An infusion quality safety digital traceability system and method based on the Internet of Things

Abstract

The present application relates to the technical field of digital infusion supervision, and particularly relates to an infusion quality safety digital tracing system and method based on an Internet of Things. The method comprises the following steps: generating infusion path space topology data according to the movement track of an infusion device from a pharmaceutical agent configuration area to a drug administration point according to a target infusion pipeline; building an intelligent Internet of Things sensing network according to the infusion path space topology data; collecting infusion multi-dimensional sensing signals using the intelligent Internet of Things sensing network, intelligently analyzing infusion disturbance events, and obtaining verified disturbance event data; processing risk event data chains according to the verified disturbance event data, and managing forward-looking risk warning instructions, so as to realize infusion quality digital tracing. The present application realizes real-time monitoring of an infusion process based on differentiated intelligent Internet of Things sensing nodes, identifies risk sources and predicts propagation paths, and completes digital tracing and forward-looking warning of infusion quality safety.

Description

Technical Field

[0001] This invention relates to the field of digital infusion monitoring technology, and in particular to an Internet of Things-based digital traceability system and method for infusion quality and safety. Background Technology

[0002] The flow of medication in infusion systems involves complex microchannel fluid dynamics, including interfacial interactions, temperature variations, and pressure fluctuations. These microscopic changes lead to quality problems such as reduced efficacy, foreign matter formation, or drug degradation. Traditional infusion systems generally employ simple mechanical control and manual monitoring, neglecting the non-uniformity of fluid behavior in three-dimensional space. Relying solely on single-point monitoring cannot comprehensively reflect the dynamic changes in the infusion process, especially in areas such as drug concentration gradients, flow rate fluctuations, and temperature changes, where there are significant monitoring blind spots. Along long-distance infusion routes, due to the combined effects of tubing bends, height differences, and drug characteristics, localized areas often experience abnormal flow rates, drug stagnation, or even backflow. These changes significantly affect the fluid dynamics within the tubing. However, traditional infusion quality management methods mainly rely on manual inspections, infusion pump monitoring, and post-infusion sampling inspections. These methods can only achieve simple monitoring and rough recording of the infusion process and cannot monitor and analyze the microscopic behavior of the infusion solution in real time. Especially in complex pipeline layout environments, traditional technologies are difficult to achieve end-to-end quality traceability, making it impossible to accurately locate and analyze abnormal points in the infusion process. Summary of the Invention

[0003] Based on this, the present invention provides an Internet of Things-based digital traceability system and method for infusion quality and safety to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an IoT-based digital traceability method for infusion quality and safety includes the following steps:

[0005] Step S1: Perform internal attitude scanning on the target infusion line to obtain internal attitude data of the infusion line; track the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal attitude data of the infusion line, and perform pipeline geometric feature analysis to generate infusion path spatial topology data;

[0006] Step S2: Construct an intelligent IoT sensing network based on the infusion path spatial topology data; perform theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data to obtain theoretical fluid state characteristic data; based on the theoretical fluid state characteristic data, use the intelligent IoT sensing network to collect infusion multi-dimensional sensing signals and perform time delay compensation to obtain spatiotemporal sensing monitoring signal data.

[0007] Step S3: Intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and perform cross-validation of disturbance events to obtain verified disturbance event data;

[0008] Step S4: Identify and track risk sources based on verified disturbance event data, and process the risk event data chain to obtain an infusion risk data chain; use an intelligent IoT sensing network to monitor disturbance events based on the infusion risk data chain to obtain a dynamic traceability map of infusion risks; manage forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

[0009] Preferably, the present invention also provides an Internet of Things (IoT)-based digital traceability system for infusion quality and safety, which executes the IoT-based digital traceability method for infusion quality and safety as described above. The IoT-based digital traceability system for infusion quality and safety includes:

[0010] The pipeline topology construction module is used to perform internal attitude scanning on the target infusion pipeline to obtain internal attitude data of the infusion pipeline; based on the internal attitude data of the infusion pipeline, it tracks the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point, and performs pipeline geometric feature analysis to generate infusion path spatial topology data;

[0011] The IoT sensing and monitoring module is used to build an intelligent IoT sensing network based on the spatial topology data of the infusion path; to perform theoretical risk characteristic analysis on the target infusion pipeline through the spatial topology data of the infusion path, and to obtain theoretical characteristic data of the fluid state; based on the theoretical characteristic data of the fluid state, the intelligent IoT sensing network is used to collect multi-dimensional sensing signals of the infusion and to perform time delay compensation, so as to obtain spatiotemporal sensing and monitoring signal data.

[0012] The disturbance intelligent analysis module is used to intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and to perform cross-validation of disturbance events to obtain verified disturbance event data.

[0013] The risk traceability and early warning module is used to identify and track risk sources based on verified disturbance event data, and to process the risk event data chain to obtain the infusion risk data chain; based on the infusion risk data chain, it uses an intelligent Internet of Things sensing network to monitor disturbance events to obtain a dynamic traceability map of infusion risks; and it manages forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

[0014] This invention organically combines microscopic fluid dynamics characteristics with macroscopic infusion management, achieving full-process, multi-dimensional monitoring from drug preparation to patient administration, breaking through the technical bottlenecks of traditional monitoring methods. The system relies on infusion tubing internal attitude scanning technology to accurately capture tubing geometric features. Multi-dimensional sensing technology based on IoT sensing networks overcomes the limitations of single-dimensional monitoring, achieving real-time monitoring of multiple parameters such as flow rate, pressure, temperature, and drug-liquid interface characteristics, and ensuring spatiotemporal consistency of data through a time delay compensation algorithm. The system's intelligent disturbance event analysis mechanism, through cross-validation of theoretical models and measured data, significantly improves the accuracy of anomaly identification, reducing the false alarm rate by at least 30%. Risk source tracing and data chain processing construct a complete quality traceability chain, completely solving the problem of broken traceability chains in traditional systems, refining problem location accuracy from the original "infusion process" to "specific links and time points," improving traceability accuracy by over 80%. The construction of a dynamic traceability map transforms static data into a dynamic risk evolution model, enabling the system to possess proactive risk warning capabilities. This shifts the risk management model from passive response to proactive prevention, with an average warning lead time of 15 minutes, providing valuable time for medical staff intervention. The system also breaks down existing data silos, achieving seamless integration and correlation analysis of data from multiple stages within the hospital, including pharmacies, nurse stations, and wards, providing comprehensive data support for medical quality control. In clinical applications, the system reduces the incidence of problems such as decreased drug efficacy, foreign body formation, and drug degradation, optimizes the allocation of medical resources, improves patient medication safety, and provides data-driven decision support for clinical medical staff. This makes the effects of infusion therapy more controllable, visible, and traceable, addressing the difficulty in predicting bubble formation and migration behavior under different spatial locations and environmental conditions in existing technologies. Therefore, the present invention provides an IoT-based digital traceability method for infusion quality and safety. This method achieves precise three-dimensional attitude scanning of the infusion tubing using a flexible guidewire with a built-in micro inertial measurement unit array. It utilizes fluid state theory characteristic data to guide the deployment of differentiated sensors and the setting of acquisition frequencies. It employs a multi-dimensional cross-validation mechanism of ultrasonic echo, infrared light transmittance, and pressure changes to accurately identify micro-disturbance events. It constructs a risk event data chain by combining spatiotemporal deviation correlation technology. It predicts the evolution characteristics of risk flow state through hydrostatic pressure sequence trend analysis of propagation path. It marks downstream high-risk nodes based on the event occurrence probability model. Finally, it forms a digital traceability system for infusion quality and safety with forward-looking early warning function, realizing intelligent monitoring and risk management of the entire medical infusion process. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the IoT-based digital traceability method for infusion quality and safety of the present invention.

