Medical equipment terminal cooperative operation and maintenance method and system based on multi-dimensional state monitoring

Abstract

This application discloses a collaborative operation and maintenance method and system for medical device terminals based on multi-dimensional status monitoring. The method includes: upon receiving a maintenance instruction from a medical device terminal, determining the connection status of the medical device terminal; if the medical device terminal is in a long-connection state, acquiring the device operation information, service flow information, and device environment information of the medical device terminal; constructing a directed topology structure for the service flow by combining the device operation information and the service flow information; collecting interaction instruction information to construct a temporal morphological feature sequence, and constructing a health latency baseline model through the temporal morphological feature sequence; constructing an environment common-mode device group based on the directed topology structure for the service flow and the device environment information, and performing group common-mode offset verification by combining the health latency baseline model and the environment common-mode device group to determine the peripheral degradation status corresponding to the medical device terminal; and executing a degradation operation and maintenance strategy based on the peripheral degradation status. This application can effectively improve the stability of medical device terminal operation.

Description

Technical Field

[0001] This application relates to the field of medical device operation and maintenance technology, and in particular to a collaborative operation and maintenance method and system for medical device terminals based on multi-dimensional status monitoring. Background Technology

[0002] With the continuous advancement of smart healthcare systems and the ongoing improvement of hospital informatization and self-service capabilities, various medical self-service terminals have been widely deployed in business scenarios such as registration, payment, and report printing. Their operational stability and business continuity have become an important foundation for ensuring the efficiency of medical services and the patient experience.

[0003] Currently, the operation and maintenance monitoring of medical device terminals mainly focuses on macro-level status perception at the system or business level. This includes assessing the overall operational status of the terminal through methods such as online device status monitoring, network connectivity detection, system operation log analysis, and business interface response statistics, and then using preset thresholds to trigger anomaly alarms and operational responses. These methods typically judge the availability or online status of the terminal from the perspective of overall device operation, which can meet basic operation and maintenance management needs to a certain extent. However, these monitoring methods primarily remain at the macro-level status assessment, lacking precise monitoring of the micro-level hardware execution processes within the terminal. In the actual operation of medical self-service terminals, multiple peripherals such as card readers, printing modules, barcode scanning modules, and touch modules typically need to collaboratively participate in task execution according to a predetermined sequence within the same business process. Complex dependencies and execution coupling exist between these peripherals in instruction scheduling, data interaction, and driver response. When a peripheral experiences performance degradation, it can cause collaborative waiting and execution sequence misalignment among multiple peripherals, leading to a significant decrease in the actual processing efficiency of the overall business process, thereby affecting the stability of the medical self-service equipment.

[0004] There is currently no good solution to the above problems. Summary of the Invention

[0005] This application provides a collaborative operation and maintenance method and system for medical device terminals based on multi-dimensional status monitoring, which is used to improve the stability of medical device terminal operation.

[0006] To achieve the above objectives, the embodiments of this application adopt the following technical solutions: Firstly, a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring is provided, applicable to a medical device terminal system connected to multiple medical device terminals. This method includes: When a maintenance command is received from a medical device terminal, the connection status of the medical device terminal is determined, including a long connection status. If the medical device terminal is in a long-connection state, activate the bypass probe pre-deployed in the medical device terminal system and obtain the device operation information, business flow information and device environment information of the medical device terminal; Construct a directed topology for business flows by combining equipment operation information and business flow information; A time-series morphological feature sequence is constructed based on the interactive command information collected by the bypass probe, and a health delay baseline model is constructed through the time-series morphological feature sequence. Based on the directed topology of the business flow and the device environment information, an environment common mode device group is constructed. The group common mode offset is verified by combining the health latency baseline model and the environment common mode device group to determine the peripheral degradation status corresponding to the medical device terminal. The peripheral degradation status includes business load driven degradation, hardware path degradation and common mode environment degradation. Based on business load-driven degradation, hardware path degradation, and common-mode environment degradation, corresponding degradation operation and maintenance strategies are executed respectively.

[0007] In another possible implementation of the first aspect, the method further includes: When the peripheral device is in a state of degradation and shows a trend of degradation, idle peripheral devices that are not in the business call chain are selected according to the directed topology of the business flow. The bypass probe sends a silent command to the idle peripheral device without causing a change in the service status, and collects the instantaneous morphological feature sequence of the idle peripheral device. Determine the temporal degradation feature sequence of the corresponding target peripheral based on the peripheral degradation status of the medical device terminal; Under the same underlying hardware topology, the timing degradation feature sequence of the target peripheral is compared with the instantaneous morphological feature sequence of the idle peripheral to determine the cause of the degradation state of the peripheral. The peripheral degradation state is corrected based on the cause of attenuation to obtain the corrected peripheral degradation state.

[0008] In another possible implementation of the first aspect, the connection state further includes an offline state, and determining the connection state of the medical device terminal includes: The medical device terminal sends heartbeat data packets to the cloud server connected to the medical device terminal system at preset heartbeat cycles. After sending a heartbeat data packet, receive the response data packet returned by the cloud server according to the preset response time window; When a response data packet is received within the preset response time window, the initial connection status of the medical device terminal and the active identifier of the communication link are determined by the round-trip time delay parameter and validity verification parameter of the response data packet. The long-term connection status of the medical device terminal is determined by combining the initial connection status and the active identifier of the communication link; If no response data packet is received within the preset response time window, the reconnection mechanism is triggered and a heartbeat data packet is sent according to the preset retry interval. If the number of consecutive retries reaches the preset retry threshold and no response data packet is received, the medical device terminal connection is determined to be abnormal, and the medical device terminal is marked as offline using the cloud server.

[0009] In another possible implementation of the first aspect, the construction of the directed topology of the service flow by combining device operation information and service flow information includes: Extract the underlying call feature information of medical device terminal peripheral operation from the device operation information, and perform hash encoding on the underlying call feature information to generate corresponding peripheral event nodes; A discrete event node set is constructed based on peripheral event nodes, and the discrete event node set is serialized and arranged according to the corresponding timestamps to obtain the terminal concurrent action timeline; By combining business flow information and terminal concurrent action timeline, a time-series association rule mining algorithm is used to analyze the time-series dependency relationship between each peripheral event node to determine the sequential following relationship and causal relationship between each peripheral event node; Multiple implicit directed edges are generated by combining sequential following relationships and causal relationships; Aggregate implicit directed edges to construct a directed topology for business flows.

[0010] In another possible implementation of the first aspect, the method further includes: Obtain historical timeline data corresponding to concurrent actions of historical terminals based on peripheral event nodes; In the directed topology of the business flow, for peripheral event nodes on different branch paths, the overlap statistical analysis of the temporal pattern of each peripheral event node is performed based on historical timeline data according to a preset time window. When the overlap probability of any two peripheral event nodes within a preset time window is less than or equal to a preset overlap threshold, it is determined that there is a mutual exclusion relationship between the two peripheral event nodes, and a corresponding mutual exclusion mapping relationship is established. Identify active event nodes corresponding to active state business paths based on the current terminal concurrent action timeline; Based on the mutual exclusion mapping relationship between the parent and child nodes and the active event nodes, the target event nodes that have a mutual exclusion relationship with the active event nodes are identified, and the idle state of the target event nodes is determined to obtain the idle state nodes. Idle peripheral resources are determined by dynamically updating the directed topology of the service flow based on idle state nodes.

[0011] In another possible implementation of the first aspect, the construction of a time-series morphological feature sequence based on the interaction command information collected by the bypass probe, and the construction of a healthy latency baseline model through the time-series morphological feature sequence, includes: The first timestamp of the instruction issuance time and the second timestamp of the hardware receipt completion are determined based on the interactive instruction information. The peripheral response timeslots are calculated by combining the first and second timestamps, and the peripheral response timeslots are constructed into raw timing data in chronological order. Collect raw time-series data of multiple identical services from the peripheral devices corresponding to the medical device terminal, and analyze and identify time disturbance components from the raw time-series data of multiple identical services. The target time series sequence is obtained by extracting the time components of stable repeating features based on the time perturbation components; Extract the time interval between each instruction in the target time sequence and calculate the ratio between any two adjacent time intervals; A multidimensional phase feature matrix is ​​constructed by combining the ratio relationships, and the multidimensional phase feature matrix is ​​arranged in chronological order to obtain a time-series morphological feature sequence; Acquire historical operating data for all peripherals in the medical device terminal, and calculate the trend and rate of change of time-series morphological feature sequences based on the historical operating data; Determine the range of characteristic changes within the future time window by combining the trend and rate of change; A healthy latency baseline model is generated based on the range of feature changes within a future time window.

[0012] In another possible implementation of the first aspect, determining the range of characteristic changes within a future time window by combining the trend of change and the rate of change includes: The matrix offset is calculated based on the multidimensional phase feature matrix within the first sliding time window and the second sliding time window respectively, wherein the duration of the first sliding time window is greater than the duration of the second sliding time window; The degradation velocity vector and degradation acceleration vector are obtained by calculating the first and second time derivatives of the matrix offset. The degradation velocity vector is used to represent the trend of change, and the degradation acceleration vector is used to represent the rate of change. Determine the corresponding peripheral type in the medical device terminal and match the physical attributes of the peripheral type in the preset database; By combining the physical properties of the peripheral device, a corresponding degenerate inertia coefficient matrix is ​​assigned to the corresponding peripheral device, and the degenerate momentum vector is obtained by combining the degenerate inertia coefficient matrix with the degenerate velocity vector. The peripheral operating state space is constructed by combining the degradation momentum vector and the degradation acceleration vector, and the health state boundary range is determined within the peripheral operating state space by combining historical health operating data from historical operating data. Map the degenerate momentum vector and degenerate acceleration vector corresponding to the current temporal morphological feature sequence to the peripheral operating state space; In the peripheral operating state space, based on the changing trend of the degradation momentum vector and the degradation acceleration vector, the evolution trajectory of the degradation momentum vector within the future time window is predicted, and the range of characteristic changes within the future time window is determined according to the degree of deviation of the evolution trajectory from the healthy state boundary range.

[0013] In another possible implementation of the first aspect, the step of constructing an environmental common-mode device group based on the directed topology of the service flow and device environment information, and performing group common-mode offset verification by combining the health latency baseline model and the environmental common-mode device group to determine the peripheral degradation state corresponding to the medical device terminal includes: The set of peripheral event nodes for nodes in the same business stage is determined based on the directed topology of the business flow, and an environmental feature vector is constructed based on the device environment information. The set of medical device terminals is filtered by combining environmental feature vectors and the business stages corresponding to the set of peripheral event nodes. Acquire the business concurrency data of the medical device terminal set within a preset time window, and determine the load phase characteristic parameters through the business concurrency data; Clustering of the medical device terminal set using load phase characteristic parameters yields an environmental common-mode device group; The real-time temporal morphological feature sequence of each medical device terminal in the environmental common mode device group is obtained, and the common mode change features in the environmental common mode device group are extracted by combining the real-time temporal morphological feature sequence. A common-mode offset compensation matrix is ​​constructed based on the common-mode variation characteristics to reflect the temporal morphological offset relationship caused by the combined effects of environmental factors and business load. Identify the target medical device terminal in the environmental common-mode device group and obtain the current temporal morphological feature sequence of the target medical device terminal; The current time-series morphological feature sequence is compared with the common-mode offset compensation matrix to obtain individual differential-mode features, so as to determine the individual differential-mode degradation features of the equipment after removing the influence of environmental common-mode. Individual differential features are input into the healthy delay baseline model for offset verification, and the corresponding peripheral degradation state is determined based on the offset results.

[0014] Secondly, this application provides a machine-readable storage medium storing instructions that cause a machine to execute the aforementioned collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring.

[0015] Thirdly, this application provides an electronic device, comprising: The memory is configured to store instructions; and The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the aforementioned collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring.

[0016] Through the above technical solution, firstly, by employing a pre-emptive connection status discrimination mechanism, accurate filtering of maintenance trigger conditions is achieved, improving the coverage of identification entry points for non-offline latent faults, such as slow response and link blockage, enhancing the accuracy of maintenance triggers, and significantly increasing the probability of discovering resource contention and waiting accumulation issues. Secondly, a non-intrusive acquisition method using bypass probes is adopted to complete multi-dimensional data aggregation without interrupting core business flows, improving the real-time capture capability of command response latency, data transmission jitter, and drive scheduling deviations, enhancing the sensitivity to subtle latency differences, and significantly improving end-to-end monitoring coverage of bus communication resource contention and waiting accumulation processes. Thirdly, by abstracting discrete multi-source information into a topological model with directional dependencies, the visualization capability of dependencies and bottleneck nodes during multi-peripheral collaborative execution is improved, enhancing the ability to trace execution link blockages and staged obstruction propagation paths, and significantly improving the efficiency of root cause localization for tasks that cannot be closed-loop. By extracting temporal features and modeling health baselines, a dynamic benchmark reference system for device communication behavior was established. This improved the early detection sensitivity of progressive degradation such as slowed response times of some peripherals and enhanced the ability to distinguish between normal latency fluctuations and abnormal waiting accumulation, significantly improving the timeliness of early warnings triggered at the initial stage of resource contention. Subsequently, an environmental common-mode correlation and group collaborative verification mechanism was introduced to improve the accuracy of distinguishing between common-cause latency deterioration and individual anomalies in multi-peripheral collaborative scenarios, significantly improving the accuracy of identifying factors affecting the operational stability of medical self-service devices. Based on business load-driven degradation, hardware path degradation, and common-mode environment degradation, corresponding degradation operation and maintenance strategies were executed, realizing a closed-loop linkage from state awareness to strategy execution. Operation and maintenance actions were triggered differently according to the degree of degradation, improving the proactive intervention capability for slowed response, link blockage, and phased blocking issues, enhancing the guarantee level of business process processing efficiency, and significantly improving the task closure success rate and overall operational stability of medical self-service devices.

[0017] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0018] Figure 1 A flowchart illustrating a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring, provided in an embodiment of this application; Figure 2 This application provides a schematic diagram of a structure for constructing a directed topology for business flows. Figure 3 This is a schematic diagram of a medical device terminal hierarchy provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0020] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0021] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0022] Figure 1 This illustration schematically shows a flowchart of a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring, according to an embodiment of this application. Figure 1 As shown in the figure, this application provides a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring, which may include the following steps.

