policy initialization stage and are difficult to cover complex scenarios. Through the structured expression of latent vectors, the particle generation network has the ability to generate a diverse and comprehensive set of remote operation and maintenance policies, laying a high-quality starting point for subsequent policy perturbation optimization and significantly improving the structural stability and initial effectiveness of the policy generation stage.

[0020] The introduction of the fireworks algorithm effectively compensates for the shortcomings of particle generation networks in terms of local accuracy. By optimizing the local perturbation of the generated strategy set, the system can quickly identify and correct the failure dimensions of the initial strategy under edge conditions and abnormal states, improving the matching and robustness of the strategy in actual execution. Simultaneously, this invention introduces a fusion mechanism based on the historical fitness mean and the distribution characteristics of the perturbation neighborhood during the optimization process of the fireworks algorithm. This not only maintains population diversity but also improves the local convergence speed of the search process, thereby constructing a high-performance operation and maintenance strategy candidate set with both breadth and depth.

[0021] More importantly, this invention achieves structural synergy between the particle generation network and the fireworks algorithm by establishing a bidirectional compensation fusion mechanism. The system matches and fuses the results of the two strategies in the cloud, preserving not only the global distribution advantages of the particle generation strategy but also leveraging the local enhancement direction of the fireworks optimization strategy. This allows for optimal reconstruction of the fusion strategy through multi-dimensional parameter adjustment. This fusion strategy demonstrates higher task adaptability, execution accuracy, and fault tolerance in subsequent remote scheduling and edge deployment, significantly improving the system's response efficiency and risk control level.

[0022] Furthermore, this invention designs a complete edge feedback acquisition and cloud model linkage update mechanism, which can adaptively optimize the strategy model based on the execution results of operation and maintenance tasks. The system collects key feedback indicators in real time, including response latency, execution success rate, and resource consumption. Through dynamic adjustment of model parameters and reconstruction of fusion factors, the particle generation network and the fireworks algorithm form a feedback-driven joint learning system, truly realizing a closed-loop self-evolution of strategy generation, optimization, execution, and update, overcoming core problems such as model rigidity, feedback disconnection, and strategy failure in existing systems.

[0023] In summary, this invention constructs a collaborative optimization method for remote operation and maintenance strategies based on particle generation networks and fireworks algorithms. This method creates an integrated intelligent remote operation and maintenance system with learning, optimization, feedback adjustment, and cloud-edge collaborative execution capabilities. It significantly improves the intelligent management level of laser equipment under complex operating conditions, reduces maintenance response time and manpower dependence, and enhances the stability and safety of equipment operation. It has significant industrial practical value and promotion potential. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0025] Figure 1 This is a flowchart of a remote operation and maintenance management method for laser equipment based on cloud collaboration proposed in this invention; Figure 2 This is a schematic diagram of a remote operation and maintenance management method for laser equipment based on cloud collaboration proposed in this invention; Figure 3 This is a data flow diagram of a cloud-based collaborative remote operation and maintenance management method for laser equipment proposed in this invention. Detailed Implementation

[0026] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0027] refer to Figure 1-3 A remote operation and maintenance management method for laser equipment based on cloud collaboration includes the following steps: S1. Collect the operating parameters of the laser equipment, construct a remote operation and maintenance management dataset, preprocess the dataset, and upload it to the cloud platform. S2. Input the preprocessed equipment status features into the particle generation network, learn the solution space distribution based on historical high-quality operation and maintenance strategies, and construct an initial remote operation and maintenance strategy set. S3. Generate multiple sparks based on the initial remote operation and maintenance strategy set, each spark representing a strategy. Apply the fireworks algorithm for local perturbation optimization to construct a candidate strategy set. S4. On the cloud platform, use the fireworks algorithm to compensate for the insufficient local accuracy in the initial remote operation and maintenance strategy generated by the particle generation network. Construct the final remote operation and maintenance strategy by combining the initial remote operation and maintenance strategy with the candidate strategies enhanced by the fireworks algorithm. S5. Deploy the final remote operation and maintenance strategy to edge nodes through the cloud platform to drive fault response, maintenance scheduling, and maintenance task execution. S6. Collect operation and maintenance feedback results and synchronize them to the cloud platform, updating the particle generation network and fireworks algorithm parameters.