[0016] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0019] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] To achieve the above objectives, please refer to Figure 1 This invention provides a digital traceability method for infusion quality and safety based on the Internet of Things, comprising the following steps:

[0021] Step S1: Perform internal attitude scanning on the target infusion line to obtain internal attitude data of the infusion line; track the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal attitude data of the infusion line, and perform pipeline geometric feature analysis to generate infusion path spatial topology data;

[0022] Step S2: Construct an intelligent IoT sensing network based on the infusion path spatial topology data; perform theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data to obtain theoretical fluid state characteristic data; based on the theoretical fluid state characteristic data, use the intelligent IoT sensing network to collect infusion multi-dimensional sensing signals and perform time delay compensation to obtain spatiotemporal sensing monitoring signal data.

[0023] Step S3: Intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and perform cross-validation of disturbance events to obtain verified disturbance event data;

[0024] Step S4: Identify and track risk sources based on verified disturbance event data, and process the risk event data chain to obtain an infusion risk data chain; use an intelligent IoT sensing network to monitor disturbance events based on the infusion risk data chain to obtain a dynamic traceability map of infusion risks; manage forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

[0025] In this embodiment of the invention, the IoT-based digital traceability method for infusion quality and safety includes the following steps:

[0026] Step S1: Perform internal attitude scanning on the target infusion line to obtain internal attitude data of the infusion line; track the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal attitude data of the infusion line, and perform pipeline geometric feature analysis to generate infusion path spatial topology data;

[0027] In this embodiment of the invention, a medical-grade silicone flexible guidewire with an array of 30 miniature inertial measurement units (MMUs) is inserted into the infusion tubing. The guidewire has a diameter of 0.5 mm, and the MMUs are spaced 5 cm apart. Each MMU includes a triaxial accelerometer, a triaxial gyroscope, and a triaxial magnetometer. The guidewire passes through the entire infusion tubing system at a uniform speed of 2 cm / s, while each MMU records spatial attitude data at a frequency of 50 Hz. The recorded data is processed using a Kalman filter algorithm to eliminate noise interference. When tracking the movement trajectory of the infusion device, the attitude data is converted into spatial attitude in Euler angle representation using a six-degree-of-freedom attitude calculation algorithm. The acceleration data is converted into displacement data using a double integration method, and zero-velocity update technology is applied to eliminate integral drift error. The obtained displacement data is sampled at 1 cm intervals to form a three-dimensional trajectory coordinate dataset of 4000-6000 sampling points. When performing pipeline geometric feature analysis, the B-spline fitting algorithm is used to extract the pipeline centerline, and parameters such as curvature and height difference are accurately calculated. Finally, the spatial topology data of the infusion path, which includes 150-200 nodes and connections, is constructed by associating it with the environmental anchored spatial coordinates through a spatial point matching algorithm.

[0028] Step S2: Construct an intelligent IoT sensing network based on the infusion path spatial topology data; perform theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data to obtain theoretical fluid state characteristic data; based on the theoretical fluid state characteristic data, use the intelligent IoT sensing network to collect infusion multi-dimensional sensing signals and perform time delay compensation to obtain spatiotemporal sensing monitoring signal data.

[0029] In this embodiment of the invention, the coordinates of the flow transition critical point, sharp bends, and sections with vertical height changes exceeding 50 cm are calculated based on the spatial topology data of the infusion path, forming preliminary key monitoring point coordinate data. The flow transition critical point is assigned to the gas-liquid interface monitoring task and equipped with a 40 MHz high-frequency ultrasonic transducer; the sharp bends are assigned to the thermal flow field disturbance monitoring task and equipped with temperature sensors and infrared transceiver pairs; the midpoint of the vertical height change section is assigned to the hydrostatic pressure calibration task and equipped with piezoresistive pressure sensors and capacitive sensors. During theoretical risk characteristic analysis, the viscosity, density, and surface tension parameters of the infusion drugs are obtained from the drug management system, and the ambient temperature is monitored and temperature compensation is performed using a sensing network. The theoretical hydrostatic pressure and additional flow resistance are calculated based on the vertical height difference and curvature data, the Reynolds number distribution within the microchannel is estimated, and the Marangoni effect risk index is calculated. Subsequently, a 100 Hz acquisition frequency was set for the sensor nodes corresponding to the high-risk turbulence zone, and 20 Hz was set for other nodes. Multidimensional sensing signal acquisition was initiated to obtain multi-source heterogeneous real-time sensing flow, including pipeline start and stop displacement sequences, measured hydrostatic pressure data, infrared light transmittance sequences, and ultrasonic echo signal flow. After digital filtering and time delay compensation, spatiotemporal sensing monitoring signal data containing absolute timestamps were generated.

[0030] Step S3: Intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and perform cross-validation of disturbance events to obtain verified disturbance event data;

[0031] In this embodiment of the invention, the intelligent analysis of infusion disturbance events first extracts the transit time and attenuation coefficient of the ultrasonic echo signal flow from the spatiotemporal sensing monitoring signal data to form an acoustic characteristic parameter sequence. When the transit time shows a jump greater than 1 microsecond, it is marked as a suspected bubble event; when the attenuation coefficient continuously increases by more than 0.2 dB, it is marked as a suspected particulate event. The Marangoni effect intensity risk index at the marked event location is queried, and simultaneously, liquid contact angle analysis is performed using the infrared transmittance sequence to identify the pipe wall contamination status. When the environmental risk index at the location of the suspected particulate event is greater than 0.7, it is confirmed as a thermally induced particulate event and assigned a high-priority code. Subsequently, cross-validation of the disturbance events is performed, extracting the infrared transmittance sequence and measured hydrostatic pressure data of the corresponding node as verification signals. The transmittance change is analyzed using Bell-Lambert law to calculate the optical disturbance intensity, and the instantaneous rate of change is calculated using the pressure data. The Euclidean distance between the optical feature vector and the pressure feature vector and each typical disturbance template is calculated. When the overall similarity exceeds 0.8, the disturbance event type is confirmed. When the similarity is between 0.6 and 0.8, it is marked as suspected confirmation. Finally, a verified disturbance event data table containing information such as event ID, type code, occurrence time, and location is generated.

[0032] Step S4: Identify and track risk sources based on verified disturbance event data, and process the risk event data chain to obtain an infusion risk data chain; use an intelligent IoT sensing network to monitor disturbance events based on the infusion risk data chain to obtain a dynamic traceability map of infusion risks; manage forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

[0033] In this embodiment of the invention, disturbance type, quantification size, occurrence location, and time are extracted from verified disturbance event data to form data on the quantified risk source to be tracked. Downstream sensing nodes of the risk source are identified, and their infusion sensing characteristic data are extracted. Template matching is performed using a dynamic time warping algorithm to assess the reproducibility of upstream and downstream node deviations, classifying them as either continuously propagating events or isolated disturbance events. For continuously propagating events, an expected spatiotemporal window for downstream node responses is set. The pressure pulse response of downstream nodes is detected, and the precise spatiotemporal coordinates of the event's start and end points are extracted to calculate the propagation spatiotemporal span. During risk event data chain processing, the propagation speed is calculated based on the event's first and last nodes, and the end-point impact time is estimated. The risk propagation is verified by high-frequency monitoring of the infusion end area using a smart IoT sensing network. When the verification is true, a highlighted dynamic line is drawn on the three-dimensional pipeline model to represent the propagation path, with key points marked with different colors, forming a dynamic traceability map of infusion risk. Based on the traceability map, the hydrostatic pressure sequence on the propagation path is extracted, its trend is analyzed to determine the path type, the event probability of downstream nodes where no event has occurred is predicted, nodes with a probability exceeding 50% are marked as high-risk nodes, and a forward-looking risk warning instruction is generated and pushed to the medical terminal to realize full-process digital traceability of infusion quality.

[0034] Preferably, step S1 includes the following steps:

[0035] Step S11: After the target infusion line is installed, the infusion line nodes are identified according to the target infusion line to obtain the infusion line node sequence data;

[0036] Step S12: Based on the infusion tubing node sequence data, use a flexible guidewire with a built-in micro inertial measurement unit array to perform attitude scanning on the inside of the target infusion tubing at a sampling frequency of 50 Hz to obtain the attitude data inside the infusion tubing.

