Medical equipment predictive maintenance method based on artificial intelligence and related equipment

Abstract

The invention discloses a medical equipment predictive maintenance method based on artificial intelligence and related equipment, and relates to the technical field of medical equipment state maintaining.The method comprises the steps that multiple maintenance instructions of multiple pieces of medical terminal equipment are received, and the long connection state of each piece of medical terminal equipment is confirmed through a server and by means of a heartbeat detection mechanism; if the medical terminal equipment is in a long connection state, combining maintenance index data corresponding to all maintenance instructions and load index data of all servers, and utilizing a load balancing technology to distribute the maintenance instructions for the servers; determining the maintenance type of each medical terminal device according to the maintenance instruction, and selecting a corresponding maintenance mode based on the maintenance type to maintain the medical terminal device; judging whether the current maintenance instruction is completed or not; and if the maintenance instruction is completed, outputting a medical terminal equipment maintenance scheme and visually displaying the medical terminal equipment maintenance scheme. The method can effectively improve the overall operation efficiency of a hospital.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of medical equipment status maintenance, and in particular to an artificial intelligence-based medical equipment predictive maintenance method and related equipment. Background Art

[0002] In the modern hospital environment, the stable operation of medical equipment is crucial for providing efficient, high-quality medical services. Cashier counters and self-service kiosks are key patient-facing service terminals in hospitals, and their operational status directly impacts both the patient experience and the hospital's efficiency. Therefore, ensuring the stable operation of medical equipment is crucial.

[0003] In daily hospital operations, a large number of patients use billing counters to settle their bills, while self-service kiosks are used for operations such as registration and checking test reports. These devices are located throughout the hospital, including outpatient halls, inpatient departments, and other high-traffic areas. Due to frequent use and complex operating environments, medical equipment is prone to hardware failures and consumable shortages. For example, mechanical component wear and aging can occur. Failure to promptly identify and resolve these issues will lead to service interruptions and patient waiting times, impacting the hospital's overall operational efficiency. Summary of the Invention

[0004] The embodiments of the present application provide an artificial intelligence-based predictive maintenance method for medical equipment and related equipment, which are used to ensure the stable operation of medical equipment and improve the overall operational efficiency of the hospital.

[0005] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions: In a first aspect, a method for predictive maintenance of medical equipment based on artificial intelligence is provided, which is applied to a server connected to multiple medical terminal devices, and includes the following steps: Receive multiple maintenance instructions for multiple medical terminal devices, and confirm the long connection status of each medical terminal device through the server and using the heartbeat detection mechanism; If the medical terminal device is in a long connection state, the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers are combined, and the load balancing technology is used to distribute the maintenance instructions to the servers; Determine the maintenance type of each medical terminal device according to the maintenance instruction, and select the corresponding maintenance method based on the maintenance type to maintain the medical terminal device; Determine whether the current maintenance instruction is completed; If the maintenance instruction has been completed, the medical terminal equipment maintenance plan is output and the medical equipment maintenance plan is visually displayed.

[0006] In a possible implementation of the first aspect, confirming the long connection status of each medical terminal device by the server and using a heartbeat detection mechanism includes: Receive the heartbeat packet sent by the medical terminal device at a predetermined time, and forward the heartbeat packet to the server, so that the server feeds back a response signal after receiving the heartbeat packet; Feedback the response signal to the medical terminal device, and determine whether the medical terminal device receives the response signal within a preset time period; If the medical terminal device does not receive a response signal within a preset time period, the medical terminal device is determined to be in an offline state; If the medical terminal device receives a response signal within a preset time period, it is confirmed that the medical terminal device is in a long connection state.

[0007] In a possible implementation of the first aspect, combining maintenance indicator data corresponding to all maintenance instructions and load indicator data of all servers and using load balancing technology to allocate maintenance instructions to servers includes: Construct a valued bipartite graph based on maintenance index data and load index data; Use the Kuhn-Munkres algorithm to solve the optimal matching of the weighted bipartite graph and output the optimal matching result; Maintenance instructions are assigned to servers according to the best matching results.

[0008] In a possible implementation of the first aspect, constructing a valued bipartite graph according to the maintenance indicator data and the load indicator data includes: Calculate the load rate of each server processing the maintenance instruction based on the load indicator data; Calculate the standardized task volume and standardized expected completion time based on maintenance indicator data; A bipartite graph is constructed based on the maintenance instructions and the servers, wherein identifiers of the plurality of maintenance instructions are used as a left vertex set of the bipartite graph, and identifiers of the plurality of servers are used as a right vertex set of the bipartite graph; For any maintenance instruction and server, the actual execution time of the maintenance instruction is determined using the standardized task volume and load rate; If the actual execution time is less than or equal to the normalized expected completion time, then an edge of the bipartite graph is established; If the actual execution time is greater than the normalized expected completion time, the edge of the bipartite graph is not established; Based on the standardized task volume and load rate, the matching degree between the maintenance instruction and the server is calculated using the matching degree algorithm; The matching degree is used as the weight value of the edge connecting the maintenance instruction and the server in the bipartite graph to obtain the assigned bipartite graph.

[0009] In a possible implementation of the first aspect, the maintenance type includes software maintenance and hardware maintenance, and selecting a corresponding maintenance method based on the maintenance type to maintain the medical terminal device includes: When the maintenance type is software maintenance, extracting version information of the software application in the medical terminal device from a preset database; Filter the updated device list based on version information and device type of medical terminal devices; Use the pre-update backup mechanism to back up the software data of the software applications in the update device list; Use breakpoint resume technology to transmit update data to the corresponding medical terminal device in the update device list and update the software application; After the update is completed, the software application functions are automatically verified using the post-update verification mechanism; If the verification fails, the rollback mechanism is automatically triggered to restore the software data to the backup state; If the verification is successful, it is determined that the software application is updated successfully.

[0010] In a possible implementation of the first aspect, the method further includes: When the maintenance type is hardware maintenance, collect historical operating status data and real-time working condition data of medical terminal equipment; Based on the dual-tree complex wavelet transform and nonlinear time series projection method, the historical operation status data is denoised and the time series characteristics and frequency domain characteristics of the historical operation status data are extracted; Fuse the time series features and frequency domain features to obtain a feature parameter set; Perform kernel principal component analysis on the feature parameter set, select the kernel principal component with a cumulative contribution rate higher than the preset threshold and map the corresponding feature vector to the kernel principal component space to obtain the test kernel principal component; The test kernel principal component was used as a covariate to construct a Weibull proportional failure rate model; Input the real-time working condition data into the Weibull proportional failure rate model to calculate the failure rate of the medical terminal equipment; Integrate the failure rate to calculate the reliability and draw the reliability curve of the medical terminal equipment; Use the feature parameter set and reliability curve to train the preset LSTM prediction model; Input real-time operating condition data into the trained LSTM prediction model, and output the predicted hardware life of the medical terminal equipment through the LSTM prediction model; Generate a hardware maintenance plan by combining the preset fault maintenance interval and the predicted hardware life.

[0011] In a possible implementation of the first aspect, performing denoising on the historical operating status data based on a dual-tree complex wavelet transform and a nonlinear time series projection method, and extracting time series features and frequency domain features of the historical operating status data includes: The dual-tree complex wavelet transform is used to perform preliminary noise reduction on the historical operating status data containing noisy signals to obtain the characteristic signal and noise signal of the historical operating status data; The nonlinear time series projection method is used to project the characteristic signal and the noise signal into different spaces to reconstruct the historical operation status data and obtain the denoised historical operation status data. The time series and frequency domain features of the historical operating status data after noise reduction are extracted through the wavelet transform algorithm.

[0012] In a possible implementation of the first aspect, generating a hardware maintenance plan based on a preset fault maintenance interval and a predicted hardware lifespan includes: Determine whether the hardware maintenance time of the medical terminal equipment in the previous cycle exceeds the preset fault maintenance interval; If the preset fault maintenance interval is exceeded, an emergency maintenance work order will be triggered; If the preset fault maintenance interval has not been exceeded, then determining whether the predicted hardware life exceeds the preset fault maintenance interval of the next cycle; If the preset fault maintenance interval for the next maintenance is exceeded, the hardware status of the medical terminal device is continuously monitored; If the preset fault maintenance interval for the next maintenance is not exceeded, the hardware maintenance of the medical terminal device will be added to the maintenance plan for the next cycle.

[0013] In a second aspect, the present application provides a machine-readable storage medium having stored thereon instructions for enabling a machine to execute the above-mentioned artificial intelligence-based predictive maintenance method for medical equipment.

[0014] In a third aspect, the present application provides a server, comprising: a memory configured to store instructions; and The processor is configured to call instructions from the memory and implement the above-mentioned artificial intelligence-based predictive maintenance method for medical equipment when executing the instructions.