[0028] This invention achieves real-time monitoring and optimization of equipment status by collecting operating parameters of laser equipment and constructing a remote operation and maintenance management dataset. A particle generation network is used to learn equipment status characteristics, and an initial set of remote operation and maintenance strategies is constructed based on historical high-quality operation and maintenance strategies. Subsequently, the fireworks algorithm is used to optimize the initial strategies with local perturbations, generating a set of candidate strategies to further improve the adaptability and execution accuracy of the strategies. By combining the candidate strategies enhanced by the fireworks algorithm with the initial strategies, a final remote operation and maintenance strategy is constructed and distributed to edge nodes through a cloud platform to ensure efficient execution of fault response and maintenance tasks. Simultaneously, the system continuously collects operation and maintenance feedback data and synchronizes it to the cloud platform, dynamically updating the parameters of the particle generation network and the fireworks algorithm to further optimize strategy adjustments and ensure efficient and accurate decision-making capabilities under different operation and maintenance scenarios. This method improves the intelligence level and response speed of remote operation and maintenance management, effectively reduces the failure rate, and enhances the stability and operating efficiency of the equipment.

[0029] In this embodiment, the operating parameters specifically include laser power, current, voltage, cooling temperature, coolant flow rate, ambient temperature, and vibration signal intensity.

[0030] This implementation method comprehensively monitors the operating status of laser equipment by real-time acquisition of operating parameters, including laser power, current, voltage, cooling temperature, coolant flow rate, ambient temperature, and vibration signal intensity. Through multi-dimensional operational data analysis, it can accurately grasp the equipment's working condition, promptly identify potential fault risks, optimize equipment management and maintenance strategies, and improve the stability and efficiency of equipment operation. This method can dynamically adjust control strategies under different operating conditions to ensure that the equipment always operates in optimal condition.

[0031] In this embodiment, S1 specifically includes: The operating parameters are sorted and indexed by device number, sampling time, and parameter fields to build a remote operation and maintenance management operating dataset. The operating dataset is preprocessed and periodically packaged and uploaded to the cloud.

[0032] This invention constructs a highly efficient remote operation and maintenance management dataset by sorting and indexing operating parameters according to device number, sampling time, and parameter fields. After preprocessing, this dataset is periodically uploaded to a cloud platform to ensure the accuracy and completeness of real-time data. Through structured data management and cloud storage, the system can quickly respond to changes in device operating status, optimize fault warnings and operation and maintenance decisions, and improve the real-time performance, accuracy, and operability of remote operation and maintenance management.

[0033] In this embodiment, S2 specifically includes: S21. Receive the running dataset uploaded to the cloud platform, perform statistical analysis and feature extraction on the running dataset, extract key behavioral parameters, state change trends, fluctuation range and interrelationships during equipment operation, and construct a unified dimension equipment state feature vector. S22. The device state feature vector is nonlinearly reduced in dimensionality using the encoding module of the particle generation network. Historical high-quality operation and maintenance strategies stored on the cloud platform are used to train the particle generation network generation module, learn the solution space distribution, and construct an initial set of remote operation and maintenance strategies. ; in, This is the initial set of remote operation and maintenance strategies. For the first A generated remote operation and maintenance policy, For strategy The One portion, The first of the historical mean strategy One portion, For strategy The first corresponding potential space The squares of the variables, where n is the dimension of the policy vector and d is the dimension of the latent space. As a regularization factor, To set the constraint threshold for the boundary of the control strategy set, To construct regularization control terms in the policy generation process, For the generated first A remote operation and maintenance strategy This is the feature vector index in the remote operation and maintenance strategy vector. The nth hidden space vector in the particle generation network One dimension, For the first The strategy in the first The squared deviation from the historical strategy mean in each dimension.