[0037] Step S13: Track and record the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal posture data of the infusion tubing, and generate three-dimensional trajectory coordinate data of the infusion path;

[0038] Step S14: Calculate a set of three-dimensional coordinate points of the infusion pipeline centerline based on the three-dimensional trajectory coordinate data of the infusion path to obtain the three-dimensional coordinate data of the pipeline centerline;

[0039] Step S15: Calculate the radius of curvature and vertical height difference between each coordinate point based on the three-dimensional coordinate data of the pipeline centerline to obtain the geometric characteristic parameters of the pipeline;

[0040] Step S16: Perform real-time spatial positioning based on the three-dimensional coordinate data of the pipeline centerline to generate environmental anchoring spatial coordinate data;

[0041] Step S17: Map and associate the pipeline geometric feature parameters with the environmental anchoring spatial coordinate data to generate infusion path spatial topology data.

[0042] In this embodiment of the invention, after the target infusion tubing is installed, a high-definition optical camera is used to capture omnidirectional images of the infusion tubing, obtaining multi-angle, high-resolution images. A pre-trained deep learning model processes the acquired images to identify key nodes on the infusion tubing, including the infusion bottle interface, drip chamber, flow regulator, three-way valve, branch connection points, and needle connection points. Pixel-level segmentation technology is used during the identification process to mark the position of each node as a coordinate point in three-dimensional space, and these coordinate points are organized into an ordered sequence according to the infusion flow direction. The output node sequence data includes the type identifier of each node, its three-dimensional spatial coordinate values ​​(X, Y, Z), and the topological connections between adjacent nodes, forming a structured infusion tubing node sequence data table.

[0043] A flexible guidewire made of medical-grade silicone, 0.5 mm in diameter and 150 cm in length, is inserted into the infusion tubing. A miniature inertial measurement unit (IMU) is embedded every 5 cm along the guidewire's length, forming an array of 30 IMUs. Each IMU contains a triaxial accelerometer, a triaxial gyroscope, and a triaxial magnetometer, connected to a data acquisition terminal via microwires inside the guidewire. The guidewire passes through the entire infusion tubing system at a constant speed of 2 cm / s, while each IMU records spatial attitude data at a frequency of 50 Hz. Each measurement point generates a data packet containing a timestamp, triaxial acceleration values, triaxial angular velocity values, and triaxial magnetic field strength values. This data is processed using a Kalman filter algorithm to eliminate noise interference, ultimately forming a continuous data stream describing the spatial attitude within the infusion tubing.

[0044] By combining timestamp information, a trajectory model of infusion equipment in a medical environment is constructed. First, a six-degree-of-freedom attitude calculation algorithm is used to fuse the acceleration, angular velocity, and magnetic field data of each measurement unit into a spatial attitude represented by Euler angles. Then, a double integration method is used to convert the acceleration data into displacement data, while zero-velocity update technology is applied to eliminate integral drift errors. By concatenating the displacement data of each measurement unit and performing spatial interpolation, a continuous trajectory from the pharmaceutical preparation area to the ward administration point is formed. The trajectory data is sampled at 1-cm intervals, with each sampling point recording the three-dimensional coordinates (X, Y, Z) in the world coordinate system, along with a timestamp and the identification information of the current tubing segment. Finally, a three-dimensional trajectory coordinate dataset of the infusion path containing 4000-6000 sampling points is generated.

[0045] Based on the 3D trajectory coordinate data of the infusion route, the centerline of the infusion tubing was extracted using a B-spline fitting algorithm. First, the original trajectory data was preprocessed to remove outliers, and the remaining data points were uniformly resampled to ensure a sample point was taken every 1 cm along the tubing length. Then, a third-order B-spline curve was constructed, and the control point positions were determined using the least squares method to best fit the resampled trajectory point set. During the fitting process, the tension parameter was set to 0.5 and the smoothing factor to 0.3 to ensure the curve both closely matched the original data and maintained smoothness. After fitting, 500-800 equidistant 3D coordinate points were extracted along the curve length. Each point consisted of X, Y, and Z components in the world coordinate system, forming a discrete 3D coordinate point sequence representing the tubing centerline.

[0046] For every three adjacent coordinate points, an arc is constructed. The radius of the circle passing through these three points is determined using Lagrange interpolation, which is the radius of curvature at that point. For each coordinate point Pi(xi, yi, zi) in the pipeline, two points Pi-1 and Pi+1 are taken before and after it, and a circle is constructed passing through these three points. The radius r is calculated using the three-point centering method. Simultaneously, the vertical height difference h = zi - z0 of each point relative to the starting point P0 is calculated. A sampling point is taken every 5 cm along the pipeline length, and the radius of curvature and height difference at that point are recorded, forming a pipeline geometric characteristic parameter table containing 100-160 records. This table includes four fields: point number, distance from the starting point, radius of curvature, and height difference.

[0047] Real-time spatial positioning is achieved by integrating a pre-deployed network of positioning base stations within the medical environment. Ten to fifteen Bluetooth Low Energy beacons are deployed throughout the medical facility, distributed across the pharmaceutical area, nurses' stations, ward corridors, and wards. Each beacon has a unique identifier and broadcasts a signal. A miniature signal receiver integrated on the guidewire receives the signal strength from multiple beacons, and the absolute position of the guidewire within the hospital is determined using trilateration. Simultaneously, visual simultaneous localization and mapping (FSMR) technology is employed. A miniature camera at the tip of the guidewire captures environmental feature points, constructing an environmental point cloud map. The Bluetooth positioning results and visual positioning results are fused using an extended Kalman filter to generate high-precision environmental anchoring spatial coordinate data, including absolute coordinates, reference beacon IDs, and environmental feature point descriptors, achieving centimeter-level spatial positioning accuracy.

[0048] Preferably, step S2, which involves building an intelligent IoT sensing network based on the infusion path spatial topology data, includes:

[0049] Based on the spatial topology data of the infusion path, the coordinates of the critical points of flow regime transition, sharp bends, and sections with vertical height changes exceeding 50 cm are marked to obtain preliminary coordinate data of key monitoring points.

[0050] Based on the preliminary key monitoring point coordinate data, the monitoring tasks are classified as follows: when the coordinate point in the preliminary key monitoring point coordinate data is a critical point of flow regime transition, it is assigned as a gas-liquid interface monitoring task; when the coordinate point in the preliminary key monitoring point coordinate data is a sharp bend section, it is assigned as a thermal flow field disturbance monitoring task; when the coordinate point in the preliminary key monitoring point coordinate data is the midpoint of a vertical height change section, it is assigned as a hydrostatic pressure calibration task.

[0051] The location configuration for the gas-liquid interface monitoring task integrates a 40 MHz high-frequency ultrasonic transducer; the location configuration for the thermal flow field disturbance monitoring task integrates a temperature sensor and an infrared transceiver pair; and the location configuration for the hydrostatic pressure calibration task integrates a piezoresistive pressure sensor and a capacitive sensor, thereby obtaining the sensing node configuration data.

[0052] Inertial measurement units are integrated based on the starting and ending points of the infusion path spatial topology data, and an IoT sensing network is built based on the sensing node configuration data to obtain an intelligent IoT sensing network.

[0053] In this embodiment of the invention, the curvature value of each point on the pipeline centerline is calculated using numerical differentiation. When the curvature exceeds 0.5 / cm, the point is marked as a sharp bend. Next, by analyzing the vertical height variation in the pipeline's geometric characteristics, when the vertical height difference between two points exceeds 50 cm, the midpoint of the line connecting the two points is marked as the midpoint of the vertical height variation section. Finally, Reynolds number abrupt change points are determined through fluid dynamics simulation. The critical points where the Reynolds number changes from laminar (less than 2000) to turbulent (greater than 2000) or vice versa are marked as flow regime transition critical points. For a standard infusion system, a total of 10-15 sharp bend points, 6-8 vertical height variation midpoints, and 4-5 flow regime transition critical points are identified, generating a structured preliminary key monitoring point coordinate data table containing point type, three-dimensional coordinate values, curvature values, height differences, and flow regime characteristic values.