[0023] S110. When a maintenance command for a medical device terminal is received, determine the connection status of the medical device terminal. In this embodiment, when the medical device terminal system receives a maintenance instruction for the target medical device terminal, it first triggers a connection status judgment process based on the maintenance instruction to confirm its current communication reachability and link stability. Maintenance instructions include, but are not limited to, remote software upgrade instructions, parameter configuration update instructions, device status detection instructions, or fault diagnosis instructions. Specifically, after receiving the maintenance instruction, the cloud server sends a status detection request to the target medical device terminal to trigger the medical device terminal to execute a heartbeat data packet transmission process based on a preset heartbeat cycle. After sending the heartbeat data packet, the system receives a response data packet returned by the cloud server within a preset response time window. When a response data packet is received within the preset response time window, the initial connection status of the medical device terminal and the active identifier of the communication link are determined based on the round-trip time parameter and validity verification parameter of the response data packet. The connection status of the medical device terminal is further determined to be either a long-term connection or a short-term unstable connection based on the initial connection status and the active identifier of the communication link. If no response data packet is received within the preset response time window, a reconnection mechanism is triggered, and heartbeat data packets are continuously sent according to the preset retry interval. When the number of consecutive retries reaches the preset retry threshold and no response data packet is received, the connection status of the medical device terminal is determined to be abnormal, and the medical device terminal is marked as offline via the cloud server. After determining the connection status of the medical device terminal, the execution strategy of maintenance instructions is tiered according to the connection status. When the connection status is a long-term connection, the maintenance instruction is executed directly; when the connection status is an unstable connection, the execution of the maintenance instruction is suspended and a secondary status confirmation is performed; when the connection status is offline, the execution of the maintenance instruction is suspended, the abnormal status is recorded, and a reconnection or alarm mechanism is triggered.

[0024] S120. If the medical device terminal is in a long connection state, deploy a bypass probe in the medical device terminal system and obtain the device operation information, business flow information and device environment information of the medical device terminal. In this embodiment, the bypass probe refers to a non-invasive data acquisition module deployed in the medical device terminal system. It is a functional unit used to passively monitor and mirror the operating status and communication behavior of the medical device terminal without participating in the main business data processing and forwarding. The bypass probe collects device operating information of the medical device terminal, including but not limited to CPU utilization, memory usage, process running status, device interface call status, and storage read / write load information. Next, the bypass probe mirrors the business communication data of the medical device terminal to obtain business flow information, including business request and response data streams, interface call frequency, data packet transmission paths, and business stage transition sequences. Finally, based on the bypass probe, it obtains information about the operating environment of the medical device terminal, including but not limited to network latency and jitter information, bandwidth usage, peripheral device connection status, and system operating environment parameters.

[0025] S130. Construct a directed topology structure for business flows by combining equipment operation information and business flow information; The underlying call feature information corresponding to peripheral operations is extracted from the collected device operation information. This underlying call feature information includes system call interface identifiers, driver access paths, and peripheral read / write operation identifiers. The underlying call feature information is then hashed to generate corresponding peripheral event nodes, thereby discretizing continuous underlying call behaviors into computable node units. Further, a discrete event node set is constructed based on the peripheral event nodes, and the discrete event node set is time-sequentially arranged according to the timestamps corresponding to each event node to obtain a terminal concurrent action timeline, representing the evolution of concurrent operations of the medical device terminal within a unit of time. Subsequently, the business flow information is aligned with the terminal concurrent action timeline, and a time-series association rule mining algorithm is used to analyze the temporal sequence and statistical co-occurrence relationships between different peripheral event nodes to determine the sequential following relationships and potential causal relationships among the peripheral event nodes. Based on this, the determined sequential following relationship and causal association relationship are mapped into multiple implicit directed edges, where each implicit directed edge is used to represent the directional dependency relationship between two peripheral event nodes. Finally, the implicit directed edges are aggregated and deredundant, eliminating duplicate dependencies and weak association edges, and generating a structured business flow directed topology.

[0026] S140. Construct a time-series morphological feature sequence based on the interactive command information collected by the bypass probe, and construct a health delay baseline model through the time-series morphological feature sequence. Next, the interaction command information between the medical device terminal and peripherals is acquired through a bypass probe. Based on the interaction command information, a first timestamp corresponding to the command issuance time and a second timestamp corresponding to the hardware acknowledgment completion time are determined to represent the start and end times of a single command execution process. Then, the peripheral response time slots are calculated by combining the first and second timestamps, and multiple peripheral response time slots are arranged in chronological order to construct raw time-series data, reflecting the latency evolution sequence of the device during continuous command interaction. Subsequently, raw time-series data from multiple identical business scenarios are collected, and different data sequences are compared and analyzed to identify time disturbance components caused by network jitter, instantaneous load fluctuations, or environmental interference. Based on this, time components with stable repetitive characteristics are extracted to form a target time-series sequence, eliminating the impact of random fluctuations on the timing structure. Furthermore, the time intervals between adjacent commands in the target time-series sequence are extracted, and the ratio between any two adjacent time intervals is calculated to represent the relative evolution characteristics of latency changes. Subsequently, a multi-dimensional phase feature matrix is ​​constructed by combining ratio relationships, and the multi-dimensional phase feature matrix is ​​arranged in chronological order to obtain a time-series morphological feature sequence that represents the dynamic evolution law of the equipment's operating state. On this basis, historical operating data is introduced to perform trend analysis and change rate calculation on the time-series morphological feature sequence to determine its evolution law in the long-term operation process, and further, the range of feature changes within the future time window is predicted by combining the change trend and change rate. Finally, a health latency baseline model is generated based on the range of feature changes within the future time window to represent the standard latency behavior boundary of the medical device terminal under normal healthy operating conditions.

[0027] S150. Construct an environmental common-mode device group based on the directional topology of the business flow and the device environment information, and perform group common-mode offset verification by combining the health latency baseline model and the environmental common-mode device group to determine the peripheral degradation status corresponding to the medical device terminal. First, based on the directed topology of the business flow, a set of peripheral event nodes in the same business phase is determined, and a corresponding environmental feature vector is constructed by combining the device environment information. The environmental feature vector is used to represent the network state, load level, and operating environment parameters of the device. Next, the environmental feature vector and the business phase information corresponding to the set of peripheral event nodes are jointly filtered to determine a set of medical device terminals with the same business semantic background and similar environmental conditions. Subsequently, the business concurrency data of the set of medical device terminals within a preset time window is obtained, and load phase feature parameters are extracted based on the business concurrency data to represent the load change pattern of the device in the time dimension. On this basis, cluster analysis is performed on the set of medical device terminals in combination with the load phase feature parameters to divide them into environmental common-mode device groups, so that the devices in the same group have a high degree of consistency in business load characteristics and operating environment. Then, the real-time temporal morphological feature sequence of each medical device terminal in the environmental common-mode device group is obtained, and common-mode change features are extracted based on the temporal feature changes of multiple devices in the group to represent the overall temporal offset trend caused by the combined effect of environmental factors and business load. A common-mode offset compensation matrix is ​​constructed based on common-mode variation characteristics to determine the systematic offset pattern of the group level over time. Based on this, a target medical device terminal is selected from the environmental common-mode device group, and its current temporal morphological feature sequence is obtained. The current temporal morphological feature sequence is then differentially analyzed with the common-mode offset compensation matrix to obtain the individual differential-mode features after removing the influence of environmental common-mode, representing the independent degradation characteristics of the target device relative to the common changes in the group. Finally, the individual differential-mode features are input into a healthy latency baseline model for offset verification. The degree of deviation from the healthy baseline model is used to determine the degradation state of the corresponding peripherals of the target medical device terminal, thereby achieving the identification of the actual degradation of the device and environmental interference factors.

[0028] S160. Based on service load-driven degradation, hardware path degradation, and common-mode environment degradation, execute the corresponding degradation operation and maintenance strategies respectively.

[0029] In this embodiment, workload-driven degradation refers to the gradual performance degradation caused by the heavy workload and dense concurrent requests carried by the medical terminal, which keeps peripheral devices in a high-load state. Hardware path degradation refers to the physical layer performance degradation of a specific hardware component inside the medical terminal, such as a card reader chip, touch screen control board, print head drive circuit, or its corresponding data path, such as an I / O bus, interface controller, or DMA channel, resulting in continuous delays in the call path related to that hardware. Common-mode environment degradation refers to the phenomenon that multiple devices or components in a certain area simultaneously exhibit similar degradation characteristics due to changes in the shared physical environment factors in which the device is located, such as abnormal power supply ripple, internal temperature rise, and enhanced electromagnetic interference. Next, after determining the peripheral degradation status of the medical device terminal, degradation operation and maintenance strategies are executed according to the degradation type. When it is determined to be service load-driven degradation, a low-load collaborative device within the same security domain as the current medical device terminal is obtained through traversal of the directed topology of the service flow. A shadow communication proxy is deployed on the low-load collaborative device. The shadow communication proxy is a lightweight proxy execution module deployed on a collaborative device within the same network security domain as the target medical device terminal. It is used to receive non-core service data redirected by bypass probes and perform corresponding communication processing or computing tasks on behalf of the target medical device terminal, thereby reducing the service load of the target medical device terminal. Then, the non-core service flow of the current medical device terminal is sliced ​​and redirected through bypass probes, and the non-core service flow is offloaded to the shadow communication proxy for execution.

[0030] When hardware path degradation is detected, the dynamic latency offset between the current timing characteristics and the healthy latency baseline model is calculated. Based on the dynamic latency offset, time-domain pre-compensation transmission is performed on the interactive commands sent to the medical device terminal. Simultaneously, the connectivity monitoring mode of the medical device terminal is adjusted to an asymmetric passive listening mode. Specifically, when hardware path degradation is detected in the medical device terminal, the interactive command information collected by the bypass probe is analyzed to extract the current timing characteristics, which are then compared with the healthy latency baseline model to obtain the dynamic latency offset characterizing the degree of hardware path performance degradation. Based on the dynamic latency offset, the interactive commands sent to the medical device terminal are pre-compensated in the time domain. Time-domain pre-compensated transmission involves shifting the command transmission time forward according to a dynamic delay offset to offset the additional transmission delay introduced by hardware path degradation. This ensures that interactive commands arrive and are executed as expected at the target execution time. Simultaneously, to reduce the communication load on the degraded hardware path, the connectivity monitoring mechanism of the medical device terminal is adaptively adjusted. The original active interactive heartbeat monitoring mode is switched to an asymmetric passive listening mode. By reducing the frequency of active reporting from the terminal side and enhancing the listening and acquisition capabilities of the bypass probe, continuous awareness of the terminal's operating status is achieved. This maintains system communication stability and delays further hardware performance degradation without increasing the hardware burden.

[0031] When a common-mode environment degradation is detected, the periodic interference trough phase of the common-mode device group is extracted, and a staggered scheduling clock is generated based on the interference trough phase. This clock is then sent to the medical device terminals within the common-mode device group to control the medical device terminals to centrally execute interactive commands within the interference trough phase window. In other words, when a common-mode environment degradation is detected, it indicates that the current environment is periodically or continuously interfering with the latency or signal of all devices in the group. The system analyzes the latency fluctuation period of the group of devices. For example, if it discovers a similar 50Hz power frequency interference characteristic or a regular electromagnetic pulse interference group concurrent latency, it can execute a group flow staggered phase polling strategy. The medical platform, based on the trough of the interference fluctuation (i.e., the time window where the interference is weaker), forcibly and uniformly reissues a phase offset clock to all medical devices and bypass probes located in the common-mode device group. This allows all devices in the group to centrally schedule their core business data packets and command interactions to be transmitted concurrently within the interference trough phase window, avoiding interference peaks. Through cluster-level time-domain phase staggered scheduling, group interference immunity is achieved.

[0032] Figure 3 This application provides a schematic diagram of a medical device terminal hierarchy, as shown in the embodiments. Figure 3 As shown, in this architecture, the cloud server is at the top layer, providing data storage, processing, and management capabilities for the entire system. Below the cloud server is the medical device terminal system, responsible for centralized management and coordination of the medical device terminals distributed in various locations. Next, the medical device terminals act as specific service nodes, each terminal connecting to the medical device terminal system. Finally, each medical device terminal is further subdivided into multiple peripherals, such as peripheral 1, peripheral 2, and peripheral 3. These peripherals collectively constitute the terminal's hardware functions, such as card reading, printing, and scanning, to support specific medical service processes.

[0033] By employing pre-connection status discrimination, the system improves its ability to identify and cover non-offline hidden faults such as slow response and link congestion, significantly enhancing the probability of detecting resource contention and waiting accumulation issues. The use of non-intrusive bypass probes to collect multi-dimensional data without interrupting business flow improves the real-time capture capability of command latency, transmission jitter, and scheduling deviations, significantly enhancing end-to-end monitoring coverage. The construction of a directed topology model of the business flow improves the visualization capability of multi-peripheral collaborative dependencies and bottleneck nodes, significantly enhancing the tracing efficiency of execution link congestion and blocking propagation paths. Through temporal feature extraction and health baseline modeling, a dynamic benchmark reference system is established, improving the early detection sensitivity of progressive degradation and significantly enhancing the timeliness of early warnings in the initial stage of resource contention. The introduction of environmental common-mode correlation and group collaborative verification improves the accuracy of identifying common-cause latency deterioration and individual anomalies, significantly enhancing the accuracy of identifying factors affecting operational stability. Based on the peripheral device degradation state, collaborative operation and maintenance strategies are executed to achieve closed-loop linkage from state awareness to strategy execution. This improves the ability to proactively intervene in issues such as slow response and link blockage, significantly enhances business process processing efficiency and assurance level, and improves the task closure success rate and overall operational stability of medical self-service equipment.

[0034] In one embodiment of this invention, the method further includes: S210. When the peripheral device degradation status is that there is a degradation trend, select idle peripheral devices that are not in the business call link according to the business flow directed topology. S220. Send a silent command to the idle peripheral device without causing a change in service status through the bypass probe, and collect the instantaneous morphological feature sequence corresponding to the idle peripheral device. S230. Determine the temporal degradation feature sequence of the target peripheral based on the degradation state of the peripheral corresponding to the medical device terminal; S240. Under the same underlying hardware topology, perform topological cross-comparison between the timing degradation feature sequence of the target peripheral and the instantaneous morphological feature sequence of the idle peripheral to determine the cause of the degradation state of the peripheral. S250. Based on the attenuation cause, the peripheral degradation state is corrected to obtain the corrected peripheral degradation state.