[0015] The above technical solution uses a heartbeat detection mechanism to confirm the persistent connection status of medical terminal devices, ensuring that servers can access real-time status information for medical terminal devices. This allows maintenance personnel to stay informed of the latest device updates, providing a foundation for timely response to device issues and effectively improving maintenance efficiency and accuracy. Allocating maintenance instructions to servers based on maintenance and load metrics enables precise allocation of maintenance resources and avoids resource waste. Real-time monitoring and historical data analysis can quickly detect abnormal changes in device component status and potential fault signs, providing timely warnings to maintenance personnel. This significantly shortens the time between fault occurrence and detection, helping to avoid equipment downtime caused by unresolved faults and ensuring the continuity of medical services. Intelligently determining the maintenance type for medical terminal devices based on task and load metrics improves the targeting and accuracy of maintenance and enhances the efficiency of fault repair. Medical equipment maintenance plans are displayed in real time on a large screen, displaying key information such as device location, operating status, utilization rate, and remaining consumables using intuitive graphics, charts, and color coding. Managers can quickly understand the overall operational status of equipment throughout the hospital, promptly identify potential problems and abnormal devices, and intuitively understand maintenance content and progress, thereby improving overall hospital operations and management.

[0016] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of an artificial intelligence-based predictive maintenance method for medical equipment provided in an embodiment of the present application; Figure 2 A schematic diagram of a structure for confirming the long connection status of each medical terminal device provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of software maintenance provided in an embodiment of the present application; Figure 4 A flowchart of hardware maintenance provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0019] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), such directional indications are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0020] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0021] Figure 1 The following schematically shows a flow chart of a method for predictive maintenance of medical equipment based on artificial intelligence according to an embodiment of the present application. Figure 1 As shown, an embodiment of the present application provides an artificial intelligence-based predictive maintenance method for medical equipment, which may include the following steps.

[0022] S110, receiving multiple maintenance instructions for multiple medical terminal devices, and confirming the long connection status of each medical terminal device through the server and using a heartbeat detection mechanism; S120: If the medical terminal device is in a long connection state, combine the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers, and use load balancing technology to distribute maintenance instructions to the servers; S130, determining the maintenance type of each medical terminal device according to the maintenance instruction, and selecting a corresponding maintenance method based on the maintenance type to maintain the medical terminal device; S140, determining whether the current maintenance instruction is completed; S150: If the maintenance instruction has been completed, the medical terminal equipment maintenance plan is output and the medical equipment maintenance plan is visually displayed.

[0023] First, multiple maintenance instructions are received from multiple medical terminal devices. The server then uses a heartbeat detection mechanism to confirm the persistent connection status of each medical terminal device. In modern medical environments, a wide variety of medical terminal devices exist, such as monitors, infusion pumps, ultrasound diagnostic equipment, and electrocardiographs. These devices generate and process large amounts of medical data and require regular maintenance and management to ensure proper operation and data accuracy. Different medical terminal devices may have various types of maintenance instructions. For example, software update instructions are used to update the device's operating system or application; parameter calibration instructions are used to calibrate the device's measurement parameters to ensure the accuracy of measurement results; and troubleshooting instructions are used to diagnose and repair device faults. The server receives multiple maintenance instructions for multiple medical terminal devices. The server is the core control and management center of the entire system. It receives maintenance instructions from various sources and then categorizes, stores, and processes them. The server must accurately send maintenance instructions to the corresponding medical terminal device based on information such as device type and status. Furthermore, the server must monitor and manage the device's status in real time. The server receives maintenance instructions from multiple medical terminal devices, which may contain information such as the device ID, instruction type, and execution time. The server also uses a heartbeat detection mechanism to confirm the persistent connection status of each medical terminal device. For each device, the server determines whether its connection status is normal, abnormal, or offline based on the heartbeat packet received. Only when the device is confirmed to be in a normal persistent connection state will the server reliably send maintenance instructions to the device, ensuring that the instructions are correctly received and executed. A persistent connection maintains a continuous communication link between the medical terminal device and the server over a period of time, rather than reestablishing a connection for each communication. This connection method improves communication efficiency and reduces the overhead of establishing and disconnecting connections.

[0024] Secondly, if the medical terminal device is in a persistent connection state, the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers are combined, and load balancing technology is used to distribute maintenance instructions to the servers. When the medical terminal device confirms a persistent connection with the server through heartbeat detection, it means that the medical terminal device is online and the communication link is stable. The server can send maintenance instructions to it and ensure that the instructions are received and executed. At this point, the system needs to further solve the problem of how to reasonably distribute maintenance tasks to multiple servers to avoid overloading some servers and idle others, thereby improving overall efficiency. Therefore, load balancing technology is used to distribute maintenance instructions to servers. Load balancing is a technology that distributes tasks to multiple servers to avoid overloading a single server. The goal is to optimize resource utilization, improve system stability, and improve responsiveness. Specifically, maintenance indicator data is obtained from the maintenance instructions, and server load indicator data is obtained from the servers. Subsequently, load balancing technology is used to evenly distribute maintenance instructions to the servers based on the maintenance indicator data and load indicator data, avoiding overloading of a single server.

[0025] Next, the maintenance type of each medical terminal device is determined according to the maintenance instruction, and the corresponding maintenance method is selected based on the maintenance type to maintain the medical terminal device. In this embodiment, the maintenance types are software maintenance and hardware maintenance. Software maintenance refers to updating, repairing, optimizing, and other operations on the software system of the medical terminal device; hardware maintenance refers to inspecting, replacing, repairing, and other operations on the hardware components of the medical terminal device. Determine the maintenance type of each medical terminal device according to the maintenance instruction When the server or management system receives a maintenance instruction for a medical terminal device, it is first necessary to analyze the content and target of the instruction to determine whether the maintenance task belongs to software maintenance or hardware maintenance, so as to subsequently select the corresponding maintenance method. Secondly, determine whether the maintenance instruction is software maintenance or hardware maintenance. If the maintenance instruction is software maintenance, the medical terminal device is maintained in the form of software maintenance. If the maintenance instruction is hardware maintenance, the medical terminal device is maintained in the form of hardware maintenance.

[0026] Subsequently, determining whether the current maintenance instruction has been completed involves a series of checks and verifications to confirm whether the maintenance task for the medical terminal device has achieved the intended objectives and met quality standards. In this embodiment, the completion of the current location instruction can be determined by verifying the functionality of the software application in the medical terminal device. Passing the verification of the software application functionality indicates that the verification is complete.

[0027] Once maintenance tasks for medical terminal devices (such as monitors and ventilators) are completed and verified, the system automatically generates a complete maintenance plan document and displays the plan contents in the form of visual charts or interfaces, making it easy for medical staff, equipment managers, or engineers to quickly review and analyze them. In this embodiment, the visual display can display the dynamic status of all devices in real time on a large screen. At the data level, data aggregation and caching technologies are used to aggregate and process various status data from different devices in real time and store it in a high-performance cache such as Memcached or Redis, ensuring rapid response to data requests from the large screen. For visualization, advanced visualization libraries such as Highcharts or Plotly are used, combining hospital business characteristics and user needs to design an intuitive, beautiful, and easy-to-understand visualization interface. Through various formats such as graphics, charts, and animations, information such as device location distribution, operating status, utilization rate, and consumables remaining can be displayed. For example, heat maps are used to display the regional distribution of device utilization, and dynamic progress bars are used to display consumables remaining. At the same time, the interactive design is integrated to support users to filter equipment status, view detailed information, trace historical data and other operations through the large screen, realize in-depth mining and utilization of data, and help managers quickly grasp the overall operation status of the equipment and make scientific decisions.

[0028] Using a heartbeat detection mechanism to confirm the persistent connection status of medical terminal devices ensures that servers receive real-time status information, allowing maintenance personnel to stay informed of the latest device updates. This provides a foundation for timely response to device issues and effectively improves maintenance efficiency and accuracy. Allocating maintenance instructions to servers based on maintenance and load metrics enables precise allocation of maintenance resources, avoiding resource waste. Real-time monitoring and historical data analysis quickly detect abnormal changes in device component status and potential fault signs, providing timely warnings to maintenance personnel. This significantly shortens the time between fault occurrence and detection, helping to avoid equipment downtime caused by unresolved faults and ensuring the continuity of medical services. Intelligently determining the maintenance type for medical terminal devices based on task and load metrics improves the targeting and accuracy of maintenance and enhances the efficiency of fault remediation. Medical equipment maintenance plans are displayed in real time on a large screen, presenting key information such as device location, operating status, utilization rate, and remaining consumables using intuitive graphics, charts, and color coding. Managers can quickly understand the overall operational status of equipment throughout the hospital, promptly identify potential problems and abnormal devices, and intuitively understand maintenance content and progress, thereby improving overall hospital operations and management.

[0029] In one implementation of this embodiment, confirming the long connection status of each medical terminal device by the server and using a heartbeat detection mechanism includes: S210, receiving a heartbeat packet sent by the medical terminal device at a predetermined time, and forwarding the heartbeat packet to the server, so that the server feeds back a response signal after receiving the heartbeat packet; S220: Feedback the response signal to the medical terminal device, and determine whether the medical terminal device receives the response signal within a preset time period; S230: If the medical terminal device does not receive a response signal within a preset time period, confirming that the medical terminal device is in an offline state; S240: If the medical terminal device receives a response signal within a preset time period, it is confirmed that the medical terminal device is in a long connection state.

[0030] Figure 2 A structural diagram of confirming the long connection status of each medical terminal device provided in an embodiment of the present application is as follows Figure 2 As shown in the figure, advanced network communication technology is required to achieve a persistent connection between the server and medical terminal devices for real-time monitoring of device status. Based on the TCP / IP protocol and utilizing a heartbeat detection mechanism, the medical terminal device periodically sends heartbeat packets to the server, which responds promptly upon receipt to confirm the connection status. If no response is received within the specified time, the server determines the connection is abnormal and attempts to reconnect. If several retries fail, the device will be marked as offline. Furthermore, load balancing technology is introduced to ensure that when the server faces a large number of persistent connection requests from medical terminal devices, it can allocate resources appropriately, ensuring the stability of each connection and data transmission efficiency.