[0034] This invention receives operational datasets uploaded to the cloud, performs statistical analysis and feature extraction on the data, extracts key behavioral parameters, state change trends, and their interrelationships during equipment operation, and constructs a unified-dimensional equipment state feature vector. These feature vectors undergo nonlinear dimensionality reduction mapping through the encoding module of a particle generation network and are trained using historical high-quality operation and maintenance strategies to learn the solution space distribution and generate an initial set of remote operation and maintenance strategies. By introducing regularization control terms and boundary constraint thresholds, the generated strategies are ensured to have good adaptability and optimization capabilities. This method effectively improves the accuracy of strategy generation, ensuring that remote operation and maintenance strategies maintain high robustness and flexibility in complex operating environments, thereby enhancing the accuracy of equipment management and fault early warning.

[0035] In this embodiment, S3 specifically includes: S31. Take each strategy in the initial remote operation and maintenance strategy set as the initial individual of the fireworks algorithm, and set the perturbation range and the upper limit of the number of optimization iterations for each initial individual. ; S32. In each iteration, local perturbation optimization is performed within the perturbation range for each initial policy, generating a spark solution set and constructing a candidate policy set. The optimization objective satisfied by the candidate policy set is: ; in, For the first One candidate strategy, For perturbation solutions, For the first A generated remote operation and maintenance policy, To use strategy The radius of the local perturbation search range centered on the target. For Centered on, with radius The solution space region, For candidate strategies One portion, The first of the historical mean strategy Quantity, For the first The weighted coefficients for each strategy dimension As a diversity regulator, This is the index for remote operations and maintenance dimensions, where n is the dimension of the remote operations and maintenance policy vector. Dimension index of the policy vector This is the feature vector index in the remote operation and maintenance strategy vector. The first candidate strategy Each component The first in the historical mean strategy Each component The square of the difference between them To take the difference between the pair of remote operation and maintenance dimension vectors with the largest difference; S33. Evaluate the fitness of all candidate strategies, select the individual with the best fitness in each round and add it to the update set, and iterate until the maximum number of rounds T is reached.

[0036] This invention uses each strategy in the initial remote operation and maintenance strategy set as the initial individual for the "spark algorithm," sets the perturbation range and the upper limit of the optimization iteration count, and generates a candidate strategy set using a local perturbation optimization method. In each iteration, a local perturbation is performed around each initial strategy, and the generated spark solution set is adjusted according to the optimization objective to ensure the quality of the candidate strategy vectors. By evaluating the fitness of the candidate strategies, the optimal strategy is selected and updated to the strategy set, and optimization continues until the maximum number of iterations is reached. This method can effectively improve the adaptability and accuracy of remote operation and maintenance strategies, optimize equipment management, fault warning, and operation and maintenance decisions, and ensure efficient and flexible operation and maintenance management in dynamic environments.

[0037] In this embodiment, S4 specifically includes: To address the shortcomings in local accuracy of the initial remote operation and maintenance (O&M) strategy set generated by particle generation networks, a fusion mechanism based on two-layer compensation is adopted for strategy construction. Each strategy in the initial strategy set is indexed and paired with its corresponding candidate strategy optimized by the fireworks algorithm. A fusion strategy is constructed through local correction and global direction offset. The fusion objective not only minimizes the deviation between the final O&M strategy and the candidate and initial remote O&M strategies, but also controls and minimizes the magnitude of the difference between the final O&M strategy and the candidate and initial remote O&M strategies. ; in, This is the final set of remote operation and maintenance strategies. For the first The final remote operation and maintenance strategy generated by the fusion The candidate solution after fusion is the first one in the initial strategy. Components of each dimension For the first One candidate strategy, For the current integration of the first Components of each dimension For the first The fusion weight coefficients of each dimension For the first Compensation adjustment factors in each dimension This is the feature vector index in the remote operation and maintenance strategy vector. For the first in the set of fusion strategies Index of each strategy This is the index for remote operations and maintenance dimensions, where n is the dimension of the remote operations and maintenance policy vector. Dimension index of the policy vector This is the global regularization coefficient. To obtain the difference between the pair of remote operation and maintenance dimension vectors with the largest difference, for The square of the deviation from the fusion target value.