[0035] When the target peripheral is determined to be in a state of degradation trend based on the aforementioned degradation judgment results, i.e., its temporal morphological characteristics have continuously deviated from the healthy latency baseline model but have not yet reached the fault threshold, the system's internal and external peripheral resources are screened based on the business flow directed topology to identify idle peripherals that can be used for collaborative operation and maintenance. In this embodiment, in the medical device terminal system, peripheral units that are the same or equivalent to the target peripheral in terms of functional type or business processing capability and are in the same business stage or have an association relationship in the business flow directed topology, but are not currently participating in the business call link execution and are in an unoccupied or low-load state, are selected. The target peripheral can be any specific hardware module participating in the medical business process, such as a card reader, barcode scanner, printer, camera, or communication module. First, the set of peripheral event nodes corresponding to the current business execution path is extracted from the business flow directed topology to determine the set of active peripherals in the business call link. Subsequently, peripheral nodes that are the same or equivalent to the target peripheral in terms of functional type and business stage are screened from the full set of peripherals in the medical device terminal system to construct a candidate peripheral set. Next, the candidate peripheral set is compared with the active peripheral set, and peripheral nodes currently participating in the business call chain are removed, thus obtaining a subset of peripherals not in the business call chain. Finally, the peripheral subset is determined as the set of related peripherals that are associated with the target peripheral and are currently in an idle state.

[0036] In this embodiment, a silent instruction that does not cause a change in business state refers to a lightweight test instruction constructed for the target peripheral. This instruction does not trigger actual business processes during execution, does not generate user-perceptible business output results, does not write to or modify persistent data in the business database, and does not change the current configuration state of the device or the system's operating state. It is only used to trigger the peripheral to complete a controlled minimum functional response to obtain its time response characteristics under the current operating environment. Silent instructions can be implemented through simulated execution, no-operation requests, or self-test triggers. For example, for printing peripherals, it could be a virtual printing request without paper output or a buffer write test instruction; for barcode scanning peripherals, it could be an internal identification process trigger instruction; and for card reading peripherals, it could be a no-read request or an interface connectivity detection instruction. Specifically, based on the aforementioned screening results, a set of idle peripherals is determined, and then the corresponding silent instruction is determined according to the peripheral type. A bypass probe injects a silent command into the interface layer or communication link of a target idle peripheral device. This injection process bypasses the main service scheduling process, thus avoiding interference with normal service execution. During the peripheral device's execution of the silent command, the bypass probe simultaneously collects key timestamp information such as the command issuance time, peripheral device response start time, and execution completion time, and records the phased responses during command execution. Based on the collected timestamp data, the response time intervals for each phase and their changing relationships are calculated, and a corresponding time interval sequence and ratio relationship are constructed according to chronological order, thereby forming a morphological feature sequence reflecting the current instantaneous response characteristics of the peripheral device.

[0037] Subsequently, based on the aforementioned constructed temporal morphological feature sequence and health latency baseline model, the operating data of the target peripheral device within multiple consecutive time windows are obtained, and the offset, fluctuation degree, and structural change characteristics relative to the health baseline model within each time window are calculated to form corresponding degradation feature vectors. The degradation feature vectors within different time windows are arranged and organized in chronological order to construct a degradation evolution sequence. Furthermore, by performing trend analysis on the degradation evolution sequence, the feature change patterns reflecting the response latency growth trend, fluctuation amplitude changes, and phase structure shifts are extracted, thereby forming a temporal degradation feature sequence that can represent the evolution process of the target peripheral device from a normal state to a degradation state.

[0038] Next, under a unified hardware operating environment and business topology constraints, by introducing an idle peripheral with functional equivalence as a real-time reference benchmark, a structured comparative analysis of the degradation evolution process of the target peripheral is performed, thereby achieving attribution identification of the degradation source. In this embodiment, the temporal degradation feature sequence of the target peripheral is topologically cross-referenced with the instantaneous morphological feature sequence of the idle peripheral to determine the cause of the peripheral degradation state, including the following steps: Under the same underlying hardware topology, determine the connection relationships and signal dependency paths between all peripherals; By combining the connection relationships and signal-dependent paths between peripherals, topological homology mapping is performed on the target peripheral and idle peripherals to determine the set of shared nodes and the set of different nodes. Based on the shared node set, the temporal degradation feature sequence of the target peripheral is mapped into a topological path sequence vector by slicing it into time windows, and the instantaneous morphological feature sequence of the idle peripheral is mapped into a topological state snapshot vector at the corresponding time. Construct a topology alignment matrix using topology state snapshot vectors and topology path sequence vectors; The node-level alignment calibration process is performed on the corresponding nodes in the topology alignment matrix to obtain the alignment result. Based on the alignment result, the temporal degradation offset and instantaneous morphological fluctuation between the target peripheral and the idle peripheral are calculated to obtain the topology cross difference matrix. The degradation factor decomposition of the topology cross difference matrix yields the service load driving factor, hardware path attenuation factor, and common mode environment disturbance factor. Based on the corresponding weight contribution relationship, the attenuation cause of peripheral degradation state is determined.

[0039] Within the hardware architecture of the medical device terminal, the hardware configuration information of the terminal is first obtained, including the motherboard interface distribution, bus structure, and access methods of each peripheral. Then, by parsing system configuration files, device driver loading information, or hardware description information, the connection relationships between each peripheral and the main control unit are identified, and it is further determined whether peripherals are indirectly connected through an intermediate control module or a shared bus. Furthermore, by combining the calling order and data interaction relationships of peripherals in actual business operations, the signal dependency paths between peripherals are analyzed. For example, whether the output of one peripheral serves as the input trigger condition for another peripheral, or whether there are serial calls, parallel collaboration, or cascading processing relationships.

[0040] Next, within the constructed underlying hardware topology, the target peripheral and the selected idle peripherals are structurally aligned and mapped along their respective connection links and signal dependency paths. The overlapping and differing portions of their topology paths are compared to distinguish between commonly affected link nodes and independent nodes affecting only a single peripheral. Based on the aforementioned determined connection relationships and signal dependency paths, the complete topology paths corresponding to the target peripheral and the idle peripherals are extracted. These topology paths include intermediate nodes at various levels from the main control unit to the peripheral, such as bus interfaces, control modules, and driver layer nodes. Subsequently, the two topology paths are subjected to source mapping calibration, i.e., alignment and matching are performed according to node type, interface attributes, or signal transmission order to identify nodes with the same structural location or functional attributes in both paths. Furthermore, nodes that exist in both paths and are successfully matched are defined as a shared node set, representing the hardware or signal link portions commonly depended upon by both types of peripherals. Meanwhile, nodes that exist only in one path or do not match in structural properties are defined as a set of differential nodes, used to represent their own independent hardware paths or exclusive dependencies. In some implementations, nodes can also be weighted to reflect their importance in signal transmission.

[0041] Based on the completed topological homogeneity mapping and the obtained shared node set, the temporal characteristics of different peripherals are uniformly projected into the same topological node space, and a comparable structured vector representation is constructed according to the time dimension to facilitate cross-peripheral topological consistency analysis. Specifically, based on the shared node set, the common path nodes of the target peripheral and idle peripherals in the underlying hardware topology are determined, and the temporal degradation feature sequence of the target peripheral is sliced ​​according to a preset time window, extracting the feature data within each time window. Subsequently, the degradation features within each time window are mapped according to the order of the shared nodes, and the feature values ​​of the corresponding nodes are filled into a vector of a unified dimension to form a sequence vector consistent with the topological path. That is, each time window corresponds to a topological path sequence vector, which is used to represent the degradation state distribution of the target peripheral along the shared topological path within that time period. For idle peripherals, their instantaneous morphological feature sequence is obtained at the corresponding time point, and similarly mapped according to the shared node set. The feature values ​​of each node at that moment are filled into a vector of the same structure to construct a topological state snapshot vector, which is used to represent the instantaneous state on the shared topological path at that moment.

[0042] Subsequently, using the topology path sequence vector as one dimension of the matrix (e.g., row dimension) and the topology state snapshot vector as another dimension (e.g., column dimension), a dimension-wise alignment calculation is performed on any pair of vectors. This can be achieved through methods such as difference calculation, similarity calculation, or weighted matching to obtain the matching degree at the corresponding nodes. The matching results between each pair of vectors are then filled into the corresponding positions in the matrix to form a topology alignment matrix. The matrix elements represent the degree of consistency or difference in the topology state between the target peripheral and the idle peripheral at a certain moment within a certain time window. Through this method, cross-peripheral and cross-time topology feature alignment modeling is achieved, providing a basis for subsequent degradation cause determination.

[0043] Next, for each pair of corresponding nodes in the topology alignment matrix—that is, the mapping position of the same shared node in the target peripheral and the idle peripheral—node-level alignment calibration is performed. This calibration may include operations such as time offset correction, amplitude normalization, and scale unification. Subsequently, based on the calibration, the difference between the temporal degradation characteristics of the target peripheral and the corresponding characteristics of the idle peripheral is calculated for each node in the time dimension, yielding the temporal degradation offset, which represents the degree of deviation of the target peripheral from the reference peripheral in the long-term trend. At the same time point, the characteristic differences between the two are calculated to obtain the instantaneous morphological fluctuation, which represents short-term fluctuations or transient anomalies. Finally, the temporal degradation offset and instantaneous morphological fluctuation corresponding to each node are organized and filled according to the topology to construct a topological cross-difference matrix. Each element of this matrix comprehensively reflects the degree of difference between the target peripheral and the idle peripheral at different nodes and time scales.

[0044] Subsequently, factor decomposition is performed on the aforementioned topological cross-difference matrix. For example, matrix decomposition or eigenvalue decomposition methods are used to represent the original difference matrix as a superposition of multiple sub-factors with independent physical meanings, where each sub-factor corresponds to an influence component from a different source. Combining the distribution characteristics and temporal variation features of each factor in the topology, semantic mapping and classification are performed. For example, factors that appear synchronously during high-concurrency business phases and change with load are classified as business load driving factors; factors that persist at specific hardware path nodes and have a stable attenuation trend are classified as hardware path attenuation factors; and factors that exhibit consistent fluctuations across multiple devices and nodes are classified as common-mode environmental disturbance factors. After factor classification, the contribution of each type of factor is quantified, for example, by calculating the energy proportion, variance contribution rate, or weighting coefficient of each factor in the overall difference matrix to obtain the corresponding weight contribution relationship. Finally, based on the weight of each factor and its proportion, the dominant attenuation cause of peripheral degradation is determined. For example, when a certain type of factor has the highest proportion, it is identified as the main source of degradation. In some implementations, composite degradation caused by the superposition of multiple factors can also be identified. By employing the above methods, we can achieve interpretable decomposition and attribution analysis of complex degradation phenomena, thereby providing a basis for the formulation of subsequent refined operation and maintenance strategies.

[0045] Finally, based on the aforementioned topology cross-comparison results, the attenuation cause category of the peripheral degradation state is determined. When the attenuation cause is environmental common-mode factors, an environmental common-mode compensation parameter is introduced. Based on the temporal morphological characteristic sequence of each device in the environmental common-mode device group, the environmental reference offset is calculated, including: calculating the average instantaneous response of idle peripherals within the same time window; calculating the offset difference of the target peripheral relative to this average; and performing a moving average processing on the difference sequence of multiple time windows to obtain the environmental common-mode offset function, the specific expression of which is shown below:

[0046] Among them, E of Δ(t) represents the environmental common-mode offset function; W represents the width of the moving average window, which can be set according to actual requirements. The larger the window, the smoother the extracted offset curve, but the more delayed the response to real changes; t represents the discrete-time index of the current calculation time. The offset at time t is determined by the difference between t and the previous W-1 times; k represents the number of lag steps within the window, ranging from 0 to W-1. k=0 represents the difference Δ(t) at the current time, k=1 represents the difference Δ(t-1) at the previous time, and so on, used to traverse the historical differences within the moving average window.

[0047] Based on this, the environmental common-mode offset function is used as the environmental common-mode compensation parameter to subtract or proportionally correct the timing degradation feature sequence of the target peripheral. That is, the environmental offset component is subtracted from the original degradation features or the environmental influence term is scaled proportionally, thereby achieving the isolation of environmental factors such as network latency fluctuations and system load fluctuations. When the attenuation is caused by the degradation of the device's own performance, the trend change of the original degradation features remains unchanged, and only local abnormal fluctuations are smoothed and filtered to enhance the stable expression of the degradation trend.

[0048] By correcting the degradation status of peripherals, the accuracy and anti-interference ability of peripheral degradation status assessment are improved, the level of differentiation between common-mode noise and individual degradation is enhanced, and over-maintenance or missed detection caused by misjudgment due to environmental fluctuations or system scheduling is avoided. This can significantly improve the confidence of peripheral status diagnosis of medical device terminals and the pertinence of collaborative maintenance strategies, thereby ensuring the long-term stability and business processing efficiency of medical self-service equipment.

[0049] In one embodiment of this invention, the connection status includes a long-connection status and an offline status. Determining the connection status of the medical device terminal includes: S310. Send heartbeat data packets to the cloud server connected to the medical device terminal system at preset heartbeat cycles through the medical device terminal. S320. After sending the heartbeat data packet, receive the response data packet returned by the cloud server according to the preset response time window; S330. When a response data packet is received within the preset response time window, the initial connection status of the medical device terminal and the active identifier of the communication link are determined by the round-trip delay parameter and validity verification parameter of the response data packet. S340. Determine the long connection status of the medical device terminal by combining the initial connection status and the active identifier of the communication link. S350. If no response data packet is received within the preset response time window, the reconnection mechanism is triggered and a heartbeat data packet is sent according to the preset retry interval. S360. When the number of consecutive retries reaches the preset retry threshold and no response data packet is received, the medical device terminal connection is determined to be abnormal, and the medical device terminal is marked as offline using the cloud server.

[0050] In this embodiment, a stable communication link is first established between the medical device terminal and the cloud server. This communication link can be a long-lived connection communication channel. Then, a heartbeat cycle parameter is configured within the medical device terminal to limit the time interval for sending heartbeat data packets. Furthermore, a timed scheduling mechanism is set up in the medical device terminal to automatically trigger the generation and transmission of heartbeat data packets when each preset heartbeat cycle arrives. The heartbeat data packet includes the medical device terminal's identification information, a current timestamp, and connection status identification information, indicating the current operating status and communication reachability of the medical device terminal. Finally, the medical device terminal sends the heartbeat data packet to the cloud server through the established communication link, enabling the cloud server to continuously monitor the online status and connection stability of the medical device terminal based on the received heartbeat data packet.

[0051] After the medical device terminal sends the heartbeat data packet, it immediately initiates a preset response time window timing mechanism. This response time window limits the maximum allowable reception time range for the cloud server's response data packet. Within this time window, the medical device terminal continuously listens for the response signal corresponding to the heartbeat data packet and determines whether it has received the response data packet returned by the cloud server. The response data packet is feedback data generated by the cloud server in acknowledgment of the heartbeat data packet, indicating that the cloud server has successfully received and parsed the heartbeat data packet. When a response data packet is successfully received within the preset response time window, the current communication link is considered reachable, and the corresponding reception time information is recorded. If no response data packet is received within the preset response time window, it is determined that there is a response delay or communication anomaly in the current communication link, triggering subsequent retry or anomaly detection procedures. The preset response time window can be set based on the historical round-trip time (RTT) statistics between the medical device terminal and the cloud server, for example, taking a multiple of the historical average RTT. The specific data can be set according to the enterprise's accuracy requirements.