[0031] The heartbeat detection mechanism confirms the persistent connection status of each medical terminal device. This mechanism monitors the connection status between a medical terminal device and a server by periodically sending lightweight signals (similar to heartbeats). After a persistent connection (a persistent communication link) is established between the server and the medical terminal device, the server periodically sends heartbeat packets to each medical terminal device (or the terminal proactively sends them). By monitoring the response to these heartbeat packets, the server can determine in real time whether the connection is functioning properly and whether the device is online. First, establishing a session connection between the medical terminal device and the server involves establishing a dedicated communication channel between the two devices, enabling secure and efficient data and information exchange. The medical terminal device sends a connection request to the server, typically including device identification and authentication information. The server verifies the request and confirms the identity and permissions of the medical terminal device. If verification is successful, the server accepts the connection request and allocates resources for the session. Subsequently, a session channel is established, establishing a dedicated communication channel between the medical terminal device and the server, such as a TCP connection or a WebSocket connection. In this embodiment, the dedicated communication protocols established between the medical terminal device and the server are TCP and IP.

[0032] Next, in response to receiving a heartbeat packet sent by the medical terminal device at a predetermined time, the server confirms receipt of the heartbeat packet and sends a response signal to the medical terminal device. In this embodiment, the predetermined time can be determined based on actual circumstances. A heartbeat packet is a lightweight data packet sent by the medical terminal device to the server at predetermined intervals, notifying the server that it is online and connected. A response signal is a confirmation signal sent by the server to the medical terminal device after receiving a heartbeat packet, notifying the medical terminal device that it has received your heartbeat packet and that the connection is functioning properly. Specifically, the medical terminal device sends a heartbeat packet to the server at predetermined times (e.g., every 30 seconds). The heartbeat packet typically contains information such as a device identifier and a timestamp. The server monitors heartbeat packets from the medical terminal device. Upon receiving a heartbeat packet, the server confirms that the medical terminal device is online. After receiving the heartbeat packet, the server immediately or with a short delay sends a response signal to the medical terminal device. The response signal may contain a confirmation code, a timestamp, or instructions. Upon receiving the response signal, the medical terminal device confirms that the connection with the server is functioning properly. If it does not receive a response signal, the medical terminal device may deem the connection abnormal and trigger a reconnection mechanism.

[0033] Determining whether the medical terminal device receives a response signal within a preset time period means that the medical terminal device checks whether it receives a confirmation signal from the server within a specified time range after sending a request or a heartbeat packet. This process is used to verify the timeliness and reliability of communication and ensure that the connection between the device and the server is normal. In this embodiment, the preset time period can be determined according to actual conditions and is used to determine whether the response signal arrives in time. For example, after the medical terminal device sends a request, it expects to receive a response from the server within 500 milliseconds. The medical terminal device sends a request or a heartbeat packet to the server, and the medical terminal device starts a timer and waits for the server's response. If a response signal is received within the preset time period, the reception time is recorded and the response time is calculated, and it is determined that the communication is normal. If no response signal is received within the preset time period, it is determined that the communication has timed out.

[0034] If the medical terminal device does not receive a response signal within the preset time period, it is determined that the medical terminal device is offline and the medical terminal device is reconnected. Specifically, the medical terminal device sends a request or heartbeat packet to the server, records the sending time, and the medical terminal device starts a timer, waiting for the server's response, and the preset time period begins. If a response signal is received within the preset time period, the reception time is recorded and the response time is calculated. The response time is used to determine that the communication is normal, and subsequent operations continue. If no response signal is received within the preset time period, it is determined that the communication has timed out. At this time, the medical terminal device confirms that it is offline. The medical terminal device starts the reconnection mechanism and attempts to re-establish the connection with the server.

[0035] If the medical terminal device receives a response signal within the preset time period, it is confirmed that the medical terminal device is in a persistent connection state. A persistent connection means that after the medical terminal device establishes a connection with the server, the connection remains alive for a long time without frequent re-handshakes (such as TCP connections and WebSocket sessions). When the medical terminal device receives a response signal within the preset time period, it indicates that the medical terminal device and the server are in a persistent connection state.

[0036] In this embodiment, the assistant program of the medical terminal device communicates with the server via a persistent connection, which is the basis for achieving real-time monitoring and management. The TCP protocol is typically used to establish persistent connections because it provides reliable, connection-oriented data transmission services that ensure that data arrives accurately and in order. To maintain the stability of persistent connections, a heartbeat detection mechanism is used. The assistant periodically sends heartbeat packets to the server to inform the server that it is active; after receiving the heartbeat packet, the server responds with a response packet to confirm that the connection is normal. If no response is received from the server within a certain period of time, the assistant will attempt to resend the heartbeat packet or reconnect to ensure that the connection is not disconnected, allowing device status data (such as location, component status, usage rate, etc.) to be continuously uploaded to the server and receiving management instructions issued by the server.

[0037] The heartbeat detection mechanism confirms the long-term connection status of each medical terminal device, ensuring real-time and reliable communication and reducing communication interruptions. Furthermore, the automatic reconnection mechanism can quickly restore communication in the event of an unexpected disconnection, improving system availability and enhancing the network connection stability of medical devices.

[0038] In one implementation of this embodiment, combining the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers and using load balancing technology to allocate maintenance instructions to the servers includes: S310, constructing a valued bipartite graph based on the maintenance index data and the load index data; S320, using the Kuhn-Munkres algorithm to solve the optimal matching of the weighted bipartite graph, and outputting the optimal matching result; S330: Distribute the maintenance instruction to the server according to the optimal matching result.

[0039] Constructing a valued bipartite graph based on maintenance and load metrics is a graph-theoretic modeling approach that associates tasks and servers and quantifies their relationships. A bipartite graph is a special type of graph whose vertices can be divided into two disjoint sets A and B. Each edge in the graph connects a vertex in set A with a vertex in set B. Set A can represent maintenance instructions for multiple medical devices (with maintenance metrics characterizing tasks such as urgency and complexity), while set B can represent servers (with load metrics characterizing server status such as CPU utilization and memory usage). Assignment involves assigning a numerical value to each edge in the bipartite graph. This value represents the cost or benefit of assigning a maintenance instruction to a server. For example, this value could represent the time or resource consumption required for the server to complete the task. By constructing such a valued bipartite graph, the maintenance instruction assignment problem can be transformed into a matching problem in graph theory, leading to the identification of the optimal task allocation solution. First, define the vertex set of the bipartite graph. In this embodiment, the vertex set of the bipartite graph is the task vertex set and the server vertex set, where the task vertex set refers to treating the maintenance instructions of each medical terminal device as a vertex, for example, A={a1, a2, ..., an}, where n is the number of maintenance instructions; the server vertex set refers to treating each server as a vertex, for example, B={b1, b2, ..., bm}, where m is the number of servers. Secondly, for each task vertex and each server vertex, it is necessary to calculate the assignment of the edges between them. This assignment can be calculated based on the maintenance index data and the load index data. After obtaining the assignment, the above vertex set and edge assignment are used to construct a bipartite graph G=(V, E), where V=A∪B is the vertex set, E is the edge set, and each edge has a corresponding assignment, thereby obtaining an assignment bipartite graph.

[0040] After obtaining the assigned bipartite graph, the Kuhn-Munkres algorithm is used to solve the optimal matching of the weighted bipartite graph and output the optimal matching result, which means finding a matching scheme in the bipartite graph with weighted edges so that the sum of the weights on the matching edges is maximized, and outputting this matching scheme. The optimal matching means finding a matching method among all possible matching schemes so that the sum of the weights of the matching edges is maximized. The Kuhn-Munkres algorithm is used to solve the optimal matching problem of weighted bipartite graphs. The goal of the algorithm is to find an optimal matching in a bipartite graph so that the sum of the weights of the matching edges is maximized. Each vertex can be assigned a label (top label). Let vertex A of vertex set A be i The top mark is lx [ i ], vertex y in vertex set B j The top mark is ly [ j ]. Initially, in order to makelx [ i ]+ ly [ j ]≥ weight ( i , j ) is always valid, usually let lx [ i ] for all vertex x i The maximum weight of the associated edge, ly [ j ]=0. Then, the Kuhn-Munkres algorithm is used to find the complete matching of equal subgraphs, which are composed of all the equal subgraphs that satisfy lx[i]+ly[j]=weight(i,j) The edge ( i , j ). In an equal subgraph, a path starting from an unmatched vertex, alternating between matched and unmatched edges, and finally reaching another unmatched vertex can be found using the Hungarian algorithm. An augmenting path can be found in the equal subgraph. If an augmenting path is found, the matched and unmatched edges on the path are swapped to obtain a larger match. If no augmenting path is found, it indicates that the augmenting path does not exist in the current equal subgraph, and the vertex index needs to be adjusted. The process of finding the augmenting path and adjusting the vertex index is repeated until a complete match (i.e., every vertex in the set has a match) is found. This complete match is output as the optimal match. After calculating the optimal match between the maintenance instructions of the medical terminal device and the server using the Kuhn-Munkres algorithm, each maintenance instruction (i.e., the maintenance instruction for the medical terminal device) is explicitly assigned to a specific server for execution based on this match. For example, if the algorithm outputs instruction A → server 1 and instruction B → server 2, the assignment is made accordingly. Server status can also be monitored in real time. If load abnormalities occur (e.g., a server is overloaded), the matching can be recalculated and the assignment adjusted.