[0054] Based on preliminary key monitoring point coordinate data, a task classification matrix is ​​used to define the function of each monitoring point. First, a classification judgment table is constructed, matching each coordinate point with the judgment rules. When a coordinate point is marked as a flow transition critical point (Reynolds number change rate greater than 200 / s), it is assigned to the gas-liquid interface monitoring task, responsible for detecting bubble formation and flow state during infusion. When a coordinate point is marked as a sharp bend (curvature greater than 0.5 / cm), it is assigned to the thermal flow field disturbance monitoring task, responsible for monitoring the thermal effects and flow field changes caused by the bend in the liquid flow. When a coordinate point is marked as the midpoint of a vertical height change section (height difference greater than 50 cm), it is assigned to the hydrostatic pressure calibration task, responsible for providing pressure benchmark values ​​and infusion resistance assessment. Through the judgment logic engine, a monitoring point task allocation table is generated, containing information such as point number, three-dimensional coordinates, monitoring type, monitoring parameters, and threshold settings, forming structured monitoring task classification data.

[0055] Based on the data categorized by monitoring task, sensors were configured for the IoT sensing nodes. For gas-liquid interface monitoring points, a 40 MHz high-frequency ultrasonic transducer (5 mm × 5 mm × 3 mm) was installed, with a transmission power of 10 mW, a sampling rate of 200 Hz, and a sensitivity achieving 0.01 mm bubble detection capability. For thermal flow disturbance monitoring points, an NTC thermistor temperature sensor with an accuracy of 0.01℃ and an infrared transceiver pair with a wavelength of 940 nm were configured, with a transmission power of 5 mW and a receiving sensitivity of -85 dBmW. For hydrostatic pressure calibration points, a piezoresistive pressure sensor with a measurement range of 0-300 mmHg and an accuracy of 0.1 mmHg was configured, along with a capacitive sensor with a detection range of 10-1000 picofarads for liquid dielectric constant detection. All sensors are sealed in medical-grade silicone and fixed to the outer wall of the pipeline. They are installed non-destructively using magnetic adsorption. Each sensing node is equipped with a 2.4 GHz low-power wireless transmission module and a 3.7V 100mAh lithium battery, forming a sensing node configuration data table.

[0056] Based on the spatial topology data of the infusion path, the starting point (infusion bottle interface in the medicine preparation area) and the ending point (patient intravenous injection point) are determined. An inertial measurement unit (IMU) consisting of a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer is installed at each of these two locations, with a sampling rate set to 200 Hz, an acceleration range of ±16g, and an angular velocity range of ±2000° / s. Then, according to the sensing node configuration data table, various sensors are installed at the designated monitoring points. All nodes are networked in a star topology, communicating using an IoT protocol stack. The physical layer uses 2.4 GHz ZigBee technology with a transmission rate of 250 kbps; the network layer uses IPv6 Over Low Power Wireless Personal Area Network (LPPAN) protocol; and the application layer uses Constraint Application Protocol (CAP). A gateway node is set up at the nurse station, equipped with an edge computing module, to run data preprocessing algorithms for data aggregation and initial anomaly screening. All nodes use time-division multiple access (TDMA), sending data to the gateway in turn according to a preset time slot (10 milliseconds / node), forming a complete intelligent IoT sensing network.

[0057] Preferably, step S2, which involves performing a theoretical risk characteristic analysis of the target infusion pipeline using infusion path spatial topology data, includes:

[0058] Obtain information about the infusion drug; this information includes the viscosity, density, and surface tension parameters under standard conditions at 25°C.

[0059] The temperature of the infusion environment is monitored using a smart Internet of Things (IoT) sensing network, generating infusion environment temperature data.

[0060] The average value is calculated based on the ambient temperature data of the infusion, and then the viscosity of the infusion drug information is temperature compensated to obtain the drug compensated viscosity data.

[0061] Extracting vertical height difference and curvature data of the infusion path based on spatial topology data;

[0062] Based on the information of the infusion drugs and the vertical height difference and curvature data of the path, the theoretical hydrostatic pressure and additional flow resistance of the target infusion pipeline are calculated in segments to obtain the theoretical fluid resistance distribution data.

[0063] The Reynolds number distribution within the microchannel is calculated using the preset infusion flow rate and the infusion drug information. Sections with Reynolds numbers greater than 1500 are marked as high-risk turbulence areas, and a theoretical Reynolds number distribution map is generated.

[0064] Based on the temperature data of the infusion environment, the temperature gradient of the physical space of the infusion is processed. When the temperature gradient exceeds 2 degrees Celsius, the Marangoni effect intensity risk index caused by the surface tension gradient is calculated by the drug compensation viscosity data to obtain the thermal effect risk index sequence.

[0065] The theoretical fluid resistance distribution data, theoretical Reynolds number distribution map, and thermal effect risk index sequence are serialized and aligned to generate theoretical characteristic data of fluid state.

[0066] In this embodiment of the invention, the information of the infusion medication is obtained through electronic tag identification in the medication management system. A high-frequency radio frequency identification (RFID) reader scans the medical-grade electronic tag on the infusion bottle to obtain the unique identification code of the medication. This identification code is used to access the hospital's medication database to extract the physicochemical parameters of the batch of medication under standard conditions at 25°C, including viscosity (mPa·s), density (g / cm³), and surface tension (mN / m). For compound infusion medications, pre-stored mixing parameters are used to calculate the physicochemical properties. For example, the viscosity of a 0.9% sodium chloride solution under standard conditions at 25°C is 1.005 mPa·s, the density is 1.0046 g / cm³, and the surface tension is 72.0 mN / m; while the viscosity of a 5% glucose solution is 1.081 mPa·s, the density is 1.0197 g / cm³, and the surface tension is 73.5 mN / m. The output data is stored in a structured table format, including drug name, batch number, expiration date, physicochemical parameter values, and measurement method identifier.

[0067] A smart IoT sensing network is used to monitor the infusion environment temperature in real time. Temperature sensors (NTC thermistors, accuracy 0.01℃) deployed at key nodes in the infusion line collect ambient temperature data every 10 seconds. A total of 12 temperature measurement points are set up in areas such as wards, corridors, and nurses' stations, covering the entire path of medication from preparation to infusion. Each sensor's temperature value includes three information items: measurement timestamp, sensor location code, and temperature value (degrees Celsius). The data packets are transmitted to the central monitoring system via a 2.4 GHz wireless transmission module. Upon receiving the data, the system performs time synchronization processing to ensure all data points have a unified time reference and removes outliers exceeding the reasonable range (15℃-40℃). The pre-processed temperature data is formed into a two-dimensional matrix according to time sequence and spatial location, recording the temperature changes of the entire infusion environment over 24 hours, generating detailed infusion environment temperature data.

[0068] Based on the ambient temperature data of the infusion environment, the average temperature value along the infusion path is calculated. Specifically, the temperature values ​​at each measurement point are weighted and averaged according to the acquisition time, with the weighting coefficient proportional to the dwell time at each measurement point. For a stable infusion process, the calculation formula is: T_average = ∑(Ti × Wi) / ∑Wi, where Ti is the temperature value (degrees Celsius) of the i-th measurement point, and Wi is the weighting coefficient (dimensionless) for that measurement point. After obtaining the average temperature, the Andrade equation is used to compensate for the drug viscosity at temperature: η(T) = η0 × e^(B × (1 / T - 1 / T0)), where η(T) is the viscosity value at temperature T (mPa·s), η0 is the viscosity value under standard conditions of 25℃ (mPa·s), B is the drug characteristic constant (Kelvin), T is the actual ambient average temperature (Kelvin), and T0 is the standard temperature of 298.15 Kelvin, generating the drug-compensated viscosity data at the current ambient temperature.

[0069] Based on the spatial topology data of the infusion route, geometric feature parameters of the pipeline are extracted. First, the three-dimensional coordinate sequence P(x, y, z) of the pipeline centerline is read from the topology data, with a distance of 1 cm between every two adjacent points. For each coordinate point, its vertical height difference relative to the starting point is calculated: Δh = z - z0, where z is the height coordinate of the current point (cm) and z0 is the height coordinate of the starting point (cm). Vertical height difference data is sampled and stored along the pipeline length at 5 cm intervals. Simultaneously, the curvature value of the pipeline at each point is calculated using the three-point method: κ = 1 / r, where r is the radius (cm) of the circle passing through the three consecutive points, determined by Lagrange interpolation. The curvature calculation formula is: κ = 4A / (a × b × c), where A is the area (square centimeters) of the triangle formed by the three points, and a, b, and c are the lengths of the three sides of the triangle (cm). The calculated vertical height difference data and curvature data are organized into a structured table, which includes four data items: point location number, distance from the starting point, vertical height difference value, and curvature value, forming a path vertical height difference and curvature data table.