[0052] When the medical device terminal receives the response data packet from the cloud server within a preset response time window, it extracts the round-trip time (RTT) parameter from the RTT. The RTT parameter represents the time delay experienced by the heartbeat data packet from being sent to the cloud server and returning the response data packet. By comparing this RTT parameter with a preset standard delay threshold, it determines whether the transmission latency level of the current communication link is within the normal range. The preset standard delay threshold can be set according to the sensitivity of medical services to response speed. For example, a lower threshold can be used for high real-time services such as emergency care and payment; a medium threshold can be used for general monitoring services; and a higher threshold can be used for background synchronization services. The specific value can be set according to actual conditions. Subsequently, a validity verification parameter is extracted from the RTT. This validity verification parameter verifies whether the RTT has been corrupted, lost, or tampered with during transmission. Verification methods can include checksum comparison, digest consistency verification, or sequence number matching verification. Furthermore, when the round-trip time (RTT) is within the preset normal range and the validity check passes, the initial connection state of the medical device terminal is determined to be a stable connection state, and the communication link is marked as active. The preset normal range can be determined by statistically analyzing the RTT of historical heartbeat communication between the medical device terminal and the cloud server, such as calculating the average historical RTT as a threshold. When the RTT deviates from the normal range but the validity check passes, it is determined to be a delayed connection state, indicating that the link has performance fluctuations but remains reachable. When the validity check fails, it is determined that there is abnormal interference or data transmission error in the communication link, and the link active status is adjusted to an unreliable state.

[0053] The connection status of the medical device terminal is determined by combining the initial connection status and the activity identifier of the communication link. In this embodiment, the initial connection status includes an available status, a delayed status, and a no-response status. The available status is when a response is successfully received within a specified time window, and the RTT is less than a preset latency threshold. The delayed status is when a response is received, but the RTT is greater than the threshold, indicating high latency in the link. The no-response status is when no response is received within the time window. The preset latency threshold can be set in the laboratory based on historical data, and the specific value can be set according to the enterprise's requirements. The activity identifier of the communication link is determined by the validity verification parameters carried by the heartbeat response packet. The activity identifier of the communication link includes active, dead, and invalid. Active means that the response packet passes the integrity verification, and the session identifier, sequence number, etc., carried are continuously updated, proving that the communication processes at both ends of the link are in normal working condition. Dead means that the integrity verification is passed, but the session identifier has expired, the sequence number is stagnant, or there is only heartbeat traffic for a long time without business data interaction, indicating that the link lacks activity. Invalid means that the integrity verification is not passed, the session identifier does not match, etc., indicating that the response may be returned by an incorrect intermediate device, or the connection context has been lost. By combining the above two dimensions, the final long-term connection status is determined. For example, if the initial connection status is available and the communication link activity indicator is active, it is determined to be a long-term connection, indicating that the network is smooth, application layer interaction is active and normal, and it is an ideal state for business operations and target monitoring. At this time, it is most suitable to activate the bypass probe for micro-data collection. If the initial connection status is unresponsive and the communication link activity indicator is absent, it is a connection pending confirmation, indicating a transitional state before triggering the reconnection mechanism, and offline is not directly determined at this time.

[0054] After sending a heartbeat data packet, the medical device terminal starts a response time window timer. If no corresponding response data packet is received by the timer expires, the connection management module first triggers a reconnection mechanism to re-establish or restore the current communication link. Subsequently, the system controls the retransmission of the heartbeat data packet according to a preset retry interval parameter. This preset retry interval limits the time interval between consecutive retries to avoid network congestion caused by high-frequency transmissions in a short period; the specific value can be set according to enterprise requirements. During each retry, the medical device terminal regenerates or reuses the heartbeat data packet and sends it to the cloud server again, while simultaneously restarting the response time window to wait for a response. This process can continue until the preset maximum number of retries is reached or a response data packet is successfully received. If a response data packet is successfully received during the retry, the communication link is considered to have returned to normal, and the reconnection mechanism is exited. If the number of consecutive retries reaches a preset threshold and no response data packet is received, the communication link is further determined to be abnormal or the device is offline.

[0055] After entering the reconnection mechanism, the medical device terminal periodically sends heartbeat data packets according to a preset retry interval, and restarts the response time window after each transmission to wait for a response, while accumulating the number of retries. When the cumulative number of retries reaches a preset retry threshold, and no response data packet is received from the cloud server during all retries, a connection anomaly judgment logic is triggered. The medical device terminal or the cloud server jointly confirms that the current communication link has been continuously failed. Subsequently, the cloud server updates the status information of the medical device terminal based on the judgment result, marking it as offline in the device status management list or operation monitoring database, and synchronously updating its connection identifier and availability status fields. In some implementations, the cloud server can also trigger an alarm mechanism or a log recording mechanism for subsequent operation and maintenance analysis and fault location. The preset retry interval can be estimated based on historical round-trip time (RTT), such as taking the average RTT; the preset retry threshold is used to limit the maximum number of consecutive retries and can be used to statistically analyze historical device communication recovery, such as the average number of retries required for recovery.

[0056] By judging the connection status of medical device terminals, the fault tolerance and anti-interruption ability of connection status determination are improved, avoiding misjudgments caused by network interruptions or response delays. The robustness and availability of the communication link between medical device terminals and cloud servers are enhanced, which can significantly improve the accuracy of remote operation and maintenance triggering of medical device terminals and the reliability of connection status monitoring.

[0057] In one embodiment of this invention, a directed topology for the service flow is constructed by combining device operation information and service flow information, including: S410. Extract the underlying call feature information of peripheral operation from the device operation information, and perform hash encoding on the underlying call feature information to generate the corresponding peripheral event node. S420. Construct a set of discrete event nodes based on peripheral event nodes, and serialize and arrange the set of discrete event nodes according to the corresponding timestamps to obtain the terminal concurrent action timeline; S430. Combining business flow information and terminal concurrent action timeline, the temporal dependency relationship between each peripheral event node is analyzed using a temporal association rule mining algorithm to determine the sequential following relationship and causal relationship between each peripheral event node. S440, Combining sequential following relationships and causal associations to generate multiple implicit directed edges; S450. Aggregate implicit directed edges to construct a directed topology structure for business flows.

[0058] Figure 2This diagram illustrates the directed topology of the service flow provided in this application, showcasing the call dependencies of local services in a medical device terminal system. The diagram includes service triggering sources, upstream service nodes, peripheral event nodes, intermediate aggregation nodes, and SDK event nodes, demonstrating the complete service chain from patient operation to responses from various peripherals. The directed arrows in the diagram correspond to the implicit directed edges of this application, accurately reflecting the sequential following relationships and causal connections between nodes. This topology provides the underlying structural basis for the subsequent construction of common-mode device groups and the identification of idle peripherals.

[0059] The system retrieves peripheral operation-related data sources from device operation information. These data sources include system call logs, driver layer interface call records, interrupt trigger records, and peripheral read / write operation records. The data sources are then parsed to extract underlying call feature information related to peripheral interaction. This underlying call feature information includes, but is not limited to, peripheral interface identifiers, called function names, operation type identifiers, access address information, and call timestamps, representing the specific execution behavior of the peripheral at the system level. Further, the extracted underlying call feature information is hashed according to a preset encoding rule, mapping the multi-dimensional feature information to a unique identifier value. This achieves unique identification and compressed representation of different peripheral operation behaviors. The hash encoding can be based on MD5, SHA-like algorithms, or a custom lightweight hash function. Finally, the hash-encoded result is defined as a peripheral event node, used as a basic analysis unit in the directed topology of the business flow.

[0060] Next, peripheral event nodes from different sources are aggregated to construct a discrete event node set. Each peripheral event node corresponds to an independent underlying peripheral call and carries unique identification information and a corresponding timestamp. Then, the nodes in the discrete event node set are sorted according to their timestamps, with earlier occurring events placed first and later events placed last, thus forming a time-ordered event sequence. During the sorting process, if multiple peripheral event nodes have the same or similar timestamps, concurrency relationships are further marked using event type identifiers or call priorities to reflect the parallel execution characteristics of multiple peripheral operations. Finally, the sorted event sequence is mapped to a terminal concurrent action timeline to represent the temporal distribution structure of multiple peripheral concurrent execution behaviors of the medical device terminal within a unit of time, providing a basic time dimension support for subsequent business flow dependency analysis and topology construction.

[0061] Subsequently, business flow information is acquired first. This information represents the flow order and call path between various business links during the execution of specific medical services by the medical device terminal. Simultaneously, a timeline of concurrent actions of the terminal is acquired, representing the arrangement and concurrent execution of various peripheral event nodes over time. The business flow information and timeline data are then aligned, using business stages or nodes as mapping benchmarks to assign peripheral event nodes to their corresponding business stage intervals. Next, a temporal association rule mining algorithm is introduced to statistically analyze the temporal sequence, frequent co-occurrence, and conditional dependencies among peripheral event nodes. For example, event combination relationships are filtered using support, confidence, or time interval constraints. Furthermore, based on the mining results, the sequential following relationship between peripheral event nodes is determined, that is, the relationship in which a certain peripheral event node occurs stably before or after another peripheral event node in time. Under the condition of combining business flow semantic constraints, the causal relationship between peripheral event nodes is deduced, that is, the relationship in which a certain peripheral operation triggers or depends on another peripheral operation in business logic or execution path; finally, a set of temporal dependencies is formed for subsequent topology construction.

[0062] Based on the results of temporal association rule mining, the sequential following relationships between peripheral event nodes are obtained. The sequential following relationship is used to represent the time series constraint that a certain peripheral event node occurs before or after another peripheral event node in time. At the same time, combined with the causal relationship extracted from the business flow information, the causal relationship is used to represent the triggering, dependency or support relationship of different peripheral event nodes in the business execution logic. Subsequently, a fusion analysis of sequential following relationships and causal relationships is performed. Under the conditions of satisfying time consistency and business semantic consistency, node pairs that simultaneously satisfy sequential constraints and causal constraints are mapped. Each node pair that satisfies the constraints is abstracted into a directed connection relationship, where the starting node corresponds to the preceding or triggering event node, and the ending node corresponds to the subsequent or dependent event node. This directed connection relationship is defined as an implicit directed edge. In this way, the implicit dependencies that were originally scattered in the time series and business semantics are transformed into a structured set of graph connections. In this embodiment, the implicit directed edges do not originate from a pre-defined interface call chain or hard-coded business process, but are based on the stable sequential following relationship of peripheral event nodes in the terminal concurrent action timeline, and the causal relationship determined by combining business flow information.

[0063] All implicit directed edges generated by sequential following relationships and causal associations are collected to form an initial edge set. This initial edge set undergoes aggregation processing, which includes merging multiple directed edges with the same start and end nodes, and accumulating or weighting their association strength or confidence to obtain uniform edge weight parameters. Directed edges with low frequency of occurrence or confidence below a preset threshold are filtered to eliminate weak associations caused by occasional occurrences or noise interference. Furthermore, multi-hop connection paths between nodes are merged to extract stable backbone path structures to represent core business flow relationships. After completing edge aggregation and filtering, all peripheral event nodes are used as the node set in the graph structure, and the aggregated implicit directed edges are used as the connection relationship set to construct a directed graph structure. Finally, a directed topology structure for the business flow is obtained.

[0064] By constructing a directed topology for business flows, we can significantly improve the efficiency of tracing complex anomaly propagation paths such as execution link blockages and phased obstructions, as well as the accuracy of root cause localization. This provides a structured analysis foundation for the subsequent construction of environmental common-mode groups and the precise implementation of collaborative operation and maintenance strategies.

[0065] In one embodiment of this invention, the method further includes: S510, Obtain historical timeline data corresponding to concurrent actions of historical terminals; S520. In the directed topology of the business flow, for peripheral event nodes on different branch paths, based on historical time axis data, the overlap statistical analysis of the temporal pattern of each peripheral event node is performed according to a preset time window. S530. When the overlap probability of any two peripheral event nodes within a preset time window is less than or equal to a preset overlap threshold, it is determined that there is a mutual exclusion relationship between the two peripheral event nodes, and a corresponding mutual exclusion mapping relationship is established. S540: Identify the active event nodes corresponding to the active state service paths based on the current terminal concurrent action timeline; S550. Based on the mutual exclusion mapping relationship between the parent and child nodes and the active event nodes, determine the target event node that has a mutual exclusion relationship with the active event node, and determine the idle state of the target event node to obtain the idle state node. S560: Based on the idle state nodes, dynamically update the directed topology of the service flow to determine idle peripheral resources.

[0066] Interactive data from the device operation logs are obtained from the historical operation records of the medical device terminal, and peripheral event nodes and their timestamp information corresponding to various peripheral operations are extracted. Then, based on the timestamps, the historical peripheral event nodes are sorted, arranging events according to their occurrence time, and marking the concurrency relationship of multiple events occurring within the same or similar time windows. According to business identifiers or time window divisions, peripheral event nodes within a continuous time range are combined into several historical concurrent action segments, and the event sequence and concurrency structure within each segment are recorded. Based on this, multiple historical time segments are spliced ​​and organized according to the chronological evolution order to construct a complete historical terminal concurrent action timeline data. This historical timeline data is used to represent the temporal distribution and concurrent execution mode of various peripheral operations during the historical operation of the medical device terminal.

[0067] Next, peripheral event node sets on different branch paths are identified in the directed topology of the business flow, and they are classified according to business stage or node type. Then, historical terminal concurrent action timeline data is acquired, and the timeline is segmented according to preset time windows, such as fixed-length time windows or sliding time windows. Within each time window, the trigger count, occurrence time location, and duration range of each peripheral event node are statistically analyzed. Furthermore, the occurrence of peripheral event nodes from different branch paths within the same time window is overlaid and analyzed to calculate their temporal overlap, for example, by statistically analyzing the frequency of overlap, the proportion of overlap duration, or the probability of overlap, to quantify the temporal overlap relationship between different nodes. This statistical process is repeated across multiple time windows to form a cross-time dimension overlap distribution characteristic, thereby identifying peripheral event node combinations with stable concurrency or temporal coupling relationships during historical operation.