[0041] By constructing a weighted bipartite graph based on maintenance indicator data and load indicator data, and using the Kuhn-Munkres algorithm to solve the optimal matching, maintenance instructions are distributed to the target servers. This ensures that the load of each server is as balanced as possible, avoiding overloading of some servers while others are idle. Load balancing optimization improves maintenance efficiency and resource utilization.

[0042] In one implementation of this embodiment, constructing a valued bipartite graph based on maintenance index data and load index data includes: S410, calculating the load rate of each server processing the maintenance instruction based on the load index data; S420: Calculate the standardized task volume and the standardized expected completion time based on the maintenance indicator data; S430: construct a bipartite graph based on the maintenance instructions and the servers, wherein identifiers of the plurality of maintenance instructions are used as a left vertex set of the bipartite graph, and identifiers of the plurality of servers are used as a right vertex set of the bipartite graph; S440: For any maintenance instruction and server, determine the actual execution time of the maintenance instruction using the standardized task volume and load rate; S450: If the actual execution time is less than or equal to the normalized expected completion time, then establish an edge of the bipartite graph; S460: If the actual execution time is greater than the normalized expected completion time, then the edge of the bipartite graph is not established; S470, calculating the matching degree between the maintenance instruction and the server based on the standardized task volume and load rate and using a matching degree algorithm; S480 , using the matching degree as the weight value of the edge connecting the maintenance instruction and the server in the bipartite graph to obtain a valued bipartite graph.

[0043] In the predictive maintenance scenario of medical equipment, the server needs to process maintenance instructions for multiple medical terminal devices. Calculating the load rate of each server processing maintenance instructions based on load index data refers to evaluating the server's ability to process maintenance instructions per unit time based on indicators related to the server's current load situation. The load rate can be understood as the speed at which the server processes maintenance instructions. It reflects the server's work efficiency and processing capacity under the existing load. By calculating the load rate, we can better understand the performance status of the server so that maintenance instructions can be reasonably allocated to avoid overloading some servers while other servers are idle, thereby improving the efficiency and stability of the entire maintenance system. The load rate formula is used to calculate the load rate of each server processing maintenance instructions as shown below:

[0044] The resource consumption coefficient is a weighted value of the utilization of resources such as CPU, memory, and I / O. For example, the resource consumption coefficient = 0.4 × CPU utilization + 0.3 × memory utilization + 0.3 × I / O utilization.

[0045] For example, if the server processes 10 instructions within 1 hour, the average CPU utilization is 60%, the memory utilization is 20%, and the I / O utilization is 20%. The resource consumption coefficient = 0.4×0.6+0.3×0.2+0.3×0.2=0.36, and the load rate = 10 / 1+0.36=7.35.

[0046] After obtaining the load rate, the standardized workload of the maintenance indicator data is calculated using the preset workload formula. This is a common operation in data preprocessing. It aims to convert maintenance indicator data of different specifications or distributions into a unified specification or specific range for more accurate comparison and analysis. The preset workload formula is as follows:

[0047] in, represents the standardized task volume, Indicates the complexity of the task, Denoted as the complexity of the benchmark task, The workload of the benchmark task.

[0048] The standardized task volume of the maintenance indicator data is calculated by the preset task volume formula. The purpose is to compare the complexity of different tasks with the benchmark task J0 and scale the task volume accordingly. For example, if task J k The complexity of is twice that of J0, then its task volume is also considered to be twice that of J0.

[0049] Subsequently, the standardized expected completion time of the maintenance indicator data is calculated using the preset expected completion time formula. In the predictive maintenance of medical equipment, different maintenance tasks have their own characteristics, such as the complexity of the task, the required resources, the degree of urgency, etc. This information constitutes the maintenance indicator data. Each task has an expected completion time, which is affected by many factors. The expected completion time of different tasks may differ in the numerical range and measurement standards, which is not convenient for direct comparison and analysis. Using the preset expected completion time formula to calculate the standardized expected completion time of the maintenance indicator data is to use the pre-set formula, combined with the maintenance indicator data, to convert the expected completion time of different tasks into a unified standard and comparable value. Doing so will help to arrange tasks more scientifically, evaluate server processing capabilities, and perform tasks and other operations. The preset expected completion time formula is as follows:

[0050] in, represents the standardized expected completion time, represents the standardized task volume, represents the workload of the benchmark task, represents the actual running time of the benchmark task, Indicates adjustment parameters. In this embodiment, the adjustment parameters can be determined according to actual needs.

[0051] The standardized expected completion time of maintenance metrics is calculated using a pre-defined expected completion time formula. The goal is to predict the completion time of new tasks based on the ratio of task size to baseline tasks and the runtime of baseline tasks. Beta adjustments can be used to optimize prediction accuracy (e.g., to account for resource contention and cache effects).

[0052] Next, a bipartite graph is constructed based on the maintenance instructions and servers. The identifiers of the multiple maintenance instructions serve as the left vertex set of the bipartite graph, and the identifiers of the multiple servers serve as the right vertex set. A bipartite graph is a special type of graph whose vertices can be divided into two disjoint sets, and each edge in the graph connects only vertices in these two sets. The identifiers of the multiple maintenance instructions can be considered the left vertex set of the bipartite graph, and the identifiers of the multiple servers serve as the right vertex set of the bipartite graph. The edges of the bipartite graph represent the association between maintenance instructions and servers, and this association can be used to optimize task allocation. First, define vertex sets: the left vertex set is the maintenance instructions, and the right vertex set is the servers. Next, list all maintenance instructions to be assigned, for example, I1: CT scanner calibration; I2: MRI cooling system inspection; I3: Ultrasound equipment software upgrade. List all available servers, for example, S1: Maintenance Team A (hardware expertise); S2: Maintenance Team B (software expertise); S3: Remote Support Group.

[0053] For any maintenance instruction and server, the actual execution time of the maintenance instruction is obtained by dividing the standardized task volume by the load rate. In order to reasonably assign maintenance instructions to servers, it is necessary to evaluate the execution time of each maintenance instruction on different servers. The standardized task volume represents the standardized value of comprehensive factors such as the complexity of the maintenance instruction itself and the required resources, while the load rate reflects the server's ability to handle tasks under the current load. The actual execution time of the maintenance instruction is obtained by dividing the standardized task volume by the load rate. The standardized task volume can be regarded as the "workload" of the task, and the load rate can be regarded as the "work efficiency" of the server in processing tasks. According to the principle of "working time = workload ÷ work efficiency", the calculated result is the estimated execution time of the maintenance instruction on a specific server, which helps to more accurately assign tasks and avoid server overload or task backlog. The formula for calculating the actual execution time is as follows:

[0054] In this embodiment, the standardized task volume refers to abstracting the complexity and resource requirements of the maintenance instructions into a numerical value of a unified dimension (such as "task unit"); the load rate refers to the server's ability to process tasks per unit time (tasks / second); and the actual execution time refers to the estimated completion time (seconds) of the maintenance instruction on the target server.

[0055] After determining the actual execution time of a maintenance order, compare it with the standardized expected completion time. If the actual execution time is less than or equal to the standardized expected completion time, an edge in the bipartite graph is established. Previously, the actual execution time of the maintenance order on the server and its standardized expected completion time were calculated. Comparing these two times effectively evaluates whether the server can complete the maintenance order within the expected time. If the actual execution time is less than the standardized expected completion time, it indicates that the server is capable of processing the maintenance order in a timely manner. In this case, an edge is established between the two in the bipartite graph, indicating that the maintenance order can be assigned to that server.

[0056] If the actual execution time is greater than the standardized expected completion time, no edge of the bipartite graph will be established. When constructing a bipartite graph of maintenance instructions and servers, the actual execution time refers to the expected processing time of the maintenance instruction on a specific server (calculated by the standardized task volume and the server load rate). The standardized expected completion time is the ideal processing time preset based on indicators such as task urgency and complexity. If the actual execution time > the standardized expected completion time, it means that the current load or processing capacity of the server cannot complete this maintenance instruction within the expected time. At this time, no edge between the two will be established, which means that the possibility of the server processing this task is excluded. This screening mechanism ensures that the bipartite graph only retains "feasible matching" and provides effective input for subsequent task allocation (such as the Hungarian algorithm to solve the optimal matching). For example, the standardized expected completion time of each maintenance instruction is denoted as T exp ; Actual execution time, recorded as T actual , Maintenance instruction 1: Texp=0.8; Server 1: T actual =0.6≤0.8, establish edge; Server 2: T actual =0.7≤0.8, establish the edge.

[0057] Maintenance instruction 2: T exp =1.2; Server 1: T actual =1.3>1.2, no edge is established; Server 2: T actual =1.1≤1.2, establish the edge.