[0070] Based on the information of the infusion drug and the data on the vertical height difference and curvature of the path, the theoretical fluid resistance distribution of the target infusion pipeline is calculated. First, the static pressure difference of each pipeline segment is calculated: ΔP_static = ρ × g × Δh, where ρ is the drug density (g / cm³), g is the acceleration due to gravity (980 cm / s²), and Δh is the vertical height difference (cm). Then, the additional flow resistance caused by pipeline curvature is calculated using the modified Hagen-Poiseuille formula: ΔP_additional = 8 × η × L × Q / (π × R) 4 )×[1+f(κ)], where η is the reagent compensation viscosity (mPa·s), L is the pipe section length (cm), Q is the flow rate (cm³ / s), R is the inner radius of the pipe (cm), and f(κ) is the curvature correction function, f(κ)=0.033×κ×D 2Where κ is the curvature value (1 / cm) and D is the inner diameter of the pipe (cm). Calculations are performed at 5 cm intervals along the length of the pipeline to generate a theoretical fluid resistance distribution data table that includes the point location, hydrostatic pressure value, and additional flow resistance value, covering the entire infusion path.

[0071] The preset infusion flow rate Q (mL / h) is obtained from the doctor's order. Combined with the infusion medication information and medication viscosity compensation data, the Reynolds number distribution within the microchannel is calculated. For a circular tube with an inner diameter of D (cm), the Reynolds number calculation formula is: Re = ρ × v × D / η, where ρ is the medication density (g / cm³), v is the average fluid velocity (cm / s), and v = 4 × Q / (π × D). 2 ×3600), where η is the reagent compensation viscosity (mPa·s). The Reynolds number is calculated every 5 cm along the length of the infusion pipeline. When the calculation results show that the Reynolds number of a certain pipeline segment is greater than 1500, that segment is marked as a high-risk turbulence zone and indicated in red; segments with Reynolds numbers between 1000 and 1500 are marked as transition zones and indicated in yellow; segments with Reynolds numbers less than 1000 are marked as laminar flow safe zones and indicated in green. The calculation results are integrated into a data table containing location number, distance from the starting point, Reynolds number, and flow regime type, and a Reynolds number distribution curve along the pipeline length is generated, forming a theoretical Reynolds number distribution map.

[0072] Based on the ambient temperature data of the infusion environment, the temperature gradient of the physical space of the infusion is calculated. The formula for calculating the temperature gradient between adjacent measurement points i and j is: ΔT / Δs=(Ti-Tj) / Lij, where Ti and Tj are the temperature values ​​(degrees Celsius) of the two points, and Lij is the spatial distance (meters) between the two points. When the detected temperature gradient exceeds 2℃ / meter, the Marangoni effect intensity risk index M caused by the surface tension gradient is calculated using the reagent compensation viscosity data. The calculation formula is: M=(dσ / dT)×(ΔT / Δs)×(D / η), where dσ / dT is the derivative of surface tension with respect to temperature (mN / m·degree Celsius), which is usually negative, approximately -0.15 mN / m·degree Celsius for water-based solutions; ΔT / Δs is the temperature gradient (degrees Celsius / meter); D is the inner diameter of the tube (meters); and η is the reagent compensation viscosity (mPa·s). When the M value is greater than 100, it is marked as a high-risk area; when the M value is between 50 and 100, it is marked as a medium-risk area; and when the M value is less than 50, it is marked as a low-risk area. The calculation results are sorted according to the pipeline location to form a sequence of thermal effect risk indices.

[0073] The theoretical fluid resistance distribution data, theoretical Reynolds number distribution map, and thermally induced effect risk index sequence were serialized and aligned. First, a unified spatial reference coordinate system was established, with the infusion inlet as the origin, and a standard sampling point set every 5 cm along the pipeline centerline. Then, spatial interpolation was performed on the three types of data to ensure that each standard sampling point had a corresponding resistance value, Reynolds number, and risk index value. Cubic spline interpolation was used to achieve a smooth data transition. The aligned data was organized into a structured table, with each row representing a spatial location point, including fields such as distance from the inlet, theoretical hydrostatic pressure, additional flow resistance, Reynolds number, flow regime type identifier, temperature gradient, and Marangoni effect risk index. A comprehensive risk score was also generated, calculated using the formula: R = 0.4 × (P / Pmax) + 0.4 × (Re / Remax) + 0.2 × (M / Mmax), where P is the theoretical fluid resistance value, Re is the Reynolds number, M is the Marangoni effect risk index, and Pmax, Remax, and Mmax are the maximum values ​​of their respective parameters. The final result is a table of theoretical characteristic data of fluid state containing 120-150 records.

[0074] Preferably, step S2, which involves acquiring multi-dimensional sensing signals for infusion based on fluid state theory characteristic data using an intelligent Internet of Things sensing network and performing time delay compensation, includes:

[0075] Based on the characteristic data analysis of fluid state theory, the high-risk area of ​​turbulence is analyzed, and the corresponding sensor nodes in the intelligent Internet of Things sensing network are set with differentiated acquisition frequencies to generate dynamic acquisition frequency configuration data. Among them, the differentiated acquisition frequency setting sets the acquisition frequency of the corresponding sensor nodes in the high-risk area of ​​turbulence to 100 Hz, and the acquisition frequency of other nodes to 20 Hz.

[0076] Based on the dynamic acquisition frequency configuration data, the intelligent IoT sensing network is dynamically monitored and configured, and the multi-dimensional sensing signal acquisition of all sensor nodes in the intelligent IoT sensing network is initiated to obtain multi-source heterogeneous real-time sensing flow; among which, the multi-source heterogeneous real-time sensing flow includes pipeline start and stop displacement sequence, measured hydrostatic pressure data, infrared light transmittance sequence and ultrasonic echo signal flow.

[0077] Digital filtering is performed on multi-source heterogeneous real-time sensing streams to generate standard multi-dimensional sensing signal data;

[0078] Compensation is performed on the time delay caused by the distance between nodes in the standard multidimensional sensing signal data, and timestamps are added to obtain spatiotemporal sensing monitoring signal data.

[0079] In this embodiment of the invention, based on theoretical fluid state characteristic data, sections with Reynolds numbers greater than 1500 in the Reynolds number distribution map are extracted as high-risk turbulence zones. A spatial mapping algorithm is used to map the coordinates in the theoretical characteristic data to sensor nodes in the intelligent IoT sensing network, accurately locating the fluid state region of each sensor node. For sensor nodes located in high-risk turbulence zones, a sampling frequency of 100 Hz is set to ensure the capture of rapidly changing turbulent pulsations; for sensor nodes located in other areas, a sampling frequency of 20 Hz is set to meet routine monitoring needs while reducing system energy consumption. A segmented configuration method is adopted, dividing the infusion pipeline into 8-12 monitoring segments, each segment assigned a corresponding sampling frequency identifier based on its risk level. Finally, a dynamic sampling frequency configuration data table is generated, containing node numbers, spatial locations, corresponding risk zone identifiers, and sampling frequency settings, serving as the basic parameters for the operation of the intelligent IoT sensing network.

[0080] Based on dynamically acquired frequency configuration data, the intelligent IoT sensing network is remotely configured via an IoT control protocol. First, the gateway node sends a frequency setting command to each sensor node, containing the node number and acquisition frequency parameters. Upon receiving the configuration, each node initiates data acquisition according to the set frequency. Inertial measurement units at the pipeline's start and end points acquire acceleration and angular velocity data at 100 Hz, generating a pipeline start-end displacement sequence. Piezoresistive pressure sensors acquire pressure data distributed along the pipeline at 20 Hz or 100 Hz (depending on the risk level of the area), forming measured hydrostatic pressure data. An infrared transceiver emits 940 nm wavelength infrared light, and the receiver measures transmittance changes; the acquisition frequency is also set according to the risk zone, generating an infrared transmittance sequence. A 40 MHz high-frequency ultrasonic transducer emits ultrasonic waves and receives echo signals; based on the echo time difference and waveform characteristics, bubbles and fluid interfaces are identified, forming an ultrasonic echo signal stream. All data is wirelessly transmitted and aggregated to a central processing unit, forming a multi-source heterogeneous real-time sensing stream.