[0068] When the overlap probability of any two peripheral event nodes within a preset time window is less than or equal to a preset overlap threshold, a mutual exclusion relationship is determined between the two peripheral event nodes, and a corresponding collateral mutual exclusion mapping relationship is established. This involves segmenting the timeline based on historical terminal concurrent action timeline data, processing the timeline according to preset time windows, and recording the occurrence of each peripheral event node within each time window. Then, for any two peripheral event nodes, the number of times they co-occur within multiple time windows and their respective total occurrences are counted, and the time overlap probability is calculated accordingly. The overlap probability can be defined as the proportion of the number of times they co-occur within the same time window to their total joint occurrences or to the total number of time windows. Further, the calculated overlap probability is compared with the preset overlap threshold. When the overlap probability is less than or equal to this threshold, it indicates that the two peripheral event nodes do not overlap in time and lack concurrent execution characteristics, thus determining that a mutual exclusion relationship exists between them. Based on this, the mutual exclusion relationship is structurally expressed, establishing a collateral mutual exclusion mapping relationship to identify that the two peripheral event nodes belong to different business branch paths and will not be called simultaneously during normal business operation. The preset overlap threshold can be calculated based on historical timeline data, the probability distribution of overlap of all peripheral event node pairs, and then the low quantile is selected as the threshold. The specific value can be set according to the actual situation.

[0069] The system identifies active event nodes corresponding to active business paths based on the current terminal concurrent action timeline. This involves mapping and aligning the current timeline with a pre-built directed business flow topology, and identifying the business path branch to which the event node belongs based on its position within the topology. Further, it filters out active peripheral event nodes by determining whether they are triggered, in execution, or continuously generating response data within the current time window. Then, based on the connectivity of active event nodes in the directed business flow topology, it determines the currently activated business path and marks it as an active business path. Finally, it outputs the active business paths and their corresponding active event node sets to reflect the current business execution status and peripheral call distribution of the medical device terminal.

[0070] Based on the mutual exclusion mapping relationship and active event nodes, target event nodes with mutual exclusion relationships with active event nodes are identified, and their idle state is determined to obtain idle state nodes. That is, based on the previously determined mutual exclusion mapping relationship, a set of candidate event nodes with mutual exclusion relationships with each active event node is retrieved. Candidate event nodes belong to different branch paths in business logic and, under normal circumstances, will not occur simultaneously with their corresponding active nodes. Further, the status of candidate event nodes is determined by checking whether they have a trigger record within the current time window, whether they appear in the current concurrent action timeline, and whether they generate a response behavior, to determine whether they are in an uncalled or unexecuted state. When a candidate event node does not generate a trigger behavior and does not participate in any business call chain within the current time window, it is determined to be an idle state node. Finally, the set of all event nodes that satisfy the mutual exclusion relationship and are in an idle state is output for subsequent bypass detection, comparative analysis, or degradation attribution processing.

[0071] Finally, based on the idle nodes, the directed topology of the business flow is dynamically updated to determine idle peripheral resources. Idle nodes are mapped to their corresponding node positions in the directed topology, and node attributes are updated with status markers, such as marking them as idle or inactive. Nodes currently in execution are marked as active. Further, node states are propagated and constrained within the topology, for example, by combining dependencies and path connectivity between nodes to update the overall active or idle status of each branch path. Based on this, a set of nodes in an idle state, not constrained by the current active path, is selected from the topology. This set of nodes is then converted into a corresponding set of peripheral resources according to the mapping relationship between nodes and peripheral resources, thus determining the currently available idle peripheral resources. These idle peripheral resources can be used for subsequent bypass detection, performance comparison analysis, or dynamic scheduling.

[0072] The above steps improve the ability to discover resource occupation conflict patterns in multi-peripheral concurrent scenarios, enhance the accuracy of inferring the idle status of peripherals in non-explicit call chains, avoid the problem of misjudgment or omission of idle resources caused by traditional static configuration methods, and can significantly improve the real-time performance and reliability of idle peripheral screening.

[0073] In one embodiment of this invention, a time-series morphological feature sequence is constructed based on interactive command information collected by a bypass probe, and a healthy latency baseline model is constructed using the time-series morphological feature sequence, including: S610, determine the first timestamp of the instruction issuance time and the second timestamp of the hardware receipt completion based on the interactive instruction information; S620. Calculate the peripheral response time slot by combining the first timestamp and the second timestamp, and construct the peripheral response time slots into the original timing data according to the time sequence. S630. Collect raw time-series data of multiple identical services, and analyze and identify time disturbance components in the raw time-series data of multiple identical services. S640. Extract the time component of stable repeating features based on the time perturbation component to obtain the target time series sequence; S650. Extract the time interval between each instruction in the target time sequence and calculate the ratio between any two adjacent time intervals. S660. Construct a multi-dimensional phase feature matrix by combining the ratio relationship, and arrange the multi-dimensional phase feature matrix in time order to obtain the time-series morphological feature sequence; S670. Obtain historical operating data and calculate the changing trend and rate of change of the time series morphological characteristic sequence based on the historical operating data; S680. Combine the trend of change and the rate of change to determine the range of characteristic changes within the future time window; S690. Generate a healthy latency baseline model based on the range of feature changes within the future time window.

[0074] The peripheral control commands issued by the medical device terminal are captured by a bypass probe or system call monitoring module. The time information corresponding to the command being sent from the application layer or driver layer to the peripheral interface is recorded and marked as the first timestamp, indicating the start time of the command issuance. Subsequently, when the peripheral completes the corresponding command processing and returns the execution result or status receipt, the receipt completion event is captured by listening to peripheral driver callbacks, interrupt signals, or data returned from the communication interface, and the time information of this event is recorded and marked as the second timestamp, indicating the end time of the peripheral response completion. The interactive command information includes command identifier, call context, and execution result identifier, used to establish a correspondence between the sending and receiving of commands, thereby ensuring a one-to-one match between the first and second timestamps.

[0075] Next, for each interactive command, the corresponding first timestamp (the time the command was issued) and second timestamp (the time the hardware acknowledgment was completed) are obtained, and the difference between the two is calculated to obtain the response time length of that peripheral operation. The response time length is defined as the peripheral response slot. Subsequently, the above calculation process is repeated for multiple interactive commands during continuous operation to obtain a set of peripheral response slot data corresponding to different time points. According to the first or second timestamp corresponding to each peripheral response slot, the slot data is sorted, with earlier slots placed at the beginning and later slots placed at the end. Finally, the sorted peripheral response slots are organized in chronological order to construct the original timing data sequence, which is used to represent the changes in the response time of the peripheral during continuous operation.

[0076] This process involves collecting raw time-series data from multiple identical business processes and analyzing this data to identify time disturbance components. Specifically, it aligns and statistically analyzes the peripheral micro-response time-series data generated during multiple executions of the same type of business, separating out unstable time components caused by random factors or changes in the external environment. First, the corresponding business process is executed multiple times under the same business type, and raw time-series data is collected during each execution. This raw time-series data consists of peripheral response time slots arranged in chronological order. Then, multiple sets of raw time-series data are aligned, using instruction sequence or business stage as a benchmark, to ensure that corresponding time slots in different execution processes have consistent semantic positions. Based on this, statistical analysis is performed on the aligned time slot positions, calculating their mean, variance, or standard deviation, etc., to represent the stability of each time slot. Further, the difference between the actual value of each time slot and the corresponding statistical mean is used as the time disturbance component. Alternatively, a smoothed sequence is obtained by processing the raw time-series data using sliding window averaging filtering, exponential weighted smoothing, etc., and the difference between the raw sequence and the smoothed sequence is extracted as the time disturbance component.

[0077] The target time series sequence is obtained by extracting stable repetitive time components based on time perturbation components. Specifically, for each time slot, the original time series data is descrambled. Descrambling can be achieved by subtracting the corresponding perturbation component from the original time slot value using one or a combination of methods to obtain the descrambled time component; or by directly selecting the mean or median of each time slot as the stable time component based on the statistical results of multiple executions. Then, the stable time components extracted from each time slot are arranged in their original time order to construct the target time series sequence. This target time series sequence represents the stable response rhythm and time structure characteristics of the peripheral device under normal operating conditions.

[0078] Next, the constructed target time series sequence is obtained. This sequence consists of multiple stable time points or stable response time slots arranged chronologically. For any two adjacent time points in the sequence, the time difference is calculated to obtain the corresponding time interval sequence. Each time interval represents the stable time span between two adjacent instructions or two adjacent response stages. Further, the ratio of any two adjacent time intervals in the time interval sequence is calculated. That is, using the previous time interval as a benchmark, the proportional relationship between the subsequent time interval and the previous time interval is calculated, thus obtaining the time interval ratio sequence. In some implementations, to avoid the influence of extreme values ​​on the results, the time intervals or ratios can be normalized or subject to threshold constraints. Finally, the time interval ratio sequence represents the relative rhythmic change characteristics between each stage in the target time series sequence. Through this method, the time series characteristics that originally depended on absolute time are transformed into relative proportional relationship characteristics, effectively reducing the impact of overall time delay offset on feature expression and providing a foundation for the subsequent construction of a time series morphological feature sequence.

[0079] Based on the aforementioned time interval ratio sequence, multiple consecutive adjacent ratios are combined according to a preset dimension. For example, several consecutive ratios are selected using a sliding window to form a feature vector. Each feature vector describes the rhythmic change characteristics within a local time segment. Subsequently, multiple feature vectors are arranged by column or row to construct a multidimensional phase feature matrix. Each row or column of the matrix corresponds to the phase feature representation of a time segment, and each dimension corresponds to the time interval ratio at different positions, used to characterize the relative time relationship between different stages. Next, the multidimensional phase feature matrix is ​​arranged according to its corresponding time order. That is, according to the chronological order of each time segment in the original time series data, the feature vectors in the matrix are sorted and organized to form a feature sequence that evolves over time. Finally, a time series morphological feature sequence is obtained, which is used to comprehensively represent the temporal structure pattern and rhythmic change law of the peripheral device during continuous operation. Through the above method, the original time interval information is transformed into a structured multidimensional feature expression, effectively enhancing the ability to characterize time series morphological changes and providing highly discriminative features for subsequent health delay baseline modeling and anomaly detection.

[0080] Multiple time-series morphological feature sequences generated during the historical operation of medical device terminals are acquired. Historical operational data can be divided into multiple consecutive time windows, each corresponding to a set of feature sequences. Subsequently, the time-series morphological feature sequences within different time windows are aligned to ensure comparability of features in the same dimension. Based on this, the values ​​of each feature dimension over time are analyzed, such as calculating the mean or representative value of the feature values ​​corresponding to each time window, thus forming a feature trajectory over time. Trend calculation is performed on the feature trajectory, using methods such as linear fitting, moving average, or regression analysis to determine its direction of change, such as identifying an upward, downward, or stable trend. Simultaneously, the feature differences between adjacent time windows are calculated, and the rate of change is determined by the ratio of the difference to the time interval, representing the speed of feature change. Smoothing or outlier filtering can also be applied to the rate of change to improve the stability of the calculation results.

[0081] To determine the range of characteristic changes within a future time window by combining the trend and rate of change, specifically, firstly, statistical analysis is performed on the multidimensional phase feature matrix within the first and second sliding time windows respectively, calculating the corresponding matrix offset. The duration of the first sliding time window is longer than that of the second sliding time window to simultaneously represent long-term trends and short-term fluctuations. Subsequently, the time derivatives of the matrix offsets are calculated to obtain the first and second time derivatives, which are used as the degradation velocity vector and degradation acceleration vector, respectively, to represent the trend and rate of change of the temporal morphological characteristics. Further, the corresponding peripheral type in the medical device terminal is determined, and the physical attributes corresponding to this peripheral type are matched in a preset database. A degradation inertia coefficient matrix is ​​assigned to the corresponding peripheral based on the physical attributes, and this degradation inertia coefficient matrix is ​​combined with the degradation velocity vector to obtain the degradation momentum vector. Based on this, a degradation momentum vector and degradation acceleration vector are used as state variables to construct an external... A state space is defined, and healthy operating data from historical operating data is used to determine the boundary range of the healthy state within the state space. Then, the degradation momentum vector and degradation acceleration vector corresponding to the current temporal morphological feature sequence are mapped to the peripheral operating state space. Finally, based on the changing trend of the degradation momentum vector and the degradation acceleration vector within the state space, the evolution trajectory of the degradation momentum vector in the future time window is predicted, and the characteristic change range in the future time window is determined according to the degree of deviation of the evolution trajectory from the boundary range of the healthy state, thereby realizing a forward-looking assessment of the peripheral performance evolution trend.

[0082] A healthy latency baseline model is generated based on the range of feature changes within a future time window. First, the range of feature changes within the future time window determined in the preceding steps is obtained. This range can be represented as the value interval of multi-dimensional phase features at future times, or its upper and lower bounds. Then, the feature change range is mapped back to the original latency space or the corresponding time interval feature space. For example, through reverse mapping or feature reconstruction, the multi-dimensional phase feature interval is converted into the corresponding peripheral response time interval or latency fluctuation range. Next, the time interval is structurally modeled, representing it as a dynamic boundary function or interval model that changes over time. Each time point corresponds to an allowable upper and lower latency limit, thus forming a healthy latency envelope over continuous time. Based on this, the dynamic boundary function or interval model is defined as the healthy latency baseline model, used to represent the normal operating range of the peripheral within the future time window. In some implementations, historical health data can also be used to smooth the baseline model or correct the confidence interval to improve the model's stability and reliability.

[0083] Through the above steps, the sensitivity to small delay differences and gradual degradation is improved, the ability to distinguish between normal fluctuations and abnormal waiting accumulation is enhanced, and the problems of missed detection and false alarm caused by the traditional single threshold judgment method are avoided. It can significantly improve the timeliness of early warning in the initial stage of resource competition and the adaptive update capability of the health baseline model.

[0084] In one embodiment of this invention, determining the range of characteristic changes within a future time window by combining the trend of change and the rate of change includes: S710. Calculate the matrix offset based on the multidimensional phase feature matrix in the first sliding time window and the second sliding time window respectively, wherein the duration of the first sliding time window is greater than the duration of the second sliding time window; S720. Calculate the first and second time derivatives of the matrix offset to obtain the degradation velocity vector and degradation acceleration vector, where the degradation velocity vector is used to represent the trend of change and the degradation acceleration vector is used to represent the rate of change. S730. Determine the corresponding peripheral type in the medical device terminal and match the peripheral physical attributes corresponding to the peripheral type in the preset database. S740. Assign a corresponding degenerate inertia coefficient matrix to the corresponding peripheral based on the physical properties of the peripheral, and use the degenerate inertia coefficient matrix and the degenerate velocity vector to perform a combined operation to obtain the degenerate momentum vector. S750: Construct the peripheral operating state space by combining the degradation momentum vector and the degradation acceleration vector, and determine the health state boundary range within the peripheral operating state space by combining historical health operating data from historical operating data. S760. Map the degradation momentum vector and degradation acceleration vector corresponding to the current temporal morphological feature sequence to the peripheral operating state space; S770. In the peripheral operating state space, based on the changing trend of the degradation momentum vector and the degradation acceleration vector, the evolution trajectory of the degradation momentum vector within the future time window is predicted, and the characteristic change range within the future time window is determined according to the degree of deviation of the evolution trajectory from the healthy state boundary range.