[0038] This invention addresses the shortcomings in local accuracy of the initial remote operation and maintenance (O&M) strategy set generated by particle generation networks (PGNs) by employing a fusion mechanism based on two-layer compensation. The method constructs the final fused strategy by indexing and pairing the initial strategy with candidate strategies optimized by the fireworks algorithm, combining local correction and global direction offset. The fusion objective not only minimizes the deviation between the final O&M strategy and the initial and candidate strategies but also optimizes the consistency and adaptability of the strategy by controlling the magnitude of the difference. By adjusting the fusion weight coefficients and compensation adjustment factors of each dimension, this method ensures high-precision strategy construction, improves the execution capability and flexibility of remote O&M strategies in complex environments, and effectively enhances the intelligence and dynamic adjustment capabilities of equipment management.

[0039] In this embodiment, the edge node specifically includes an edge communication module, a local execution control unit, an edge perception and feedback acquisition module, and an edge computing processing unit.

[0040] This implementation method, by setting up edge nodes and integrating an edge communication module, a local execution control unit, an edge sensing and feedback acquisition module, and an edge computing processing unit, effectively improves the response speed and processing capacity of the remote operation and maintenance system. The edge communication module ensures high-speed data transmission between the device and the cloud platform, while the local execution control unit can execute operation and maintenance strategies in real time on-site, reducing reliance on the central platform. The edge sensing and feedback acquisition module collects device status data in real time, providing accurate feedback information for decision-making. The edge computing processing unit performs data processing and analysis locally, improving the system's processing efficiency and reliability. Through this edge intelligent architecture, the system can achieve more efficient and accurate operation and maintenance management, effectively reducing latency and enhancing the real-time performance and intelligence of device management.

[0041] In this embodiment, S5 specifically includes: S51. Based on the optimization features of each strategy in the final fusion strategy set, construct a structured scheduling instruction set for edge node execution on the cloud platform. S52. The scheduling instruction set is sent to the corresponding edge nodes through the remote communication module, and the activation status of each instruction is determined based on a multi-factor fusion scoring mechanism. After the instruction is triggered, the edge node drives the execution of remote operation and maintenance actions, including fault response, maintenance scheduling, and maintenance tasks. ; in, For the first The comprehensive scheduling score of each remote scheduling instruction. For candidate policies in the set Index in This is the set of indexes for currently active candidate policies. This is the feature vector index in the remote operation and maintenance strategy vector. For the first The first fusion strategy in the Components in each dimension For the first The importance weight of each integration strategy dimension For resource scheduling factor weight coefficients, For indexing resource items, For the current task, the first The utilization rate of each resource unit. For the first The fusion strategy corresponds to the first edge node. Availability score for each resource unit As a time conflict penalty factor, For remote operation and maintenance dimension indexing, For the first The planned execution time point for a remote operation and maintenance strategy The currently activated number The execution time of remote operation and maintenance policies executed by each edge node is the index number of the remote operation and maintenance scheduling instruction currently being evaluated, n is the dimension of the remote operation and maintenance policy vector, and m is the number of remote operation and maintenance policies that have been scheduled.

[0042] This invention constructs a structured scheduling instruction set on a cloud platform based on optimized features from a final fusion strategy set, ensuring that each instruction accurately executes remote operation and maintenance tasks. The scheduling instruction set is distributed to edge nodes via a remote communication module, and a multi-factor fusion scoring mechanism is used to evaluate the activation status of each instruction. The system intelligently decides the execution order of instructions based on factors such as resource availability, task occupancy ratio, and time conflicts. This method not only improves resource utilization but also effectively reduces execution latency, achieving efficient execution of fault response, maintenance scheduling, and maintenance tasks. By dynamically adjusting the scheduling strategy, the system can flexibly adapt to different operation and maintenance needs, enhancing the intelligence and responsiveness of remote operation and maintenance management.

[0043] In this embodiment, the operation and maintenance feedback results specifically include task execution success rate, response latency, task execution time and resource consumption; In this embodiment, S6 specifically includes: S61. During the execution of remote operation and maintenance tasks, continuously collect operation and maintenance feedback data from edge nodes and periodically synchronize it to the cloud platform; S62. In the cloud platform, update the particle generation network. Based on the changes in response latency and resource consumption levels in the feedback data, correct the spatial sampling distribution of the particle generation network. Utilize the success rate and fitness changes of the fusion strategy in actual execution to adjust the generation module parameters of the particle generation network, including the fusion strategy vector output structure and feature mapping method. S63. On the cloud platform, the key parameters of the fireworks algorithm are updated. Based on the timeliness and local convergence characteristics of the fusion strategy set in the execution of edge nodes, the perturbation range, spark generation quantity and search granularity in the fireworks algorithm are updated.