[0081] Digital filtering was applied to the multi-source heterogeneous real-time sensing flow to eliminate noise interference and outliers. First, a Kalman filter was applied to the pipeline start-stop displacement sequence. The state transition matrix was set based on a physical motion model, and the noise covariance matrix was calibrated using historical data. After filtering, the displacement accuracy was improved to 0.1 mm. A band-stop filter was used for the measured hydrostatic pressure data, with cutoff frequencies set at 0.5-1 Hz and 15-20 Hz to filter out periodic interference caused by respiration and heartbeat. The infrared transmittance sequence was processed using a Butterworth low-pass filter with a cutoff frequency of 10 Hz and an order of 4, preserving information on liquid composition changes while filtering out ambient light fluctuations. The ultrasonic echo signal flow was denoised using wavelet transform with a db4 wavelet basis function, a decomposition level of 5, and a soft thresholding method to determine the threshold. After reconstruction, the signal-to-noise ratio was improved by 15 dB. All sensor signals were filtered and standardized to a unified data format, including four fields: sensor number, data type, numerical value, and quality identifier, forming standard multidimensional sensor signal data.

[0082] Time delay compensation is performed on standard multidimensional sensor signal data. First, the actual flow velocity v (cm / s) of the drug solution in the infusion line is measured. This is calculated by dividing the preset infusion flow rate Q (mL / h) by the cross-sectional area A (cm²) of the line: v = Q / (A×3600). Then, the time delay between adjacent sensor nodes is calculated: Δt = L / v, where L is the distance between nodes (cm). For fluid events propagating from upstream node i to downstream node j, the time series of signals acquired by the downstream node is shifted forward by Δtij to achieve time alignment. A piecewise linear interpolation algorithm is used to fill the data gaps caused by the time shift. A unified reference timestamp is added to the aligned data, accurate to the millisecond level, in the format "year-month-day hour:minute:second.millisecond". Simultaneously, the absolute acquisition time and the time offset relative to the infusion start time of each data point are recorded. After time delay compensation and timestamp annotation, a spatiotemporal sensing monitoring signal data table is generated, containing sensor number, data type, value, quality identifier, absolute timestamp, and relative time offset.

[0083] Preferably, step S3 includes the following steps:

[0084] Step S31: Extract the theoretical reference value of the corresponding position from the fluid state theoretical characteristic data based on the node position of the spatiotemporal sensing monitoring signal data, and generate a real-time-reference data pair;

[0085] Step S32: Assess the degree of deviation of the real-time-benchmark control data pair by perceptual features, and perform infusion abnormality weighting to generate infusion abnormality weight data;

[0086] Step S33: Based on the infusion anomaly weight data, intelligent judgment of disturbance events is performed through spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data to generate priority micro-disturbance event codes;

[0087] Step S34: Perform temporal clustering analysis based on priority micro-perturbation event codes to identify and extract individual perturbation events and generate independent perturbation event sample data;

[0088] Step S35: Extract the infrared transmittance sequence and measured hydrostatic pressure data of the same node when the event to be verified occurs based on the independent disturbance event sample data to obtain the disturbance verification signal data;

[0089] Step S36: Quantify the optical path disturbance intensity caused by particles or bubbles blocking the infrared light transmittance sequence in the disturbance verification signal data to generate an optical disturbance intensity value;

[0090] Step S37: Calculate the instantaneous pressure change rate of the measured hydrostatic pressure data in the disturbance verification signal data based on the average instantaneous pressure change rate of the previous second to obtain the instantaneous pressure change rate.

[0091] Step S38: Cross-validate the priority micro-perturbation event codes based on the optical perturbation intensity value and the instantaneous pressure change rate to obtain the validated perturbation event data.

[0092] In this embodiment of the invention, the node location information is first extracted. A spatial mapping algorithm is used to match the sensor node coordinates with the coordinates in the theoretical fluid state characteristic data, with the matching accuracy controlled within ±2.5 mm. For each sensor node location, three theoretical benchmark values—theoretical hydrostatic pressure, theoretical Reynolds number, and thermally induced effect risk index—are extracted from the theoretical fluid state characteristic data. Cubic spline interpolation is used to process the theoretical values ​​of non-directly corresponding points, ensuring that each monitoring point obtains an accurate theoretical reference value. The measured hydrostatic pressure data, ultrasonic echo signal flow intensity value, and infrared light transmittance value obtained from real-time monitoring are paired one-to-one with the theoretical benchmark values ​​at the corresponding locations, forming a structured data table containing node number, spatial location, timestamp, measured value, theoretical value, and difference. Each record in the table constitutes a real-time-benchmark comparison data pair, generating approximately 500-800 data pairs, covering the entire infusion process.

[0093] Evaluate the degree of deviation of perceptual features for real-time-reference comparison data pairs. First, calculate the normalized deviation value for each data pair: D = (V_real - V_theory) / V_theory, where V_real is the measured value and V_theory is the theoretical reference value. Set different deviation thresholds for different types of monitoring data: the deviation threshold for pressure data τ_p = ±5%, the deviation threshold for infrared light transmittance τ_o = ±8%, and the deviation threshold for ultrasonic echo intensity τ_u = ±12%. When |D| > τ, it is determined as abnormal, and calculate the abnormal weight W = |D| / τ. Further, assign a regional weight coefficient K to each abnormality according to the risk zones in the theoretical feature data: high turbulence risk zone K = 1.5, large vertical height change zone K = 1.2, high Marangoni effect risk zone K = 1.3, other zones K = 1.0. Finally, calculate the infusion abnormality weight as W' = W × K. Generate an infusion abnormality weight data table containing node number, timestamp, data type, deviation value, and abnormal weight. Data with an abnormal weight greater than 1.5 is marked as a high-priority abnormality and identified in red.

[0094] Based on the infusion abnormality weight data, combine the spatio-temporal perception monitoring signal data and the fluid state theoretical feature data to perform intelligent judgment of perturbation events. First, use the sliding window method to analyze the spatio-temporal perception monitoring signal data. The window length is set to 2 seconds and the step size is 0.5 seconds. Detect the signal mutation features within each window. Decompose the signal into 5 scales using wavelet transform and extract the feature coefficients. Match the extracted feature coefficients with the feature templates of six typical perturbation patterns (bubble generation, particle blockage, liquid stratification, pipeline kink, external force extrusion, flow rate fluctuation) preset, and calculate the matching degree S using cosine similarity. When S > 0.75, it is initially identified as a perturbation event of the corresponding type. Assign a priority P = S × W' to each identified perturbation event in combination with the infusion abnormality weight W'. The priority P is divided into three levels: P > 2.0 is level one (urgent), 1.0 < P ≤ 2.0 is level two (warning), and 0.5 < P ≤ 1.0 is level three (prompt). Finally, generate a priority micro-perturbation event code containing perturbation type code (1-6), occurrence time, duration, occurrence location, influence range, and priority.

[0095] Temporal clustering analysis was performed based on priority micro-perturbation event coding to extract independent perturbation events. First, an event spatiotemporal distribution matrix was constructed, with the horizontal axis representing time (accuracy in seconds) and the vertical axis representing spatial location (accuracy in centimeters). Matrix elements were assigned the priority P of the micro-perturbation event at the corresponding spatiotemporal point. The density-based spatial temporal clustering algorithm DBSCAN was used to cluster the matrix, setting the temporal neighborhood radius εt = 3 seconds, the spatial neighborhood radius εs = 10 centimeters, and the minimum number of points MinPts = 5. A class propagation factor α = 0.8 was introduced during clustering to ensure the propagation correlation of events along the infusion direction. For each cluster, its spatiotemporal boundary, core point distribution, priority mean, and variance were extracted to generate a feature description of the independent perturbation event. Clusters with a distance less than a threshold (10 centimeters and 2 seconds) were merged by calculating the shortest distance matrix between clusters. Finally, 10-15 independent perturbation events were identified, each event including a type identifier, time window, spatial range, and priority score, forming the independent perturbation event sample data.