[0085] First, a multidimensional phase feature matrix is ​​obtained from the continuous time series, and a feature matrix sequence is constructed in chronological order. Then, two sliding time windows are set: the first sliding time window covers a longer time period to reflect the overall changes of the features over a longer time range, and the second sliding time window covers a shorter time period to reflect rapid changes or local fluctuations of the features within a short period. Within each time window, the multidimensional phase feature matrices corresponding to the start and end times of that time window are selected, or multiple matrices within the time window are averaged or weighted to obtain the feature matrix representing that time window. By performing a difference operation on the feature matrices within the time window, for example, calculating the difference between the matrix at the end time and the matrix at the start time, the corresponding matrix offset is obtained, which represents the overall change magnitude of the features within that time window. Repeating the above process for the first and second sliding time windows respectively yields the matrix offsets at long and short time scales. By comparing or jointly analyzing the two offsets, stable trend changes and short-term disturbance changes can be distinguished, thus providing a basis for subsequent calculations of change trends and rates of change.

[0086] The matrix offset sequence is calculated based on a continuous time window. The matrix offset can be represented as a multidimensional vector or matrix sequence that varies with time. Subsequently, under discrete-time conditions, the matrix offsets at adjacent time points are differentially calculated to approximate the first-order time derivative. That is, the difference between the matrix offset at the current time and the previous time is calculated and divided by the corresponding time interval to obtain the degradation velocity vector. The degradation velocity vector is used to represent the direction and rate of change of the multidimensional features within the current time period, i.e., the overall trend of change. Based on this, the degradation velocity vector is again differentially calculated, i.e., the degradation velocity vectors at adjacent time points are differentially calculated and divided by the time interval to obtain the second-order time derivative, corresponding to the degradation acceleration vector, which is used to represent the change in the trend itself, i.e., the degree of change in the rate of change. In some implementations, the derivative results can also be smoothed or outlier suppressed to improve computational stability.

[0087] Peripheral identification information, such as device ID and driver name, is obtained from device operation information, system configuration files, or driver loading information. Based on this identification information, peripheral types are identified. Matching can be performed using a pre-set database to categorize specific peripherals into their corresponding peripheral types. In this embodiment, peripheral types can include printing peripherals, scanning and recognition peripherals, input interaction peripherals, image acquisition peripherals, and communication peripherals. The pre-set database is constructed based on the correlation between peripheral types and their corresponding physical attributes in historical data, including device operation records, performance monitoring data, and device specifications. Statistical analysis and summarization of the response characteristics, load characteristics, and stability characteristics of different types of peripherals during actual operation are performed to extract representative physical attribute parameters. Peripheral types and their corresponding physical attributes are then structured and stored to form a database for rapid matching. The physical attributes of the peripherals corresponding to the peripheral types are matched using the pre-set database. Physical attributes include, but are not limited to, response latency characteristics, mechanical inertia characteristics, processing throughput, thermal stability characteristics, and communication transmission characteristics. After peripheral type identification is completed, the physical attribute parameters of the corresponding peripheral type are retrieved from the database based on the identification results.

[0088] Based on the peripheral type determined in the preceding steps, corresponding physical attribute parameters, such as response delay coefficient, thermal stability coefficient, mechanical inertia coefficient, or data processing hysteresis characteristics, are retrieved from a preset database. These physical attribute parameters are normalized, and a degenerate inertia coefficient matrix is ​​constructed based on the dimensional structure of the multidimensional phase characteristics. This matrix can be a diagonal matrix or a weighted matrix, where each element represents the degree of hysteresis in response to changes in the corresponding feature dimension. Next, the degenerate inertia coefficient matrix is ​​multiplied or weighted with the currently calculated degenerate velocity vector, modulating the original change trend according to the peripheral's physical characteristics in different dimensions to obtain a degenerate momentum vector. This degenerate momentum vector not only reflects the direction and speed of feature changes but also incorporates the peripheral's response inertia factor, making the characterization of degenerate behavior more consistent with actual physical characteristics. Through this method, the transformation from a simple description of change velocity to a description of change momentum with physical meaning is achieved.

[0089] A peripheral operating state space is constructed by combining the degradation momentum vector and degradation acceleration vector, and the healthy state boundary range is determined within the peripheral operating state space by incorporating historical healthy operating data from historical operating data. Specifically, the degradation momentum vector and degradation acceleration vector calculated at each moment are concatenated or combined to form a unified state feature vector. This state feature vector is then used as coordinates to map the operating state of the peripheral at different time points to discrete points in the peripheral operating state space. Subsequently, sample data in normal or healthy states are selected from the historical operating data, and the corresponding state feature vectors are mapped to the state space to form a set of healthy sample points. Next, the distribution of the healthy sample point set is modeled. Methods such as statistical envelope, cluster analysis, or density estimation can be used to determine the concentrated region of healthy samples in the state space, and the healthy state boundary range is constructed based on this concentrated region. For example, this can be done by calculating the value range of each dimension of features, constructing a multidimensional confidence region, or generating an envelope boundary surface. By using the above method, spatial modeling of the normal operating state of peripherals is achieved, so that the operating state at any subsequent moment can be evaluated by the relationship between its position in the state space and the health boundary, thereby providing a basis for judgment for degradation detection and trend prediction.

[0090] The degradation momentum vector and degradation acceleration vector corresponding to the current temporal morphological feature sequence are mapped to the peripheral device's operating state space. Specifically, based on the current temporal morphological feature sequence, the corresponding degradation momentum vector and degradation acceleration vector are calculated using the aforementioned method. These two vectors are then concatenated or merged according to a preset combination method to form a unified state feature vector. For example, the degradation momentum vector can be used as the first half of the feature dimension, and the degradation acceleration vector as the second half. This state feature vector is used as coordinate input and mapped to the constructed peripheral device operating state space, where each dimension of the state space corresponds one-to-one with the dimension of the feature vector, thus determining a unique mapping point in the state space. In some implementations, the feature vector can be normalized or scaled before mapping to ensure comparability between different dimensions.

[0091] In the constructed state space, the degenerate momentum vector corresponding to the current moment is first determined as the initial state position, and the corresponding degenerate velocity vector and degenerate acceleration vector are obtained. Then, at discrete time steps, the degenerate momentum vector is recursively predicted. For example, within each time step, based on the current degenerate momentum vector superimposed with the change determined by the degenerate velocity vector and degenerate acceleration vector, the predicted momentum vectors corresponding to multiple future moments are calculated step by step, thus forming an evolution trajectory in the state space. In some implementations, linear extrapolation, a second-order motion model, or a constrained recursive model can be used to predict the trajectory, and smoothing processing can be combined to improve stability. Next, the predicted evolution trajectory is compared and analyzed with the pre-determined healthy state boundary range, calculating the distance or deviation of each predicted point in the trajectory from the healthy boundary, for example, determining whether it has entered the region outside the boundary or is close to the boundary critical region. Finally, based on the degree of deviation, the characteristic change range within the future time window is determined. For example, the projection range of the trajectory within the healthy boundary is defined as the normal fluctuation range, the range close to the boundary is defined as the warning range, and the range exceeding the boundary is defined as the abnormal or degenerate range. The above method enables trajectory-level prediction of future performance changes based on the current state, and transforms the prediction results into a range of feature changes that can be used for judgment.

[0092] The above steps improve the ability to predict progressive degradation trends, enhance the personalized accuracy requirements for the differences in physical characteristics of different peripherals, and avoid the fitting distortion problem of traditional linear extrapolation methods for nonlinear degradation processes. They can significantly improve the adaptability and predictive reliability of the health delay baseline model, and provide dynamic decision-making basis for early warning of peripheral degradation status and collaborative operation and maintenance strategies.

[0093] In one embodiment of this example, an environmental common-mode device group is constructed based on the directed topology of the service flow and device environment information. Then, a group common-mode offset verification is performed by combining the health latency baseline model and the environmental common-mode device group to determine the peripheral degradation state corresponding to the medical device terminal, including: S810. Determine the set of peripheral event nodes for nodes in the same business stage based on the directed topology of the business flow, and construct an environmental feature vector based on the device environment information. S820: Combine environmental feature vectors and peripheral event node sets to filter the set of medical device terminals corresponding to the business stages; S830: Obtain the business concurrency data of the medical device terminal set within a preset time window, and determine the load phase characteristic parameters through the business concurrency data; S840. Clustering the medical device terminal set using load phase characteristic parameters yields an environmental common-mode device group. S850: Obtain the real-time temporal morphological feature sequence of each medical device terminal in the environmental common mode device group, and extract the common mode change features in the environmental common mode device group by combining the real-time temporal morphological feature sequence. S860. Construct a common-mode offset compensation matrix based on common-mode variation characteristics to reflect the temporal morphological offset relationship caused by the combined effects of environmental factors and service load. S870. Identify the target medical device terminal in the environmental common mode device group and obtain the current temporal morphological feature sequence of the target medical device terminal; S880. Perform a difference operation between the current time-series morphological feature sequence and the common-mode offset compensation matrix to obtain individual differential mode features, so as to determine the individual differential mode degradation features of the equipment after stripping away the influence of environmental common modes. S890. Input the individual differential mode features into the health delay baseline model for offset verification, and determine the degradation status of the corresponding peripheral device based on the offset results.

[0094] Based on the directed topology of the business flow, a set of peripheral event nodes for nodes in the same business stage is determined, and an environmental feature vector is constructed based on device environment information. This means that peripheral event nodes are divided into stages based on the directed topology of the business flow. Stage division can be based on functional nodes or execution steps in the business process; for example, peripheral event nodes for the same business step, such as scanning, payment, and printing, are grouped into the same set of business stage nodes. Subsequently, all peripheral event nodes in the same business stage are selected from the topology and combined into a set of peripheral event nodes to represent the peripheral operation behavior involved in that business stage. Further, device environment information of the medical device terminal during operation is obtained. This environmental information includes, but is not limited to, parameters such as network latency, system load, CPU utilization, memory usage, temperature status, and geographical location. Then, the environmental information is normalized and features are extracted. Environmental parameters of different dimensions are encoded according to a unified scale to construct an environmental feature vector, where each feature dimension corresponds to an environmental influencing factor. Finally, the environmental feature vector is associated with the corresponding set of peripheral event nodes to represent the peripheral operating status of that business stage under specific environmental conditions. By adopting the above method, joint modeling of business stages and environmental factors can be achieved, providing a foundation for subsequent environmental common-mode equipment group division and degradation impact analysis.

[0095] This method combines environmental feature vectors and peripheral event node sets to filter medical device terminal sets. First, the target business stage is determined based on the directed topology of the business flow, and the corresponding peripheral event node set is obtained. Then, historical or current operating data is retrieved from multiple medical device terminals to filter those that are in the business stage and have executed the corresponding peripheral event node set. Environmental feature vectors for each candidate terminal are obtained, and based on preset similarity rules, such as Euclidean distance and cosine similarity, terminals whose environmental feature vectors match or are similar to the target environmental conditions are selected. Segmentation or clustering of environmental features can also be performed to improve filtering accuracy. Finally, medical device terminals that simultaneously meet the criteria of being in the same business stage and having similar environmental features are aggregated to form a medical device terminal set.

[0096] The next step is to acquire the business concurrency data of the medical device terminal set within a preset time window. This involves statistically analyzing the business execution status of each device in the set within that time window, obtaining the number of business triggers, concurrent tasks, or simultaneous active peripheral calls per unit time, thus forming the business concurrency data. Next, load phase characteristic parameters are determined using this data. This involves constructing a concurrency time series from the business concurrency data in chronological order and normalizing or smoothing it to eliminate scale differences and the impact of instantaneous fluctuations. Load phase characteristic parameters are then extracted from the concurrency time series. For example, by analyzing the periodicity, peak position, and rise and fall rhythm of concurrency changes, the phase position and rhythm characteristics of load changes in the time dimension can be determined. Sliding window statistics, frequency domain analysis, or phase coding methods can be used to process the concurrency series to obtain multidimensional load phase characteristic parameters.

[0097] For each device in the medical device terminal set, its corresponding load phase characteristic parameters are obtained. These parameters may include the periodicity of concurrent load changes, peak positions, phase offsets, and change rhythms, and are constructed into a feature vector of a unified dimension. The feature vector is then normalized. Further, based on the load phase characteristic vector, a pre-defined clustering algorithm is used to classify the medical device terminals by similarity, such as Euclidean distance or cosine similarity. The degree of feature similarity between devices is calculated, and the devices are divided into several categories according to the degree of similarity. K-means clustering and hierarchical clustering can also be used for grouping, and the grouping results are optimized using clustering effect evaluation metrics. Finally, medical device terminals with similar load phase characteristics are grouped into the same group and defined as the environmental common-mode device group. This approach ensures that devices within the same group exhibit consistency in load changes and environmental impacts, thus providing a reliable foundation for subsequent common-mode offset analysis and degradation attribution.

[0098] Subsequently, for each medical device terminal in the environmental common-mode device group, its corresponding temporal morphological feature sequence is acquired in real time, and the feature sequences of each device are time-aligned according to a unified time benchmark. In the aligned multi-device feature sequences, statistical analysis is performed on the feature values ​​of each device's corresponding dimension within the same time window, such as calculating the mean, median, or weighted average, to extract the changing trends simultaneously presented in most devices. By performing difference analysis or correlation analysis on the multi-device feature sequences, highly consistent change components among multiple devices are identified, and this part is defined as the common-mode change feature, used to represent the common impact of environmental factors or business load on the entire device group. The common-mode feature can be smoothed or dimensionality reduced to enhance its stability and expressive power. In this embodiment, the extraction of common-mode change features within the environmental common-mode device group by combining real-time temporal morphological feature sequences includes the following steps: We construct weighting coefficients by combining the local signal quality parameters corresponding to the real-time temporal morphological feature sequence of each medical device terminal, and then use the weighted time warping method to perform time axis alignment processing on the real-time temporal morphological feature sequence of the medical device terminal using the weighting coefficients. Determine the business flow information of medical device terminals, and extract the upstream and downstream business logic relationships between medical device terminals through the business flow information; A feature adjacency structure is constructed based on the aligned real-time temporal morphological feature sequence and the upstream and downstream relationships of business logic; Under the constraint of the characteristic adjacency structure, the nonlinear dominant morphology within the environmental common mode device group is extracted, and the nonlinear dominant morphology is used as the initial common mode expression of the common mode variation characteristics; The common-mode initial expression is separated by combining the pre-acquired load phase characteristic parameters, and the environmental common-mode components are obtained by separating the characteristic components related to load changes from the characteristic components unrelated to load changes. The novelty index is calculated based on the difference between the environmental common mode component and the historical common mode features in the preset database. The common mode components of the environment are hierarchically updated based on the novelty index to obtain the dynamically updated common mode change characteristics.