[0044] This invention continuously collects operational feedback data from edge nodes and synchronizes it to a cloud platform, ensuring real-time monitoring and optimization of remote operation and maintenance tasks. Based on changes in response latency and resource consumption levels in the feedback data, the system dynamically updates the particle generation network, adjusts the spatial sampling distribution and generation module parameters, and improves the adaptability and accuracy of the strategy. Simultaneously, by combining the success rate and fitness changes of the fusion strategy in actual execution, the system optimizes the strategy output structure and feature mapping method. By updating the key parameters of the fireworks algorithm, optimizing the perturbation range, the number of sparks generated, and the search granularity, it ensures rapid response and efficient execution under different operating conditions. This method, through adaptive adjustment, significantly improves the intelligence level and response efficiency of remote operation and maintenance management.

[0045] Example 1: To verify the feasibility of this invention in practice, it was applied to a smart manufacturing enterprise. This enterprise mainly engages in the research and development and mass production of laser precision processing equipment, with core equipment including 78 units of various models of laser cutting machines, laser welding equipment, and laser cleaning machines. Previously, the enterprise's equipment operation and maintenance mainly relied on a combination of manual inspection and regular maintenance. The average response time for each fault exceeded 4 hours, especially at night or during holidays, when response efficiency and diagnostic accuracy dropped significantly. Minor faults often failed to be detected in time and developed into downtime accidents, affecting the stability of production line delivery and causing high losses due to downtime.

[0046] To address the above issues, the company introduced the cloud-based collaborative remote operation and maintenance management method for laser equipment proposed in this invention in September 2024. The deployment includes: installing high-frequency vibration sensors, real-time power monitoring modules, and integrated temperature / flow rate acquisition modules on each laser device to construct an edge acquisition unit; preprocessing the data through edge computing devices to extract key operational characteristics such as power, current, voltage, cooling temperature, coolant flow rate, and equipment vibration intensity; and uniformly encoding and uploading the device number, acquisition time, and parameter fields to the cloud database. A particle generation network strategy generation module is deployed on the cloud platform, using 4560 high-quality operation and maintenance cases accumulated by the company over the past two years to construct training samples, embedding features and modeling strategy patterns for the equipment's operating status; simultaneously, a fireworks algorithm optimization module is deployed to optimize local perturbations and compensate for the particle network's shortcomings in local accuracy. Finally, after optimization through a fusion compensation mechanism, the strategy is distributed from the cloud platform to the edge control module to execute automated fault response and scheduling plans. The feedback module collects the response latency, execution success rate, task duration, and resource consumption of each operation and maintenance task, transmitting this data back to the cloud in real time for closed-loop updates of the strategy model.

[0047] For example, at 2:43 AM on October 12, 2024, a laser cutting machine in the company's workshop experienced an abnormal Z-axis control offset. Traditionally, such faults cannot be detected at night and are only confirmed during morning inspections. However, after deploying the system of this invention, the sensor detected abnormal power and vibration amplitude within 3 seconds of the offset occurring. After edge node preprocessing and matching through a fusion strategy, it was determined to be a potential Z-axis drive stepping fault. The system automatically sent the strategy to the edge control module, instructing personnel to calibrate the drive module and reset the control logic on-site. The entire process took approximately 11 minutes, effectively preventing production line downtime due to Z-axis abnormalities and saving an estimated 82,000 yuan in daily production capacity.

[0048] During the three months of implementing this invention, a total of 62 minor early warning faults in the equipment were identified and handled, of which 42 were proactively identified for the first time, completing strategy generation, scheduling, and fault resolution without manual intervention. The average system response time was reduced from 264 minutes to 22 minutes, and the task execution success rate increased from 89.4% to 98.7%. The stable operating time of the equipment increased by 12.8% year-on-year, and the average monthly maintenance cost decreased by approximately 17.3%, greatly alleviating the pressure on maintenance manpower and the risk of fault misjudgment.