[0096] Based on independent disturbance event sample data, a verification signal is extracted for each disturbance event. First, the core time point t0 (accurate to milliseconds) and core spatial location s0 (accurate to millimeters) of the event are determined. At the sensor nodes at the core location, the infrared transmittance sequence and measured hydrostatic pressure data within the time window [t0-2 seconds, t0+3 seconds] are extracted, with a sampling rate uniformly set to 100 Hz, generating a time series containing 500 sampling points. For cases where core location sensor data cannot be directly obtained, data from the two closest upstream and downstream sensor nodes are selected, and the signal at the core location is synthesized using a distance-weighted average method: V(s0) = [V(s1)×(s2-s0) + V(s2)×(s0-s1)] / (s2-s1), where s1 and s2 are the upstream and downstream node locations, and V(s1) and V(s2) are the corresponding sensor values. The extracted signal is subjected to zero-phase Butterworth filtering with a passband range of 0.5-20 Hz to generate smooth perturbation verification signal data, which includes three data columns: timestamp, infrared transmittance value, and hydrostatic pressure value.

[0097] The infrared transmittance sequence from the extracted perturbation verification signal data is used to quantify the optical perturbation intensity. First, the baseline transmittance T0 before the perturbation is calculated, taking the average value within 1 second before the event. Then, the minimum transmittance Tmin during the perturbation period (t0 to t0+2 seconds) is calculated. The optical perturbation depth is calculated as ΔT = T0 - Tmin, and the relative intensity of the optical perturbation is ΔT / T0. The Bell-Lambert law is used to analyze the light intensity attenuation: I = I0 × e^(-μL), where I is the transmitted light intensity, I0 is the incident light intensity, μ is the absorption coefficient, and L is the optical path length. The equivalent particle or bubble size d = ln(T0 / Tmin) / μ is obtained through inverse kinematics, where μ is the optical absorption coefficient of the drug solution (typically 0.2-0.5 / mm). The rate of change of the perturbation is calculated by combining the maximum value of the first derivative of the transmittance time curve, |dT / dt|max. The duration of the perturbation, τ, is determined by the full width at half maximum (FWHM) method. Finally, the optical perturbation intensity eigenvector F = [ΔT / T0, d, |dT / dt|max, τ] is generated.

[0098] The instantaneous rate of change of pressure was calculated from the measured hydrostatic pressure data in the extracted disturbance verification signal data. First, a reference pressure P0 was determined, and the average pressure within 1 second before the event occurred was taken. Pressure data after the event was sampled at equal intervals of 10 milliseconds. The rate of change of pressure relative to the reference pressure at each sampling point was calculated: dP / dt = (P(t) - P0) / (t - t0), where P(t) is the pressure value at time t, and t0 is the moment the event occurred. The extreme points of the rate of change of pressure were identified, and the maximum positive rate of change (dP / dt)max and the maximum negative rate of change (dP / dt)min were recorded. The characteristic frequency f of the pressure fluctuation was calculated by performing a Fast Fourier Transform on the pressure data and taking the frequency component with the largest amplitude. The duration Tp of the pressure disturbance was determined by finding the moment when the pressure recovered to within ±2% of the reference pressure.

[0099] Cross - verify the encoding of priority micro - disturbance events based on the optical disturbance intensity value and the instantaneous change rate of pressure. First, construct a feature pattern library of six types of typical disturbances. Each type of disturbance contains an optical feature vector F standard template and a pressure feature vector G standard template. Calculate the Euclidean distances between the optical feature vector and the pressure feature vector of the disturbance event to be verified and each template: DF = ||F - Fstandard||, DG = ||G - Gstandard||. The comprehensive similarity calculation is S = 1 / (1 + wF×DF+wG×DG), where wF and wG are weight coefficients, and their values range from 0.3 to 0.7 according to different disturbance types. When S > 0.8, confirm the type of the disturbance event; when 0.6 < S ≤ 0.8, mark it as "suspected confirmation"; when S ≤ 0.6, mark it as "unable to confirm". Add the verification results to the encoding of priority micro - disturbance events to form a verified disturbance event data table containing event ID, type encoding, occurrence time, location, duration, influence range, priority, verification status, and similarity.

[0100] Preferably, the intelligent judgment of disturbance events based on infusion - anomaly weight data, spatio - temporal perception monitoring signal data, and fluid - state theoretical feature data in step S33 includes:

[0101] Extract the time - domain features of the ultrasonic echo signal flow in the spatio - temporal perception monitoring signal data, calculate its transit time and attenuation coefficient, and obtain an acoustic feature parameter sequence;

[0102] Based on the infusion - anomaly weight data and the acoustic feature parameter sequence, conduct a preliminary disturbance identification. When the transit time in the acoustic feature parameter sequence has a jump greater than 1 microsecond, mark it as a suspected bubble event; when the attenuation coefficient in the acoustic feature parameter sequence continuously increases by more than 0.2 decibels, mark it as a suspected particle event, and obtain preliminary acoustic disturbance - marked event data;

[0103] According to the location where the event occurs in the preliminary acoustic disturbance - marked event data, query the Marangoni effect intensity risk index at this location from the fluid - state theoretical feature data to obtain the event environment risk index;

[0104] Analyze the liquid contact angle based on the infrared light transmittance sequence in the spatio - temporal perception monitoring signal data, and identify the local pipe - wall pollution status to obtain local pipe - wall pollution status data;

[0105] Conduct a risk - weighted judgment on the events in the preliminary acoustic disturbance - marked event data. When the event environment risk index at the location where the suspected particle event occurs is greater than 0.7, confirm the event as a heat - induced effect - induced particle event, otherwise mark it as an ordinary particle event, and obtain disturbance event data classified by cause;

[0106] The disturbance event data are coded according to their causes. Thermally induced particulate events are assigned high-priority codes, while ordinary particulate events are assigned regular codes. The causes of disturbance events are correlated based on local pipe wall contamination data to generate priority micro-disturbance event codes.

[0107] In this embodiment of the invention, time-domain features are extracted from the ultrasonic echo signal stream in the spatiotemporal sensing monitoring signal data, and real-time analysis is achieved using a high-speed digital signal processor. First, a Hilbert transform is applied to the original ultrasonic echo signal to extract the signal envelope. For the ultrasonic echo waveform at each sampling point, the time interval between the transmitted pulse and the first echo peak is calculated, which is the transit time Tt of the ultrasonic wave in the infusion pipeline, with a calculation accuracy controlled within 0.01 microseconds. Simultaneously, the attenuation coefficient α is calculated based on the ratio of the amplitudes of two consecutive echo peaks: α = -20 × lg(A2 / A1) / 2d, where A1 is the amplitude of the first echo (volts), A2 is the amplitude of the second echo (volts), and d is the sound wave propagation distance (millimeters). Ultrasonic feature parameters are extracted along the infusion pipeline at 5-centimeter intervals, forming an acoustic feature parameter sequence containing location number, timestamp, transit time value (microseconds), and attenuation coefficient value (decibels / mm), with a data update frequency of 10 Hz.

[0108] Preliminary perturbation identification is performed based on acoustic feature parameter sequences and infusion anomaly weight data. First, a transit time jump detector is constructed, and the difference in transit time within the window is calculated using a sliding window method (window length 0.5 seconds, step size 0.1 seconds): ΔTt(i) = Tt(i) - Tt(i-1). When |ΔTt(i)| > 1 microsecond is detected, it is marked as a suspected bubble event, and the timestamp, location number, and transit time jump value of the event are recorded. Simultaneously, an attenuation coefficient trend analyzer is constructed to calculate the slope of the attenuation coefficient for N = 5 consecutive sampling points: Where i is the sampling point number, and α(i) is the attenuation coefficient value at the i-th point. When k > 0.04 dB / s and the cumulative increase in attenuation coefficient over 5 consecutive samples exceeds 0.2 dB, it is marked as a suspected particulate event, and the timestamp, location number, and rate of change of attenuation coefficient of the event are recorded. The two types of event data are merged to generate a preliminary acoustic disturbance marking event data table containing event type, occurrence time, occurrence location, and acoustic characteristic change values.