[0099] In this embodiment, to address the problem that the extracted common mode variation features only reflect the current state and are difficult to adapt to changes in the environment over time, the common mode variation features are subjected to structured modeling and dynamic updating. Specifically, the common mode variation features extracted at each time point are represented in a pattern and continuously accumulated and updated to form a reusable common mode morphological expression. Based on this, a time decay mechanism is introduced, assigning higher weights to recently observed common mode features and gradually decreasing weights to historical common mode features, thereby achieving continuous tracking of environmental change trends and weakening the impact of outdated information. Furthermore, during the update process, an adaptive update strategy is executed based on the similarity between the current common mode variation features and historical patterns. When the current common mode variation features and historical patterns meet preset similarity conditions, the corresponding patterns are incrementally updated; when the similarity conditions are not met, new pattern expressions are introduced to expand the overall representation capability, thereby achieving adaptive representation of environmental and business pattern changes. Based on this, the common mode baseline is dynamically reconstructed by combining the current observed common mode features and historical accumulated information, so that the common mode expression no longer depends on single observation results, but rather forms a stable and evolvable reference baseline based on the fusion of historical and current information. This technical solution effectively resists chronic environmental drift caused by factors such as temperature changes, power supply aging, and long-term load fluctuations, incorporating gradual environmental changes into the normal model range and avoiding misjudgments as abnormal states. Simultaneously, because common-mode variation characteristics are continuously absorbed and updated, the differential-mode characteristics after common-mode stripping are purer, thereby improving the ability to identify subtle equipment degradation. Furthermore, the continuous online update mechanism avoids the failure issues of traditional static models during long-term operation, maintaining stable performance without frequent model reconstruction. Moreover, the more accurate common-mode baseline provides a reliable reference for subsequent individual differential-mode calculations based on common-mode compensation, thereby improving the accuracy and stability of equipment degradation state determination.

[0100] In this embodiment, the real-time temporal morphological feature sequence includes local signal quality parameters. These parameters characterize the extraction confidence of the feature sequence at different time points and can be determined based on the jitter of the local temporal morphology represented by the feature variance of the feature vector. The local signal quality parameters are normalized, with a value range of [0,1]. A larger value indicates a more reliable feature representation at the corresponding time point. Then, the local signal quality parameters at each time point are converted into matching weights through a preset nonlinear mapping function, ensuring that high-quality segments receive higher weight values ​​and low-quality segments receive lower weight values. The nonlinear mapping function is preferably an sigmoid function, with the following expression:

[0101] Where w(t) represents the weight at time t; q(t) represents the local signal quality parameter at time t, with a value range of [0,1]; e is the base of the natural logarithm, approximately equal to 2.71828; k is the slope control parameter, used to adjust the steepness of the weight change. It is greater than 0; q0 is the weight transition threshold, with a value between (0,1). q0 can be obtained by collecting the characteristic reconstruction error distribution of normal equipment under steady-state operation in historical operating data, and setting q0 as the quality score corresponding to the upper quartile or median of this distribution. k can be inversely proportional to the standard deviation σq of the quality parameter over its entire lifecycle. If the quality parameter is usually concentrated in a narrow interval around q0, then k should be larger to improve the function's resolution, amplify small quality differences as much as possible, and thus finely distinguish high / low quality segments. Conversely, if the quality parameter is widely distributed, then k should be smaller to maintain smoothness.

[0102] Next, the local distance metric in the dynamic time warping algorithm is weighted and corrected using weighting coefficients to construct a weighted cumulative distance matrix. Specifically, for the first and second feature sequences to be aligned, when calculating the local distance of any alignment point (i, j) on the dynamic warping path, a joint quality weight, determined by the weighting coefficients of the first feature sequence at time i and the second feature sequence at time j, is introduced. This joint quality weight is multiplied by the local distance to obtain the weighted local distance. Subsequently, dynamic programming is recursively performed based on the weighted local distance to obtain the weighted cumulative distance matrix, and the optimal warping path is determined through path backtracking. During the path backtracking process, a weight gain term composed of the joint quality weights is introduced to guide the warping path to preferentially pass through high-quality alignment points with larger joint quality weights. This ensures accurate matching of the alignment result within the reliable feature extraction time period, while allowing flexible and rapid traversal through noisy segments with poor signal quality, avoiding excessive stretching and distortion of the time axis caused by forcibly fitting chaotic waveforms.

[0103] The process involves determining the business flow information of medical device terminals, including acquiring request logs, API call records, and transaction tracing data generated during business execution. This business flow information characterizes the transfer of business requests between different medical device terminals. Based on this information, business instances are partitioned, and within each instance, the operational behaviors of the medical device terminals are sorted according to timestamps to construct corresponding business execution sequences. Based on the business execution sequences and the call or data transfer relationships between medical device terminals, the upstream and downstream relationships of business logic between different medical device terminals are extracted. Specifically, the medical device terminal that occurs first in the business execution sequence and provides input data or triggering conditions for subsequent devices is identified as the upstream device, and the medical device terminal that occurs later and depends on the output of the preceding device is identified as the downstream device.

[0104] The connection relationships between medical device terminals are determined based on the upstream and downstream relationships of business logic, and an initial adjacency matrix is ​​constructed accordingly, establishing connections between medical device terminals with business dependencies. Based on the aligned real-time temporal morphological feature sequence, the feature similarity between different medical device terminals is calculated. Feature similarity is used to characterize the consistency of changes among each medical device terminal under a unified time reference. Furthermore, the connection relationships and feature similarity are fused to construct a weighted adjacency structure, enabling feature associations to be established between medical device terminals with only business relationships, and the association strength is determined based on the degree of feature similarity. Preferably, directional constraints are introduced when constructing the weighted adjacency structure to distinguish the degree of influence between upstream and downstream business processes; and the weighted adjacency structure is normalized to obtain a stable feature adjacency representation, thereby limiting the scope of common-mode feature extraction and suppressing interference from irrelevant devices on the common-mode analysis results.

[0105] Under the constraints of the constructed feature adjacency structure, nonlinear dominant morphologies within the environmental common-mode device group are extracted, and these morphologies are used as the initial expression of common-mode variation characteristics. Specifically, the set of medical device terminals participating in the analysis and their connection relationships are determined based on the feature adjacency structure, and a weighted adjacency representation is constructed based on these connections. Based on the real-time temporal morphological feature sequence aligned to the time axis, a corresponding group feature matrix is ​​constructed to characterize the collaborative change state of each medical device terminal under a unified time reference. Furthermore, combined with the weighted adjacency representation, structural constraints are applied to the group feature matrix, and nonlinear feature extraction processing is performed under these constraints to obtain a low-dimensional dominant morphological representation that can characterize the overall change trend of the medical device terminal group. The nonlinear feature extraction processing is used to suppress individual differences and noise interference while preserving the group's temporal change structure. Finally, the low-dimensional dominant morphological representation is determined as the initial expression of the common-mode variation characteristics of the environmental common-mode device group for subsequent common-mode offset modeling and individual differential mode feature analysis.

[0106] The initial common-mode expression is feature-separated by combining pre-acquired load phase characteristic parameters, and the feature components related to load changes are separated from those unrelated to load changes to obtain the environmental common-mode components. First, paired data sequences aligned in the time dimension between the initial common-mode expression and the load phase characteristic parameters are obtained. These paired data sequences are input into a pre-constructed parameterized separation model to perform component decomposition on the initial common-mode expression. The parameterized separation model contains two output branches: the first branch uses the load phase characteristic parameters as conditional input to generate load disturbance common-mode components strongly correlated with load changes; the second branch is decoupled from the load phase characteristic parameters to generate environmental common-mode components unrelated to load changes. After obtaining the initial separation results, disturbance response consistency correction is also performed. Based on the characteristics that real environmental disturbances typically have a slow time constant and exhibit consistent behavior across multiple devices in the same physical space, the time series of the environmental common-mode components is validated for reasonableness. If high-frequency, inconsistent random fluctuations are detected in the environmental common-mode components, residual load jitter and noise interference are eliminated through low-pass filtering or cross-device smoothing to obtain the corrected environmental common-mode components. The pre-built parameterized separation model includes a shared encoder and two parallel decoding branches. The shared encoder compresses the input common-mode initial representation into a latent feature representation. The first decoding branch takes the latent feature representation and load phase feature parameters as joint conditional inputs and outputs a load perturbation common-mode component related to load changes. The second decoding branch takes only the latent feature representation as input, does not accept load phase feature parameters, and outputs an environmental common-mode component unrelated to load changes. The outputs of the two decoding branches are summed by a summing layer to form the reconstructed signal of the common-mode initial representation. During training, a joint loss function is constructed, including a reconstruction loss term and a statistical independence penalty term. The reconstruction loss term constrains the difference between the reconstructed signal and the original common-mode initial representation. The statistical independence penalty term maximizes the statistical independence between the environmental common-mode component and the load phase feature parameters, for example, by minimizing their mutual information estimate. Using the common-mode initial representation sequence and the load phase feature parameter sequence aligned within a historical time period as training samples, the model is iteratively optimized by minimizing the joint loss function until convergence.

[0107] During the equipment's healthy operation phase, multiple environmental common-mode components are collected, clustered to form several typical pattern prototypes, and stored in a database. Each record contains a feature vector H of the pattern. k Meta-information such as frequency of occurrence and last update time. For the currently extracted environmental common-mode component, its relationship with each historical pattern H in the database is calculated. k The difference can be calculated based on cosine distance, and then the nearest neighbor distance between the current component and all historical patterns is taken as the novelty index. Alternatively, the mean of K-nearest neighbor distances or the maximum similarity can be used as the index. This indicates that if even the most similar historical patterns are far apart, the current pattern is considered very novel.

[0108] The environmental common-mode components are hierarchically updated according to the novelty index to obtain dynamically updated common-mode change features, that is, a two-layer memory structure is constructed, including long-term feature storage and short-term feature buffering. The long-term feature storage stores the verified stable environmental common-mode patterns and their statistical information, which are retained for a long time after being written and updated slowly; the short-term feature buffering stores the novel or sporadic patterns to be verified, which have a short retention time and are used to observe whether they appear repeatedly. According to the calculated novelty index S obtained above, compare it with the preset low threshold θlow and high threshold θhigh. When S≤θlow, it is a normal mode, that is, the currently extracted environmental common-mode component is highly similar to a certain historical common-mode feature in the preset database. This means that this environmental disturbance is a fluctuation form that has repeatedly occurred during the long-term operation of the device and has been verified as normal, belonging to the known knowledge of the system; when θlow<S<θhigh, it is mildly novel, that is, a transitional mode. Mild novelty means that there is a perceptible deviation between the currently extracted environmental common-mode component and the known patterns in the historical pattern library, but it has not reached the level of being completely unfamiliar. It lies between familiarity and unfamiliarity and is in a fuzzy transitional zone. The system can neither confirm that it is a natural variant of the existing pattern nor assert that it is a completely new unknown disturbance; when S≥θhigh, it is highly novel, that is, an abnormal or completely new pattern. Among them, the completely new pattern means that it persists or appears repeatedly, indicating that there has been a permanent and new structural change in the device operating environment. For example: after the relocation of the department, different frequencies of background vibration are introduced in the new computer room, new equipment generating specific electromagnetic interference is added, or seasonal changes bring temperature and humidity combinations that have never been seen before. The update operation for the normal mode is to perform exponential weighted averaging on the currently extracted environmental common-mode component and the most matching historical mode, and the specific expression is as follows:

[0109] Among them, H k (new) represents the kth historical common-mode pattern after update, which is the new version of the historical pattern obtained after integrating the information of the currently extracted environmental common-mode component and will be written back to the long-term feature storage; H k (old) is the kth historical common-mode pattern before update. That is, the original version of the historical pattern stored in the database that is most matching the currently extracted environmental common-mode component; C E is the currently extracted environmental common-mode component, that is, the pure environmental time-series morphological feature separated from the initial common-mode expression in this round and independent of load changes; α is the learning rate, that is, the smoothing coefficient, and its value range is 0<α≤1, preferably taking a larger value in the normal mode, such as 0.3~0.5, and the specific value can be set according to the actual situation calibrated by the enterprise in the laboratory.

[0110] Additionally, mildly novel update operations update the best-matching pattern in the long-term feature store with a small learning rate; simultaneously, the current environmental common-mode component is written to the short-term feature buffer and its frequency of occurrence is marked. Further logic can be added: if the same pattern appears consecutively in the short-term buffer a preset threshold set by the system (the specific value can be set according to enterprise requirements), it indicates a trend, and it is promoted to the long-term feature store. Highly novel update operations are isolated and do not update the long-term feature store initially. The current environmental common-mode component is stored in the short-term feature buffer, and an anomaly marker is triggered. If the pattern no longer appears within the observation window, it is removed from the short-term buffer; if it continues to appear stably, it is included in the long-term feature store as a new environmental pattern after manual or automatic verification. After the hierarchical update is executed, the updated best-matching pattern in the long-term feature store is combined with the current observation to reconstruct the common-mode change characteristics at the current moment, for use in the subsequent construction of the common-mode offset compensation matrix.

[0111] By implementing tiered updates, the system can adapt to gradual environmental drift, such as monthly changes in computer room temperature and humidity. Through continuous and smooth updates using regular patterns, the new environmental state is gradually absorbed into the baseline, avoiding false alarms. Furthermore, the system learns gradually from new environmental states, such as changes in power supply environment after departmental renovations. Through repeated exposure to transitional patterns, these changes are validated and eventually incorporated into long-term memory, enabling the model to self-evolve.