[0049] Table 1. Comparison of the effects of remote operation and maintenance management of laser equipment based on cloud collaboration.

[0050] Table 1 demonstrates that this invention not only significantly improves the efficiency of remote strategy generation and scheduling response, but also achieves substantial improvements in several key indicators such as equipment operation stability, labor input costs, and intelligent strategy execution. This fully proves that the invention possesses high engineering feasibility, practicality, and promotional value. The system can be widely applied to remote intelligent operation and maintenance scenarios for industrial equipment such as laser processing, CNC machine tools, and intelligent manufacturing production lines, exhibiting good versatility and scalability.

[0051] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A cloud-based collaborative laser device remote operation and maintenance management method, characterized in that, Comprise the following steps: S1, collect the operating parameters of the laser equipment, construct the operation data set of remote operation and maintenance, preprocess the operation data set, and upload to the cloud platform; S2, input the equipment state characteristics obtained by preprocessing into the particle generation network, learn the solution space distribution based on historical high-quality operation and maintenance strategies, and construct an initial remote operation and maintenance strategy set; S3, generate multiple sparks according to the initial remote operation and maintenance strategy set, each spark representing a strategy, apply the firework algorithm for local perturbation optimization, and construct a candidate strategy set; S4, compensate for the local precision deficiency in the initial remote operation and maintenance strategy generated by the particle generation network on the cloud platform, and construct the final remote operation and maintenance strategy through the initial remote operation and maintenance strategy and the candidate strategy optimized by the firework algorithm; S5, download the final remote operation and maintenance strategy to the edge node through the cloud platform to drive fault response, maintenance scheduling and maintenance task execution; S6, collect operation and maintenance feedback results and synchronize to the cloud platform to update the particle generation network and the firework algorithm parameters. 2.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The operating parameters specifically include laser power, current, voltage, cooling temperature, cooling liquid flow rate, ambient temperature and vibration signal intensity. 3.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The S1 specifically comprises: Sort and index the operating parameters according to equipment number, sampling time and parameter field, construct the operation data set of remote operation and maintenance, preprocess the operation data set, and periodically package and upload to the cloud. 4.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The S2 specifically comprises: S21, receive the operation data set uploaded to the cloud platform, perform statistical analysis and feature extraction on the operation data set, extract key behavior parameters, state change trend, fluctuation range and mutual correlation in the equipment operation process, and construct a unified dimension equipment state feature vector; S22, perform nonlinear dimension reduction mapping on the equipment state feature vector through the encoding module of the particle generation network, use the historical high-quality operation and maintenance strategies stored on the cloud platform to train the particle generation network generation module, learn the solution space distribution, and construct an initial remote operation and maintenance strategy set: ； wherein, is an initial set of remote operation policies, is the th generated remote operation policy, is the th component of the policy is the th component of the historical mean policy, is the th component of the policy is the square of the th variable of the latent space corresponding to the policy vector, n is the dimension of the policy vector, d is the dimension of the latent space, is a regularization factor, is a constraint threshold to control the boundary of the policy set, is a regularization control item in the process of constructing the policy generation, is the th generated remote operation policy, is the feature vector index in the remote operation policy vector, is the th dimension of the latent space vector in the particle generation network, is the square of the deviation of the th policy in the th dimension from the historical policy mean.