[0058] Then, for any maintenance instruction and server, the matching degree between the maintenance instruction and the server is calculated based on the standardized task volume and load rate using a preset matching formula. The matching formula is used to quantify the "suitability" of a maintenance instruction for a specific server to execute. Its core is to calculate the compatibility or efficiency between the two using two key indicators: standardized task volume (the workload of the maintenance instruction) and server load rate (the speed at which the server processes tasks). The preset matching formula is as follows:

[0059] in, Indicates the matching degree, represents the workload of the benchmark task, Indicates the server load indicator, Indicates the amount of tasks for the node with the largest amount of tasks, Indicates the load index of the node with the highest server load index.

[0060] When k=I and i=I, w=1, which may indicate a special case, such as a match between the maintenance instruction with the largest workload and the server with the largest load capacity. In this case, the matching degree is 1, and the weight of the edge is also defined as 1.

[0061] When k≠I and i≠I, the matching degree is calculated using the preset matching degree formula. This is the case when calculating task J in general. k With server S i The weight of the corresponding edge is calculated by dividing the difference between the task volume and the load index by the benchmark task volume R I With the benchmark load indicator v I The weight value is obtained by multiplying the two. This means that the weight takes into account the relative differences in task volume and server load indicators.

[0062] In other cases (i.e., when the above two conditions are not met), w = 0, which means that the association weight between the task and the server is 0, which may mean that there is no association or a very weak association between them.

[0063] Finally, the matching degree is used as the weight of the edge connecting the maintenance instruction and the server in the bipartite graph, resulting in a valued bipartite graph. In this bipartite graph, the matching degree reflects the degree of adaptation between the maintenance instruction and the server. Using the matching degree as the edge weight, we can construct a valued bipartite graph, providing data support for subsequent optimization and allocation.

[0064] By constructing a valued bipartite graph using maintenance and load indicator data, we can accurately assess the server's ability to handle maintenance instructions, avoid resource overload or idleness, and accurately allocate resources, which not only improves operational efficiency but also overall execution efficiency.

[0065] In one implementation of this embodiment, selecting a corresponding maintenance method based on the maintenance type to maintain the medical terminal device includes: S510: When the maintenance type is software maintenance, extract the version information of each software application from a preset database; S520, filtering out an update device list based on version information and device type of the medical terminal device; S530, backing up the software data of the software applications in the update device list using a pre-update backup mechanism; S540, using breakpoint resume technology to transmit the update data to the corresponding medical terminal device in the update device list, and update the software application; S550. After the update is completed, automatically verify the functions of the software application using a post-update verification mechanism; S560: If the verification fails, the rollback mechanism is automatically triggered to restore the software data to the backup state; S570: If the verification is successful, it is determined that the software application is updated successfully.

[0066] Figure 3 A structural diagram of software maintenance provided for an embodiment of the present application, first of all, the system should have a version management function to record the version information, update log and compatibility requirements of each application service. During the update, the server filters out a list of devices that need to be updated based on the type of device, the current application service version and other information. Then, the breakpoint resume technology is used to ensure that the update data can be fully transmitted to the terminal device even in the case of unstable network, avoiding update failures due to network interruptions. At the same time, in order to prevent problems from occurring during the update process that affect the normal use of the device, a pre-update backup and post-update verification mechanism is introduced. Before the update, the application services and related data on the device are backed up; after the update is completed, the application service is automatically functionally verified to ensure that the updated application service can operate normally. If the verification fails, it will automatically roll back to the backup state to ensure the stability of the device and business continuity.

[0067] First, extracting the version information of each software application from the preset database means that the version information of each software application in the medical terminal device needs to be obtained or retrieved from the preset database. In this embodiment, the preset database is an existing database with a pre-set structure and content, which may contain relevant information of various software applications, such as name, version, release date, function description, etc. In this embodiment, there is a small assistant program in each medical terminal device. The small assistant program has the local service version management function and records the currently installed version information of each service. At the same time, it keeps synchronization with the server to obtain the latest version information and update log. In this way, the assistant can determine whether the local service needs to be updated and upgraded, and clarify the content and scope of the update.

[0068] Subsequently, using two key criteria, software version information and medical terminal device type, the target device set requiring software updates is precisely identified from the large number of devices. Version information refers to different versions of software applications (e.g., V1.0, V2.3.5), which typically include update content (e.g., new features, compatibility optimizations, bug fixes, etc.); device type refers to the category of medical terminal devices (e.g., new devices, old devices, etc.). Different devices may support different software versions or update policies. By matching the compatibility of software versions and device types, and updating rules (e.g., mandatory updates, recommended updates), a list of devices requiring updates is screened. The software application version currently in the medical terminal device is compared with the latest version. If it is not the latest version of the application software, the medical terminal devices currently supporting these applications are obtained and a list of update devices is compiled.

[0069] Next, a pre-update backup mechanism is used to back up the software data of the software applications on the device list. Before updating software applications, to prevent data loss, corruption, or other issues during the update process, the pre-update backup mechanism copies the software applications and their associated data on the device to be updated and stores them elsewhere. This allows the software applications and their associated data to be restored to their pre-update state if any issues arise during the update, ensuring business continuity and data security. The pre-update backup mechanism backs up the existing data of the software applications on the device before updating them to prevent data loss or corruption due to update anomalies (such as installation failures, version compatibility issues, and data conflicts). This backup allows for rapid data recovery in the event of an update failure, ensuring business continuity and data security. Specifically, first, identify the list of software applications to be updated and determine the types of data to be backed up, such as configuration files, user data, and databases. Next, select a backup method. You can choose the appropriate method based on the importance of your data and how often it's updated. For example, a full backup backs up all relevant data for the software application; an incremental backup backs up only the data that's changed since the last backup, saving storage space and time; and a differential backup backs up only the data that's changed since the last full backup. Then, back up the software data for the software applications listed in the update device list and store the backup data in a secure and reliable location, such as an external hard drive, tape library, or cloud storage. Categorize, label, and archive the backup data for quick retrieval and restoration when needed. Record information such as the time of the backup operation, the backup method, the backup data scope, and the backup storage location.

[0070] Use breakpoint-resume transfer technology to transmit update data to the corresponding medical terminal devices in the update device list and update the software applications. This technology allows data transfer (such as downloading a software update package) to resume at the point where it was interrupted upon reconnection, without having to restart from the beginning. This technology effectively saves time and bandwidth, improving the success rate and efficiency of software updates. First, collect and organize the software application data that needs to be updated. During the transfer process, the amount of data transferred is recorded. If the transfer is interrupted, the transfer can be resumed from the point where it was interrupted. This is typically achieved by saving transfer progress information on the server and client. After receiving and storing the updated data, the medical terminal device performs the appropriate update operations based on the type and content of the updated data, such as replacing old files or modifying configuration files. After the update is complete, verify that the software application is functioning properly to ensure the update was successful.

[0071] Subsequently, after the update is completed, the functions of the software application are automatically verified using the post-update verification mechanism. The post-update verification mechanism refers to a systemic verification of the software's functions, performance, compatibility and other key indicators through a pre-set automated process after the software application update is completed, to ensure that the updated software can operate normally and meet the expected results. Its core purpose is to reduce the cost of manual verification through automated means, quickly discover and intercept unqualified updates, and avoid affecting the user experience or system stability. Through automatic verification, problems that may arise after the update can be discovered and repaired in a timely manner to ensure that the software quality meets expectations. In this embodiment, the assistant program of the medical terminal device will also download the update package according to the update package address provided by the server, and perform an integrity check on the downloaded update package. After the verification is passed, the assistant stops the currently running service according to the predetermined update process, backs up the original service files, then decompresses and installs the update package to the specified directory, and finally restarts the service to complete the update and upgrade process.

[0072] If verification fails, a rollback mechanism is automatically triggered, restoring the software data to a backup state. This mechanism automatically restores the system to its pre-update stable state (typically based on pre-update backup data or versions) when an anomaly (such as a functional failure or data error) is detected during the verification phase of a software update or system change. Its core purpose is to quickly prevent problems with the version from impacting user experience while reducing the cost and time of manual intervention. In this embodiment, the assistant manages various local services, including hardware, voice, payment, and logging. It precisely controls the startup and shutdown of each service based on the device's operational requirements and server instructions. For example, during device startup and initialization, the assistant starts each essential service in a pre-set order. When the device requires maintenance or a service experiences a malfunction, it can stop the relevant service to prevent it from impacting other functions or causing system instability.

[0073] If the verification is successful, the software application is confirmed to have been updated successfully. This verification success means that after the software application is updated, a pre-defined verification mechanism automatically checks the updated software's functionality, performance, and data integrity. If all check items meet expectations (i.e., "verification is successful"), the system automatically determines that the update process has been successfully completed, and no further manual intervention or rollback is required. This process is a critical part of the software update closed loop, ensuring that users ultimately use the new version that is functional, stable, and reliable.

[0074] By extracting the version information of each software application from the preset database and updating the software application, and using the rollback mechanism to confirm whether the software update is successful, it can effectively improve maintenance efficiency and accuracy, enhance data security and reliability, and significantly improve the maintenance efficiency, data security, system stability and user experience of medical terminal equipment.