[0112] A common-mode offset compensation matrix is ​​constructed based on common-mode variation characteristics. Before constructing the matrix, the temporal feature dimensions of all peripherals must be standardized to ensure that the real-time temporal morphological feature sequence structure output by all devices is completely consistent, including command interaction, business processing, and anomaly types. Next, common-mode variation characteristics are extracted from the environmental common-mode device group. This extraction can be based on the median method, taking the median of the temporal values ​​of the same dimension for all devices within the group. This method effectively filters out interference from individual abnormal devices that have severely degraded within the group and is more robust than the mean method. The final extracted common-mode variation characteristics are a vector with the same dimension as the temporal feature sequence. For example, the common-mode feature vector of a certain group is [125ms, 48ms, 22ms, ...], with a total of 12 elements, each element corresponding to the group's average offset in one temporal dimension. The common-mode offset compensation matrix is ​​not a static two-dimensional table, but a dynamically updated condition-offset mapping library. First, the external conditions are discretized and binned. That is, in order to transform continuous environmental and load values ​​into indexable matrix rows, all external features need to be discretized. Environmental features are binned as follows: temperature is divided into 4 levels [0-10℃, 10-20℃, 20-30℃, 30-40℃], humidity is divided into 3 levels [0-30%, 30-60%, 60-100%], and power supply voltage is divided into 3 levels [210-220V, 220-230V, 230-240V], for a total of 4×3×3=36 environmental combinations. Load characteristics are binned: concurrency is divided into 4 levels: [0-10, 10-50, 50-100, 100+]; registration service ratio is divided into 3 levels: [0-30%, 30-70%, 70-100%]; peak time periods are divided into 4 levels: [morning peak, noon peak, evening peak, night], resulting in a total of 4×3×4=48 load combinations. This yields 36×48=1728 environment + load combinations, corresponding to 1728 rows of the matrix. Next, the matrix index and structure are established. The row index uniquely corresponds to one environment + load combination. The column index corresponds to a predefined temporal morphological feature dimension, for example, 12 columns. The matrix element is the value in the i-th row and j-th column, representing the average common-mode offset of the j-th temporal dimension for all devices under the i-th environment + load combination. All 1728 environment + load combinations are traversed. For each combination, all common-mode device groups under that combination in the current system are found. Extract the common mode variation feature vector of the group, and fill the 12 elements of the vector into the corresponding rows and columns of the matrix. If there are not enough devices in a certain combination, such as less than 3, then the common mode features of adjacent combinations are used for interpolation to fill the gap.

[0113] Within a shared-mode device group, target medical device terminals are identified based on preset conditions. These conditions may include, but are not limited to, devices requiring maintenance or those requiring focused monitoring. Subsequently, for each target medical device terminal, its interactive command information is acquired in real-time using bypass probes, and a temporal morphological feature sequence corresponding to its current operational stage is constructed using existing methods. During the acquisition process, the feature sequence undergoes time alignment, normalization, or denoising to ensure comparability with the features of other devices within the group. Ultimately, the current temporal morphological feature sequence of the target medical device terminal is obtained.

[0114] The current temporal morphological feature sequence is differentially analyzed with the common-mode offset compensation matrix to obtain individual differential-mode features, which are then used to determine the device's individual differential-mode degradation features after removing the influence of environmental common-mode. In other words, the aforementioned common-mode compensation model is used to correct the current temporal features of the target device, removing the overall offset caused by environmental factors and business load, and retaining only the features reflecting the device's own performance changes. Specifically, real-time environmental and load data of the target device are collected, and the corresponding row of the common-mode offset compensation matrix is ​​located according to the unified binning rules used during matrix construction. The N-dimensional standard common-mode offset vector of that row is extracted, where N is a predefined temporal feature dimension. The current N-dimensional temporal morphological feature sequence of the target device is element-wise subtracted from the common-mode offset vector to obtain the initial individual differential-mode features. If the number of devices in the group generating the common-mode vector is less than 3, the historical common-mode vector of that device under the same conditions for 24 hours is used instead, and outliers are filtered and the results of three consecutive time windows are exponentially weighted and smoothed. Finally, the differential-mode features are Z-score standardized to obtain the final pure individual degradation features, which are then input into the healthy latency baseline model for offset verification.

[0115] Finally, the individual characteristics of the device after the environmental and load co-mode effects have been removed are compared and analyzed with the pre-built health reference model to determine the degree of deviation of the current operating state from the normal baseline, and to determine whether the peripheral device has performance degradation and the type of degradation. First, the individual differential features obtained in the aforementioned steps are acquired and converted into a feature expression form consistent with the health latency baseline model, such as the corresponding time interval feature or multi-dimensional phase feature representation. Then, the individual differential features are input into the health latency baseline model for matching and comparison, calculating their offset in each feature dimension. For example, the difference between the current feature value and the reference interval of the corresponding time point in the baseline model is calculated, or it is determined whether it falls within the healthy range defined by the baseline model. Further, the offset results are judged based on the magnitude and trend of the offset. For example, when the offset is within the allowable range, it is judged as a normal state; when the offset is close to the boundary, it is judged as performance fluctuation or slight degradation; when the offset exceeds the threshold range, it is judged as a significant degradation state. In some implementations, the degree of degradation can be further refined by combining the offset duration or offset growth rate. Finally, the degradation state of the corresponding peripheral is determined based on the offset verification results and used as the basis for subsequent operation and maintenance decisions or strategy execution.

[0116] By verifying the common-mode offset of the group, the degradation status of the peripheral devices corresponding to the medical device terminal is determined. This improves the accuracy of distinguishing between common-cause latency degradation and individual true degradation in multi-peripheral collaborative scenarios, enhances the ability to compensate for the coupling interference of environmental factors and business load, and avoids the misjudgment problem caused by environmental fluctuations when a single device is monitored independently. It can significantly improve the confidence and anti-interference ability of peripheral degradation status determination, and provide a highly reliable status diagnosis basis for the differentiated triggering of collaborative operation and maintenance strategies, thereby ensuring the stability of medical self-service equipment operation and business processing efficiency.

[0117] This application also provides a machine-readable storage medium storing instructions that cause a machine to execute the above-described collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring.

[0118] This application also provides an electronic device, including: The memory is configured to store instructions; and The processor is configured to retrieve instructions from memory and, when executing instructions, implement the aforementioned collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring.

[0119] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0124] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0125] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (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, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0126] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0127] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for collaborative operation and maintenance of medical equipment terminals based on multi-dimensional state monitoring, characterized in that, A medical device terminal system applicable to connecting multiple medical device terminals, including: When a maintenance command is received from a medical device terminal, the connection status of the medical device terminal is determined, including a long connection status. If the medical device terminal is in a long-connection state, activate the bypass probe pre-deployed in the medical device terminal system and obtain the device operation information, business flow information and device environment information of the medical device terminal; Construct a directed topology for business flows by combining equipment operation information and business flow information; A time-series morphological feature sequence is constructed based on the interactive command information collected by the bypass probe, and a health delay baseline model is constructed through the time-series morphological feature sequence. Based on the directed topology of the business flow and the device environment information, an environment common mode device group is constructed. The group common mode offset is verified by combining the health latency baseline model and the environment common mode device group to determine the peripheral degradation status corresponding to the medical device terminal. The peripheral degradation status includes business load driven degradation, hardware path degradation and common mode environment degradation. Based on business load-driven degradation, hardware path degradation, and common-mode environment degradation, corresponding degradation operation and maintenance strategies are executed respectively.

2. The method of claim 1, wherein, The method further includes: When the peripheral device is in a state of degradation and shows a trend of degradation, idle peripheral devices that are not in the business call chain are selected according to the directed topology of the business flow. The bypass probe sends a silent command to the idle peripheral device without causing a change in the service status, and collects the instantaneous morphological feature sequence of the idle peripheral device. Determine the temporal degradation feature sequence of the corresponding target peripheral based on the peripheral degradation status of the medical device terminal; Under the same underlying hardware topology, the timing degradation feature sequence of the target peripheral is compared with the instantaneous morphological feature sequence of the idle peripheral to determine the cause of the degradation state of the peripheral. The peripheral degradation state is corrected based on the cause of attenuation to obtain the corrected peripheral degradation state.

3. The method according to claim 1, characterized in that, The connection status also includes an offline status. Determining the connection status of the medical device terminal includes: The medical device terminal sends heartbeat data packets to the cloud server connected to the medical device terminal system at preset heartbeat cycles. After sending a heartbeat data packet, receive the response data packet returned by the cloud server according to the preset response time window; When a response data packet is received within the preset response time window, the initial connection status of the medical device terminal and the active identifier of the communication link are determined by the round-trip time delay parameter and validity verification parameter of the response data packet. The long-term connection status of the medical device terminal is determined by combining the initial connection status and the active identifier of the communication link; If no response data packet is received within the preset response time window, the reconnection mechanism is triggered and a heartbeat data packet is sent according to the preset retry interval. If the number of consecutive retries reaches the preset retry threshold and no response data packet is received, the medical device terminal connection is determined to be abnormal, and the medical device terminal is marked as offline using the cloud server.

4. The method according to claim 1, characterized in that, The construction of the directed topology structure for the service flow by combining equipment operation information and service flow information includes: Extract the underlying call feature information of medical device terminal peripheral operation from the device operation information, and perform hash encoding on the underlying call feature information to generate corresponding peripheral event nodes; A discrete event node set is constructed based on peripheral event nodes, and the discrete event node set is serialized and arranged according to the corresponding timestamps to obtain the terminal concurrent action timeline; By combining business flow information and terminal concurrent action timeline, a time-series association rule mining algorithm is used to analyze the time-series dependency relationship between each peripheral event node to determine the sequential following relationship and causal relationship between each peripheral event node; Multiple implicit directed edges are generated by combining sequential following relationships and causal relationships; Aggregate implicit directed edges to construct a directed topology for business flows.

5. The method according to claim 4, characterized in that, The method further includes: Obtain historical timeline data corresponding to concurrent actions of historical terminals based on peripheral event nodes; In the directed topology of the business flow, for peripheral event nodes on different branch paths, the overlap statistical analysis of the temporal pattern of each peripheral event node is performed based on historical timeline data according to a preset time window. When the overlap probability of any two peripheral event nodes within a preset time window is less than or equal to a preset overlap threshold, it is determined that there is a mutual exclusion relationship between the two peripheral event nodes, and a corresponding mutual exclusion mapping relationship is established. Identify active event nodes corresponding to active state business paths based on the current terminal concurrent action timeline; Based on the mutual exclusion mapping relationship between the parent and child nodes and the active event nodes, the target event nodes that have a mutual exclusion relationship with the active event nodes are identified, and the idle state of the target event nodes is determined to obtain the idle state nodes. Idle peripheral resources are determined by dynamically updating the directed topology of the service flow based on idle state nodes.

6. The method according to claim 1, characterized in that, The process of constructing a time-series morphological feature sequence based on interactive command information collected by a bypass probe, and then constructing a healthy latency baseline model using the time-series morphological feature sequence, includes: The first timestamp of the instruction issuance time and the second timestamp of the hardware receipt completion are determined based on the interactive instruction information. The peripheral response timeslots are calculated by combining the first and second timestamps, and the peripheral response timeslots are constructed into raw timing data in chronological order. Collect raw time-series data of multiple identical services from the peripheral devices corresponding to the medical device terminal, and analyze and identify time disturbance components from the raw time-series data of multiple identical services. The target time series sequence is obtained by extracting the time components of stable repeating features based on the time perturbation components; Extract the time interval between each instruction in the target time sequence and calculate the ratio between any two adjacent time intervals; A multidimensional phase feature matrix is ​​constructed by combining the ratio relationships, and the multidimensional phase feature matrix is ​​arranged in chronological order to obtain a time-series morphological feature sequence; Acquire historical operating data for all peripherals in the medical device terminal, and calculate the trend and rate of change of time-series morphological feature sequences based on the historical operating data; Determine the range of characteristic changes within the future time window by combining the trend and rate of change; A healthy latency baseline model is generated based on the range of feature changes within a future time window.

7. The method according to claim 6, characterized in that, The method of determining the range of characteristic changes within a future time window by combining the trend and rate of change includes: The matrix offset is calculated based on the multidimensional phase feature matrix within the first sliding time window and the second sliding time window respectively, wherein the duration of the first sliding time window is greater than the duration of the second sliding time window; The degradation velocity vector and degradation acceleration vector are obtained by calculating the first and second time derivatives of the matrix offset. The degradation velocity vector is used to represent the trend of change, and the degradation acceleration vector is used to represent the rate of change. Determine the corresponding peripheral type in the medical device terminal and match the physical attributes of the peripheral type in the preset database; By combining the physical properties of the peripheral device, a corresponding degenerate inertia coefficient matrix is ​​assigned to the corresponding peripheral device, and the degenerate momentum vector is obtained by combining the degenerate inertia coefficient matrix with the degenerate velocity vector. The peripheral operating state space is constructed by combining the degradation momentum vector and the degradation acceleration vector, and the health state boundary range is determined within the peripheral operating state space by combining historical health operating data from historical operating data. Map the degenerate momentum vector and degenerate acceleration vector corresponding to the current temporal morphological feature sequence to the peripheral operating state space; In the peripheral operating state space, based on the changing trend of the degradation momentum vector and the degradation acceleration vector, the evolution trajectory of the degradation momentum vector within the future time window is predicted, and the range of characteristic changes within the future time window is determined according to the degree of deviation of the evolution trajectory from the healthy state boundary range.

8. The method according to claim 1, characterized in that, The process of constructing an environmental common-mode device group based on the directed topology of the service flow and device environment information, and performing group common-mode offset verification by combining the health latency baseline model and the environmental common-mode device group to determine the peripheral degradation status corresponding to the medical device terminal includes: The set of peripheral event nodes for nodes in the same business stage is determined based on the directed topology of the business flow, and an environmental feature vector is constructed based on the device environment information. The set of medical device terminals is filtered by combining environmental feature vectors and the business stages corresponding to the set of peripheral event nodes. Acquire the business concurrency data of the medical device terminal set within a preset time window, and determine the load phase characteristic parameters through the business concurrency data; Clustering of the medical device terminal set using load phase characteristic parameters yields an environmental common-mode device group; The real-time temporal morphological feature sequence of each medical device terminal in the environmental common mode device group is obtained, and the common mode change features in the environmental common mode device group are extracted by combining the real-time temporal morphological feature sequence. A common-mode offset compensation matrix is ​​constructed based on the common-mode variation characteristics to reflect the temporal morphological offset relationship caused by the combined effects of environmental factors and business load. Identify the target medical device terminal in the environmental common-mode device group and obtain the current temporal morphological feature sequence of the target medical device terminal; The current time-series morphological feature sequence is compared with the common-mode offset compensation matrix to obtain individual differential-mode features, so as to determine the individual differential-mode degradation features of the equipment after removing the influence of environmental common-mode. Individual differential features are input into the healthy delay baseline model for offset verification, and the corresponding peripheral degradation state is determined based on the offset results.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to execute a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring, as claimed in any one of claims 1 to 8.

10. An electronic device, characterized in that, include: The memory is configured to store instructions; as well as The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement a collaborative operation and maintenance method for medical device terminals based on multi-dimensional status monitoring according to any one of claims 1 to 8.

Images