5. The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The S3 specifically comprises: S31, set the disturbance range and the upper limit of the optimization iteration number of each initial individual as the initial individual of the firework algorithm for each strategy in the initial remote operation and maintenance strategy set ; S32, in each iteration, perform local perturbation optimization around each initial strategy within the perturbation range, generate a spark solution set and construct a candidate strategy set, and the optimization objective satisfied by the candidate strategy set is: ； wherein, is the th candidate strategy, is the perturbation solution, is the th generated remote operation strategy, is the strategy centered local perturbation search range radius, is the solution space region centered at with radius , is the th component of the candidate strategy, is the th component of the historical average strategy, is the th weighting coefficient of the strategy dimension, is the diversity adjustment factor, is the remote operation dimension index, n is the dimension of the remote operation strategy vector, is the dimension index of the strategy vector, is the feature vector index in the remote operation strategy vector, is the square of the difference between the th component of the candidate strategy and the th component of the historical average strategy , is the difference value of the pair of remote operation dimension vectors with the largest difference. S33, evaluate the fitness of all candidate strategies, select the individual with the optimal fitness in each round to join the update set, and iterate until the maximum number of rounds T is reached. 6.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The S4 specifically comprises: For the deficiency of the initial remote operation and maintenance strategy set generated by the particle generation network in local precision, a fusion mechanism based on double-layer compensation is used for strategy construction, each strategy in the initial strategy set and the corresponding candidate strategy optimized by the firework algorithm are indexed and paired, and a fusion strategy is constructed through local correction and global direction offset, the fusion target not only minimizes the deviation between the final remote operation and maintenance strategy and the candidate strategy and the initial remote operation and maintenance strategy, but also controls the difference amplitude between the final remote operation and maintenance strategy and the candidate strategy and the initial remote operation and maintenance strategy: ； wherein, is a final remote operation and maintenance policy set, is a final remote operation and maintenance policy generated by fusion, is a component of the initial policy in the is a candidate solution after fusion, and is a component of the initial policy in the is a candidate policy, is a component of the current fusion in the is a fusion weight coefficient of the is a component of the current fusion in the is a fusion weight coefficient of the is a compensation adjustment factor of the is a feature vector index in the remote operation and maintenance policy vector, is an index of the is a compensation adjustment factor of the is a feature vector index in the remote operation and maintenance policy vector, is an index of the is a policy in the fusion policy set, is a remote operation and maintenance dimension index, and n is a dimension of the remote operation and maintenance policy vector, is a dimension index of the policy vector, is a global regularization coefficient, is a difference value of a pair of remote operation and maintenance dimension vectors with the largest difference, is a square of a deviation between the and the fusion target value. 7.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The edge node specifically comprises an edge communication module, a local execution control unit, an edge perception and feedback collection module, and an edge computing processing unit. 8.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The S5 specifically comprises: S51, constructing a structured scheduling instruction set for edge node execution based on the optimization characteristics of each policy in the final fusion policy set on the cloud platform; S52, issuing the scheduling instruction set to the corresponding edge node through the remote communication module, and determining the activation state of each instruction based on a multi-factor fusion scoring mechanism, after the instruction is triggered, the edge node drives the execution of remote operation and maintenance actions including fault response, repair scheduling, and maintenance tasks: ； wherein, is the th is the index of the candidate strategy in the set is the set of current activatable candidate strategy indexes, is the index of the feature vector in the remote operation strategy vector, is the th is the component of the th is the importance weight of the th is the resource scheduling factor weight coefficient, is the index of the resource item, is the occupation ratio of the current task to the th is the availability score of the th is the time conflict penalty factor, is the remote operation dimension index, is the planned execution time point of the th is the execution time point of the remote operation strategy executed by the th is the index number of the remote operation scheduling instruction currently being evaluated, n is the dimension of the remote operation strategy vector, and m is the number of scheduled remote operation strategies.​​​​​​​​​ 9.The cloud-based collaborative laser device remote operation and maintenance management method of claim 1, wherein, The operation and maintenance feedback result specifically comprises a task execution success rate, a response time, a task execution time, and a resource consumption amount; The cloud-based cooperative laser device remote operation and maintenance management method according to claim 1 is characterized in that, The S6 specifically comprises: S61, continuously collecting operation and maintenance feedback data of the edge node during the execution of the remote operation and maintenance task, and periodically synchronizing to the cloud platform; S62, updating the particle generation network in the cloud platform, correcting the spatial sampling distribution of the particle generation network according to the response time change and resource consumption level in the feedback data, adjusting the generation module parameters of the particle generation network including the fusion policy vector output structure and the feature mapping mode according to the success rate and adaptability change of the fusion policy in actual execution; S63, updating the key parameters of the firework algorithm in the cloud platform, updating the disturbance range, the number of sparks generated, and the search granularity in the firework algorithm according to the timeliness and local convergence characteristics of the fusion policy set in the edge node execution.

Images