[0075] In one implementation manner of this embodiment, the method further includes: S601. When the maintenance type is hardware maintenance, collect historical operating status data and real-time operating status data of the medical terminal device; S602, performing denoising on the historical operating status data based on dual-tree complex wavelet transform and nonlinear time series projection method, and extracting time series features and frequency domain features of the historical operating status data; S603: Fusing the time series features and the frequency domain features to obtain a feature parameter set; S604: Perform kernel principal component analysis on the feature parameter set, select kernel principal components whose cumulative contribution rates are higher than a preset threshold and map the corresponding feature vectors to the kernel principal component space to obtain test kernel principal components; S605, constructing a Weibull proportional failure rate model using the test kernel principal component as a covariate; S606: Input the real-time operating condition data into a Weibull proportional failure rate model to calculate the failure rate of the medical terminal device; S607, integrating the failure rate to obtain reliability, and drawing a reliability curve for the medical terminal device; S608: Using the feature parameter set and the reliability curve to train a preset LSTM prediction model; S609: Input the real-time working condition data into the trained LSTM prediction model, and output the hardware predicted life of the medical terminal device through the LSTM prediction model; S610: Generate a hardware maintenance plan based on the preset fault maintenance interval and the predicted hardware life.

[0076] Figure 4 A flowchart of a hardware maintenance process provided in an embodiment of the present application is shown as follows: Figure 4As shown, historical operating status data is first collected and extracted, and features are fused to obtain feature parameters. Kernel principal component analysis is performed on the feature parameters. Kernel principal components and corresponding eigenvectors with cumulative contributions exceeding a threshold are selected, and test kernel principal components are obtained through feature mapping. A Weibull proportional failure rate model is used to generate failure rate and reliability curves. A pre-set LSTM prediction model is trained using the feature parameter set and reliability curve. Real-time operating condition data is input into the trained LSTM prediction model, which then outputs the predicted hardware lifespan of the medical terminal device. A hardware maintenance plan is then derived based on the pre-set maintenance interval and the predicted hardware lifespan.

[0077] When the maintenance type is hardware maintenance, the historical operating status data of the medical terminal device can be collected through a preset database. The historical operating status data may include basic information of the medical terminal device, such as model, serial number, production date, installation date, operating records, and other information. The collection of real-time operating condition data can be achieved using sensors. The real-time operating condition data may include operating status such as the current working mode, operating status (normal, abnormal, standby); performance indicators such as real-time temperature, humidity, pressure, current, voltage, etc.; communication status such as network connection status and data transmission rate. In this embodiment, the historical operating status data may also be temperature data and vibration signal data. The vibration signal data can be used to determine whether the hardware of the medical terminal device has mechanical failures, such as cracks. The temperature data can be used to determine whether the electronic components of the medical terminal device are too hot, thereby causing electronic component failures and affecting the use of the medical terminal device.

[0078] Next, the historical operating status data was denoised using the dual-tree complex wavelet transform and the nonlinear time series projection method, extracting both the temporal and frequency domain features of the historical operating status data. The dual-tree complex wavelet transform utilizes two parallel real wavelet trees (odd-numbered and even-numbered trees) to generate the real and imaginary parts of a complex wavelet, respectively, to form a two-dimensional complex wavelet transform. Compared to traditional wavelet transforms, it offers advantages such as translation invariance, directional selectivity, and low redundancy, making it suitable for processing noisy, nonstationary signals (such as equipment vibration and current fluctuations). It can effectively separate noise (such as sensor errors and environmental interference) from the historical operating status data while preserving the true signal characteristics. The nonlinear time series projection method, targeting nonlinear systems (such as the complex operating states of medical equipment), uses phase space reconstruction techniques (such as the delayed coordinate method) to project univariate time series into a high-dimensional space, revealing hidden dynamical structures. Projection methods (such as nonlinear extensions of principal component analysis (PCA) and independent component analysis (ICA)) further project the high-dimensional data into a low-dimensional space, removing redundant information and preserving key features. Specifically, first, historical operating status data is collected from medical terminal devices, including various sensor data and equipment operating parameters. Secondly, the cleaned data is subjected to a dual-tree complex wavelet transform to obtain wavelet coefficients at different scales. The wavelet coefficients are then filtered by setting a threshold to remove the wavelet coefficients corresponding to the noise components. The filtered wavelet coefficients are inversely transformed to obtain the denoised signal. Next, the denoised signal is reconstructed into phase space using a nonlinear time series projection method, expanding it to a high-dimensional phase space. The background noise and characteristic signals in the signal are separated using a local projection method. The separated characteristic signals are further processed, such as smoothing and filtering, to improve signal quality. Finally, time series features such as mean, variance, and autocorrelation function are extracted from the denoised signal. These features can reflect the statistical characteristics and dynamic changes of the signal in the time domain. The denoised signal is converted from the time domain to the frequency domain by a Fourier transform or wavelet transform. Features are extracted in the frequency domain, such as power spectral density, spectral entropy, center frequency, etc. These features can reflect the energy distribution and frequency characteristics of the signal in the frequency domain, thereby obtaining the time series characteristics and frequency domain characteristics of the historical operating status data.

[0079] After obtaining time series and frequency domain features, they are fused to obtain a feature parameter set. This integrates the time and frequency domain features extracted from the time series data to form a more comprehensive and robust feature representation for tasks such as medical device condition monitoring and fault diagnosis. Specifically, the time and frequency domain features can be fused together to obtain a feature parameter set using a serial fusion method. This method involves directly concatenating the time series and frequency domain features to form a long vector. This method is simple and intuitive, but may overlook correlations between features. Alternatively, features can be fused using weighted fusion, which assigns different weights to different features and then performs a weighted sum. Weights can be determined based on feature importance, relevance, or through a learning algorithm. Finally, after the fusion process, a feature parameter set containing both time series and frequency domain information is obtained. This parameter set can be used for subsequent analysis, modeling, or decision-making.

[0080] Kernel principal component analysis (KPCA) is performed on the feature parameter set. Kernel principal components with cumulative contribution rates exceeding a preset threshold are selected and their corresponding eigenvectors are mapped into the kernel principal component space to obtain the test kernel principal components. KPCA uses a kernel function to map low-dimensional nonlinear data into a high-dimensional linear space, followed by PCA, to extract key features from nonlinear data. This method can reduce the dimensionality of the fused feature parameter set (including time series and frequency domain features), retaining the most representative features and removing redundant information. The cumulative contribution rate represents the proportion of the total variance of the original data explained by the sum of the variances of the first k kernel principal components. If a preset threshold α (e.g., 95%) is set, the minimum k kernel principal components with cumulative contribution rates exceeding α are selected. The kernel principal component space is a low-dimensional space formed by the selected kernel principal components and is used to characterize the main variation patterns of the original high-dimensional features. Specifically, the fused feature parameter set is input into the kernel function, which can be a polynomial kernel function or a radial basis function (RBF) function. The kernel matrix is calculated using the selected kernel function. Next, the kernel matrix K is centered to obtain K~ to eliminate bias caused by the number of samples. Perform eigenvalue decomposition on the centralized kernel matrix K~ to obtain the eigenvalue λ and the corresponding eigenvector α. Sort the eigenvalues from largest to smallest, calculate the cumulative contribution rate of the first m eigenvalues, and select the smallest k such that the cumulative contribution rate exceeds the preset threshold α (such as 90% or 95%). The corresponding first k eigenvectors are the coefficients of the kernel principal component. Using the selected eigenvectors, calculate the formula for the kernel principal component (i.e., the representation of the sample in the kernel principal component space) as shown below:

[0081] Among them, t k is the kth kernel principal component, x is the sample to be mapped, and α is the selected eigenvector.

[0082] Map all test samples to the kernel principal component space to obtain the test kernel principal component matrix.

[0083] Next, the test kernel principal component is used as a covariate to construct a Weibull proportional failure rate model. The Weibull proportional failure rate model is a commonly used life prediction model in reliability engineering, used to describe the law of equipment failure rate changes over time. Its mathematical expression is:

[0084] in, represents the failure rate (instantaneous failure probability) of the equipment at time t when the covariate X is given; represents the baseline failure rate (the failure rate when the covariate is 0, determined by the Weibull distribution); X represents the covariate vector (such as the test kernel principal in this example); It represents the regression coefficient of the covariate, reflecting the influence of each feature on the failure rate.

[0085] In this example, the covariate is the test kernel principal component. Kernel principal component analysis (KPCA) is a nonlinear dimensionality reduction technique that maps high-dimensional data to a low-dimensional space while preserving the data's key features. Therefore, the test kernel principal component can serve as an effective characteristic parameter reflecting the equipment's operating status. The baseline failure rate is adjusted using the covariate X to describe the impact of the equipment's health status on the risk of failure. When X includes features reflecting equipment degradation (such as the kernel principal component in this example), the model can be used to predict the equipment's remaining useful life (RUL) or health status trend. Substituting the test kernel principal component as a covariate into the basic form of the Weibull proportional failure rate model yields the Weibull proportional failure rate model.

[0086] Subsequently, the real-time operating data is input into the Weibull proportional failure rate model to calculate the failure rate of the medical terminal equipment. The failure rate represents the probability of the equipment failing per unit time after a certain moment. The real-time collected operating data of the medical terminal equipment is combined with the Weibull proportional failure rate model to evaluate and predict the failure probability of the equipment under the current operating conditions. First, the trained Weibull model parameters are called to calculate the real-time failure rate. If the regression coefficient corresponding to the covariate increases, the failure rate If the regression coefficient corresponding to the covariate is less than zero, the failure rate decreases when the covariate increases.

[0087] After obtaining the failure rate, the reliability is calculated by integrating the failure rate and plotting the reliability curve for the medical terminal device. Reliability is the probability that the device will complete its specified function under specified conditions and within a specified time. By integrating the failure rate function, the device reliability function can be obtained. The integration starts at 0 (or the time the device begins to be used) and ends at the time of interest, t. The reliability calculation formula is generally expressed as:

[0088] in is the failure rate function.

[0089] To plot a reliability curve for medical terminal equipment, prepare a series of time points t and their corresponding reliability values R(t) based on the reliability calculation results. Using charting software or programming tools (such as MATLAB or Python), plot the reliability curve for the medical terminal equipment using time points t as the horizontal axis and reliability values R(t) as the vertical axis. The reliability curve visually illustrates the reliability trend of the equipment over time.

[0090] A preset LSTM prediction model is trained using a set of feature parameters and reliability curves. The LSTM model's characteristic gating mechanism addresses the vanishing gradient problem of traditional RNNs. In equipment reliability prediction, the reliability curve itself represents time-varying time series data. Feature parameters (such as operating conditions and environmental parameters) exhibit temporal correlations with reliability, making LSTM a suitable tool for capturing dynamic patterns. Using a set of feature parameters and reliability curves to train a preset LSTM prediction model combines multi-dimensional feature data from equipment operation with reliability indicators. Using a deep learning model to uncover temporal patterns in this data, it can predict the future state of the equipment (such as failure rate and remaining life).

[0091] Real-time operating condition data is input into the trained LSTM prediction model, which then outputs the predicted hardware lifespan of the medical terminal device. Specifically, the LSTM layer extracts time series features step by time to capture device performance degradation trends (such as continuously rising temperature and abnormal fluctuations in vibration frequency). The fully connected layer maps the features to lifespan prediction values and outputs the remaining lifespan in days, thereby obtaining the predicted hardware lifespan of the medical terminal device.

[0092] Finally, a hardware maintenance plan is generated by combining the preset maintenance intervals and the predicted hardware lifespan. By comprehensively considering the equipment's preset maintenance intervals and predicted hardware lifespan, a scientific and reasonable hardware maintenance plan is developed to optimize maintenance schedules, reduce the risk of equipment failure, extend equipment lifespan, and ensure safe and stable equipment operation. The preset maintenance intervals are pre-set routine maintenance cycles for equipment (e.g., maintenance every 1000 hours of operation, inspections every three months, etc.) based on industry standards, equipment design specifications, or historical maintenance experience. The predicted hardware lifespan is a data-driven prediction model that predicts the remaining usable time of the hardware based on real-time equipment operating data (such as temperature, vibration, voltage, and operating time).

[0093] Through data processing, feature extraction, model building, and predictive analysis, we can fully and accurately understand the operating status of medical terminal equipment, improve the accuracy of maintenance plans, enhance the accuracy of fault prediction and the reliability of hardware life prediction, and ensure the stable operation of medical equipment. In one implementation of this embodiment, denoising the historical operating status data is performed based on the dual-tree complex wavelet transform and the nonlinear time series projection method, and the time series features and frequency domain features of the historical operating status data are extracted, including: S710, performing preliminary noise reduction on the historical operating status data containing noisy signals using a dual-tree complex wavelet transform to obtain a characteristic signal and a noise signal of the historical operating status data; S720, using a nonlinear time series projection method to project the characteristic signal and the noise signal into different spaces to reconstruct the historical operating status data, thereby obtaining the historical operating status data after noise reduction; S730: Extract the time series features and frequency domain features of the noise-reduced historical operating status data using a wavelet transform algorithm.

[0094] The dual-tree complex wavelet transform (DWT) is used to perform preliminary noise reduction on historical operating status data containing noisy signals, obtaining the characteristic signal and noise signal of the historical operating status data. This involves decomposing the original noisy signal into subband signals at different scales and directions, effectively separating the characteristic signal (i.e., useful information) from the noise signal. First, the DWT performs a multi-level decomposition of the noisy signal. The DWT involves two parallel real wavelet transforms (real and imaginary), generating wavelet coefficients for the real and imaginary parts, respectively. Each level of decomposition decomposes the signal into low-frequency subbands (approximate components) and high-frequency subbands (detail components). The low-frequency subbands contain the main signal features, while the high-frequency subbands primarily contain noise. Next, the wavelet coefficients of the high-frequency subbands are thresholded to remove noise. Thresholding methods can be soft or hard. Soft thresholding sets coefficients below the threshold to zero and subtracts the threshold from coefficients above the threshold, thereby retaining some useful information and suppressing noise. Hard thresholding sets coefficients below the threshold to zero, while leaving coefficients above the threshold unchanged. Subsequently, the processed wavelet coefficients (including the coefficients of the low-frequency subband and the coefficients of the high-frequency subband after threshold processing) are used to reconstruct the signal through the inverse dual-tree complex wavelet transform. The reconstructed signal is the denoised signal, which contains the main features of the original signal, while the noise is effectively suppressed.

[0095] Using nonlinear time series projection to project characteristic signals and noise signals into different spaces, historical operating status data is reconstructed to obtain denoised historical operating status data. This involves separating the characteristic signals (i.e., useful information) from the noise signals in the historical operating status data through nonlinear time series projection and projecting them into different spaces. The projected characteristic signals are then used to reconstruct the data, yielding denoised historical operating status data. Nonlinear time series projection is a method for processing time series data that captures nonlinear characteristics in the data and projects different components (such as characteristic signals and noise signals) into different spaces or subspaces. Specifically, the nonlinear time series projection method projects the preprocessed data into different spaces or subspaces. In this process, characteristic signals and noise signals, due to their different statistical characteristics and dynamic behaviors, are separated into different projection spaces. Within the projection space, the characteristic signals and noise signals are identified and separated. Using the separated characteristic signals, a backprojection or reconstruction algorithm is used to obtain denoised historical operating status data.

[0096] The wavelet transform algorithm is used to extract the time series and frequency domain features of the historical operating status data after noise reduction. Wavelet transform is a signal processing technology that can decompose a signal into wavelet coefficients of different scales and positions, thereby providing local information of the signal in time and frequency. The original signal can be decomposed into wavelet coefficients of different scales through wavelet transform. These coefficients reflect the time series characteristics of the signal at different scales. For example, at a coarser scale, the wavelet coefficients may reflect the long-term trend of the signal; at a finer scale, the wavelet coefficients may reveal the short-term fluctuations or mutation points of the signal. In the operating status monitoring of medical terminal equipment, the wavelet transform can be used to extract the trend of the equipment operating status over time, such as the periodic changes or abnormal fluctuations of key indicators such as temperature and pressure.

[0097] By denoising the historical operating status data using the dual-tree complex wavelet transform (DTCWT) and nonlinear time series projection method, and extracting time series features and frequency domain features, the failure mode and abnormal status of the equipment can be identified, thereby enhancing fault diagnosis and prediction capabilities.

[0098] In one implementation of this embodiment, generating a hardware maintenance plan based on a preset fault maintenance interval and a predicted hardware life includes: S810: Determine whether the hardware maintenance time of the medical terminal device in the previous cycle exceeds the preset fault maintenance interval; S820. If the preset fault maintenance interval is exceeded, an emergency maintenance work order is triggered; S830: If the preset fault maintenance interval has not been exceeded, determine whether the predicted hardware life exceeds the preset fault maintenance interval of the next cycle; S840: If the preset fault maintenance interval for the next maintenance is exceeded, continuously monitor the hardware status of the medical terminal device; S850: If the preset fault maintenance interval for the next maintenance has not been exceeded, the hardware maintenance of the medical terminal device is added to the maintenance plan for the next cycle.

[0099] First, determine whether the medical terminal device's hardware maintenance time in the previous cycle exceeded the preset maintenance interval. In this embodiment, the preset maintenance interval can be a time interval based on factors such as the device type, purpose, importance, and manufacturer recommendations. It is used to determine how often the device should be maintained to prevent failures. Hardware maintenance can include preventive maintenance (regular inspections, cleaning, calibration, etc.) and fault maintenance (repair or replacement of parts after a device failure). The actual number of days from the "last maintenance time" to the "current assessment time" is calculated and compared with the preset interval. If the actual number of days is greater than the preset number of days, it is considered a timeout (maintenance was not timely); if the actual number of days is less than the preset number of days, it is considered a timeout (maintenance is compliant).

[0100] If the preset maintenance interval is exceeded, an emergency maintenance work order is triggered. When the hardware maintenance time for a medical terminal device exceeds the preset maintenance interval, the system automatically or manually generates an emergency maintenance work order to notify the relevant maintenance personnel to immediately perform maintenance on the device to prevent failure or performance degradation due to lack of timely maintenance. In other words, if, for some reason (such as insufficient maintenance personnel or improper maintenance scheduling), the device is not maintained within the preset maintenance interval, it is said that the maintenance time has exceeded the preset interval. When the maintenance time exceeds the preset interval, the system (such as device management software or maintenance management system) automatically generates an emergency maintenance work order. This work order contains basic device information, maintenance requirements, urgency, and the specific content of the maintenance required. Once the work order is generated, the relevant maintenance personnel are immediately notified to perform maintenance on the device.

[0101] If the preset fault maintenance interval has not been exceeded, the hardware predicted lifespan is determined to be longer than the preset fault maintenance interval for the next cycle. When checking the hardware maintenance status of medical terminal equipment, if it is found that the maintenance time of the equipment in the previous cycle did not exceed the preset fault maintenance interval (i.e., the equipment received timely maintenance), the next step is to determine whether the hardware predicted lifespan of the equipment is long enough to cover the preset fault maintenance interval for the next cycle.

[0102] If the preset fault maintenance interval for the next maintenance is exceeded, the hardware status of the medical terminal device will be continuously monitored. When the hardware maintenance time of the medical terminal device is expected to exceed the next preset fault maintenance interval, the hardware status of the device needs to be continuously and closely monitored to ensure that the device can continue to operate safely during the maintenance interval. If potential problems are found, timely measures should be taken. If potential problems or abnormal changes are found in the equipment, relevant maintenance personnel will be notified immediately.

[0103] If the preset maintenance interval for the next maintenance has not expired, the hardware maintenance for the medical terminal device will be added to the next maintenance schedule. If the hardware maintenance time for the medical terminal device has not yet reached or exceeded the preset maintenance interval, the hardware maintenance task for the device will be scheduled for the next maintenance cycle. This ensures that the device receives regular and systematic maintenance, thereby maintaining its normal operation and extending its service life. Specifically, first, check the time of the last hardware maintenance for the medical terminal device to confirm whether it has not exceeded the preset maintenance interval for the next maintenance. After confirming that the maintenance time has not expired, evaluate the current status of the device, including its operating status and performance indicators, to ensure that the device is in good condition before being added to the next maintenance schedule. Based on the device evaluation results and the preset maintenance interval, the hardware maintenance task for the device will be added to the next maintenance schedule. The plan should clearly define the specific content, time, and person responsible for the maintenance. The maintenance plan will be communicated to relevant maintenance personnel and management personnel to ensure that they are aware of and prepared to perform the maintenance tasks for the next cycle.

[0104] By combining preset maintenance intervals with the predicted hardware lifespan to generate hardware maintenance plans, it's possible to promptly identify whether equipment has received timely maintenance, improving the timeliness and effectiveness of equipment maintenance. Furthermore, combining preset maintenance intervals with the predicted hardware lifespan to generate hardware maintenance plans enables intelligent maintenance management, reduces human interference, and improves the accuracy and efficiency of maintenance management.

[0105] The present application also provides a machine-readable storage medium having instructions stored thereon, which are used to enable a machine to execute the above-mentioned artificial intelligence-based predictive maintenance method for medical equipment.

[0106] The present application also provides a server, comprising: a memory configured to store instructions; and The processor is configured to call instructions from the memory and implement the above-mentioned artificial intelligence-based predictive maintenance method for medical equipment when executing the instructions.

[0107] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Thus, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0109] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0111] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0112] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0113] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0114] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0115] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A predictive maintenance method for medical equipment based on artificial intelligence, characterized in that: The method comprises the following steps: Receive multiple maintenance instructions for multiple medical terminal devices, and confirm the long connection status of each medical terminal device through the server and using the heartbeat detection mechanism; If the medical terminal device is in a long connection state, the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers are combined, and the load balancing technology is used to distribute the maintenance instructions to the servers; Determine the maintenance type of each medical terminal device according to the maintenance instruction, and select the corresponding maintenance method based on the maintenance type to maintain the medical terminal device; Determine whether the current maintenance instruction is completed; If the maintenance instruction has been completed, the medical terminal equipment maintenance plan is output and the medical equipment maintenance plan is visually displayed.

2. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 1, characterized in that: The server uses a heartbeat detection mechanism to confirm the long connection status of each medical terminal device, including: Receive the heartbeat packet sent by the medical terminal device at a predetermined time, and forward the heartbeat packet to the server, so that the server feeds back a response signal after receiving the heartbeat packet; Feedback the response signal to the medical terminal device, and determine whether the medical terminal device receives the response signal within a preset time period; If the medical terminal device does not receive a response signal within a preset time period, the medical terminal device is determined to be in an offline state; If the medical terminal device receives a response signal within a preset time period, it is confirmed that the medical terminal device is in a long connection state.

3. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 1, characterized in that: Combining the maintenance indicator data corresponding to all maintenance instructions and the load indicator data of all servers, and using load balancing technology to distribute maintenance instructions to servers includes: Construct a valued bipartite graph based on maintenance index data and load index data; Use the Kuhn-Munkres algorithm to solve the optimal matching of the weighted bipartite graph and output the optimal matching result; The maintenance instructions are assigned to the server according to the best matching result.

4. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 3, characterized in that: Constructing a valued bipartite graph based on maintenance index data and load index data includes: Calculate the load rate of each server processing the maintenance instruction based on the load indicator data; Calculate the standardized task volume and standardized expected completion time based on maintenance indicator data; A bipartite graph is constructed based on the maintenance instructions and the servers, wherein identifiers of the plurality of maintenance instructions are used as a left vertex set of the bipartite graph, and identifiers of the plurality of servers are used as a right vertex set of the bipartite graph; For any maintenance instruction and server, the actual execution time of the maintenance instruction is determined using the standardized task volume and load rate; If the actual execution time is less than or equal to the normalized expected completion time, then an edge of the bipartite graph is established; If the actual execution time is greater than the normalized expected completion time, the edge of the bipartite graph is not established; Based on the standardized task volume and load rate, the matching degree between the maintenance instruction and the server is calculated using the matching degree algorithm; The matching degree is used as the weight value of the edge connecting the maintenance instruction and the server in the bipartite graph to obtain the assigned bipartite graph.

5. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 1, characterized in that: Maintenance types include software maintenance and hardware maintenance. Based on the maintenance type, select the corresponding maintenance method to maintain medical terminal equipment, including: When the maintenance type is software maintenance, extracting version information of the software application in the medical terminal device from a preset database; Filter the updated device list based on version information and device type of medical terminal devices; Use the pre-update backup mechanism to back up the software data of the software applications in the update device list; Use breakpoint resume technology to transmit update data to the corresponding medical terminal device in the update device list and update the software application; After the update is completed, the software application functions are automatically verified using the post-update verification mechanism; If the verification fails, the rollback mechanism is automatically triggered to restore the software data to the backup state; If the verification is successful, it is determined that the software application is updated successfully.

6. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 5, characterized in that: The method further comprises: When the maintenance type is hardware maintenance, collect historical operating status data and real-time working condition data of medical terminal equipment; Based on the dual-tree complex wavelet transform and nonlinear time series projection method, the historical operation status data is denoised and the time series characteristics and frequency domain characteristics of the historical operation status data are extracted; Fuse the time series features and frequency domain features to obtain a feature parameter set; Perform kernel principal component analysis on the feature parameter set, select the kernel principal component with a cumulative contribution rate higher than the preset threshold and map the corresponding feature vector to the kernel principal component space to obtain the test kernel principal component; The test kernel principal component was used as a covariate to construct a Weibull proportional failure rate model; Input the real-time working condition data into the Weibull proportional failure rate model to calculate the failure rate of the medical terminal equipment; Integrate the failure rate to calculate the reliability and draw the reliability curve of the medical terminal equipment; Use the feature parameter set and reliability curve to train the preset LSTM prediction model; Input real-time operating condition data into the trained LSTM prediction model, and use the LSTM prediction model to output the predicted hardware life of the medical terminal equipment; Generate a hardware maintenance plan by combining the preset fault maintenance interval and the predicted hardware life.

7. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 6, characterized in that: Based on the dual-tree complex wavelet transform and nonlinear time series projection method, the historical operation status data is denoised, and the time series characteristics and frequency domain characteristics of the historical operation status data are extracted, including: The dual-tree complex wavelet transform is used to perform preliminary noise reduction on the historical operating status data containing noisy signals to obtain the characteristic signal and noise signal of the historical operating status data; The nonlinear time series projection method is used to project the characteristic signal and the noise signal into different spaces to reconstruct the historical operation status data and obtain the denoised historical operation status data. The time series and frequency domain features of the historical operating status data after noise reduction are extracted through the wavelet transform algorithm.

8. The artificial intelligence-based predictive maintenance method for medical equipment according to claim 6, characterized in that: Combining the preset failure maintenance interval and hardware predicted life to generate a hardware maintenance plan includes: Determine whether the hardware maintenance time of the medical terminal equipment in the previous cycle exceeds the preset fault maintenance interval; If the preset fault maintenance interval is exceeded, an emergency maintenance work order will be triggered; If the preset fault maintenance interval has not been exceeded, then determining whether the predicted hardware life exceeds the preset fault maintenance interval of the next cycle; If the preset fault maintenance interval for the next maintenance is exceeded, the hardware status of the medical terminal device is continuously monitored; If the preset fault maintenance interval for the next maintenance is not exceeded, the hardware maintenance of the medical terminal device will be added to the maintenance plan for the next cycle.

9. A machine-readable storage medium, characterized in that The machine-readable storage medium stores instructions for enabling a machine to execute the artificial intelligence-based predictive maintenance method for medical equipment according to any one of claims 1 to 8.

10. A server, characterized in that: include: a memory configured to store instructions; as well as A processor is configured to call the instructions from the memory and to implement the artificial intelligence-based predictive maintenance method for medical equipment according to any one of claims 1 to 8 when executing the instructions.

Images