[0109] Based on the preliminary acoustic disturbance marker event data, the precise location coordinates (X, Y, Z) of each event are determined. Spatial indexing techniques are used to quickly locate the corresponding location record in the fluid state theoretical characteristic data, and the Marangoni effect intensity risk index M at that location is queried. For locations that are not directly corresponding, a bilinear interpolation method is used to calculate the estimated value: M(x, y) = M(x1, y1)(x2-x)(y2-y) / (x2-x1)(y2-y1) + M(x2, y1)(x-x1)(y2-y) / (x2-x1)(y2-y1) + M(x1, y2)(x2-x)(y-y1) / (x2-x1)(y2-y1) + M(x2, y2)(x-x1)(y-y1) / (x2-x1)(y2-y1), where (x1, y1), (x2, y1), (x1, y2), and (x2, y2) are the coordinates of four known points surrounding the interpolation point. The query result is standardized to the range of 0-1: M' = M / Mmax, where Mmax is the maximum value of the Marangoni effect intensity risk index in the theoretical feature data. Assign a corresponding event environment risk index to each disturbance event, and generate a structured data table containing the event ID, location of occurrence, original value of the Marangoni effect intensity risk index, and standardized risk index value.

[0110] Based on the infrared transmittance sequence from the spatiotemporal sensing monitoring signal data, liquid contact angle analysis was performed. First, time-series data of the infrared transmittance T at each monitoring node was extracted, with a time window of 5 minutes and a sampling interval of 1 second. The contact angle calculation method based on a dynamic light intensity attenuation model was applied: θ = arccos(1 - 2 × δT / (T0 × τ)), where θ is the liquid contact angle (degrees), δT is the short-term fluctuation amplitude of transmittance, T0 is the baseline transmittance, and τ is the surface tension coefficient. When the calculated contact angle θ is greater than the preset threshold θth = 65°, it is identified as a region of enhanced hydrophobicity, indicating pipe wall contamination. Simultaneously, the long-term trend of transmittance was analyzed, and the linear regression coefficient k of the 4-hour moving average was calculated. When k < -0.005 / hour and the cumulative decrease exceeds 5%, it is marked as progressive pipe wall deposition. Based on the analysis results, each node in the pipeline was marked as a "clean zone," "initial contamination zone," or "severe contamination zone," generating local pipe wall contamination status data including node location, contamination type, contact angle value, and contamination degree index.

[0111] Risk-weighted assessments were performed on events in the preliminary acoustic disturbance marker data. First, judgment rules were established: when the environmental risk index of a suspected particulate event location is greater than 0.7, and the pipe wall contamination status at that location is "initial contamination zone" or "severe contamination zone," the event is confirmed as a thermo-induced particulate event; otherwise, it is marked as a normal particulate event. For suspected bubble events, when the hydrostatic pressure at the location is below -50 mmHg and the environmental risk index is greater than 0.5, it is marked as a negative pressure-induced bubble event; otherwise, it is marked as a regular bubble event. Automatic judgment was achieved using a decision matrix method. The matrix columns represent event types (particulates / bubbles), the columns represent risk factors (environmental risk index / pipe wall contamination status / hydrostatic pressure), and the matrix elements represent judgment thresholds. Through matrix operations, a cause type label was assigned to each disturbance event, generating a disturbance event data table containing event ID, basic type, cause type, and judgment criteria for cause classification.

[0112] The disturbance event data categorized by cause are encoded. A hierarchical encoding structure is adopted, with the encoding format "XYZ-ABCD", where X represents the basic event type (1 for bubbles, 2 for particles), Y represents the cause type (1 for conventional, 2 for thermo-induced, 3 for negative pressure induced), and Z represents the severity (1-5, 5 being the most severe); ABCD represents the event timestamp encoding, accurate to the minute. The severity Z is calculated based on the theoretical risk index of the disturbance signal strength, duration, and location: Z = ceil(0.4×S + 0.3×D + 0.3×R), where S is the normalized value of the signal strength, D is the normalized value of the duration, R is the theoretical risk index, and ceil is the floor function. A priority coefficient P = 1.5 is assigned to thermo-induced particle events, a priority coefficient P = 1.3 is assigned to negative pressure induced bubble events, and a priority coefficient P = 1.0 is assigned to conventional events. Finally, an event coding table is generated, which includes information such as event ID, event code, priority coefficient, associated pollution status, and warning level, forming a complete priority micro-perturbation event coding dataset.

[0113] Preferably, step S4, which involves identifying and tracking risk sources based on verified disturbance event data and performing risk event data chain processing, includes:

[0114] Based on the verified disturbance event data, the disturbance event type, quantification size, occurrence location and occurrence time are extracted to obtain the data of the quantified risk sources to be tracked;

[0115] Identify downstream sensing nodes based on the location of occurrence in the data of the quantified risk sources to be tracked, and extract downstream infusion sensing feature data;

[0116] Based on downstream infusion sensing characteristic data and data on risk sources to be tracked and quantified, spatiotemporal deviations are correlated to assess whether deviations are reproduced at upstream and downstream nodes, and abnormal propagation event type data is generated; among them, abnormal propagation event type data includes isolated disturbance events and continuous propagation events.

[0117] When the abnormal propagation event type data is a continuous propagation event, the downstream propagation path data is determined based on the infusion path spatial topology data; when the abnormal propagation event type data is an isolated disturbance event, the data of the risk source to be tracked and quantified is sent to the terminal device for potential risk warning and event removal processing.

[0118] Based on downstream propagation path data, the verified disturbance event data is processed into a risk event data chain to obtain the infusion risk data chain.

[0119] In this embodiment of the invention, a structured data extraction algorithm is used to analyze the risk source characteristics based on verified disturbance event data. First, basic information fields for each event are extracted from the verified disturbance event data table, including a unique event identifier, disturbance type code (bubbles / particles / liquid stratification / pipeline kinking / external force compression / flow velocity fluctuation), occurrence timestamp (accurate to millimeters), and occurrence spatial coordinates (X, Y, Z, accurate to millimeters). For bubble-type events, the equivalent diameter of the bubble is calculated using acoustic impedance analysis of the ultrasonic echo signal: d = v × Δt / 2, where v is the propagation speed of ultrasound in the liquid (1480 m / s), and Δt is the echo time jump value (microseconds). For particle-type events, the particle agglomeration size is calculated using optical attenuation coefficient inversion: L = -ln(T / T0) / α, where T is the transmittance at the disturbance peak, T0 is the baseline transmittance, and α is the light absorption coefficient of the liquid (0.2-0.5 / mm). Each risk source is assigned a unique code WF-XXXXX-YY, where XXXXX is the time sequence number and YY is the risk source type code, forming a data table of quantifiable risk sources to be tracked that contains basic information and quantitative characteristic parameters of the risk source.

[0120] Based on the data of the quantifiable risk sources to be tracked, downstream sensing nodes are identified by combining the spatial topology data of the infusion path. First, the precise location of the risk source in the infusion pipeline is determined, and a downstream node query is constructed from this starting point. A depth-first search algorithm is used to traverse the pipeline topology map, with the search range set to all sensing nodes within 50 cm downstream of the risk source location. For pipeline branch areas, all paths are recursively traversed to identify all downstream nodes. For each identified downstream node, its location coordinates, pipeline distance L (cm) from the risk source, and estimated arrival time t = L / v (seconds, where v is the flow velocity in cm / s) are recorded. Multidimensional sensing data of these downstream nodes within the time window [t0+t-2, t0+t+5] are extracted from the spatiotemporal sensing monitoring signal data, where t0 is the time of risk source occurrence. The extracted downstream node sensing data includes infrared light transmittance sequences, measured hydrostatic pressure data, and ultrasonic echo signal characteristics, with a sampling frequency uniformly set to 50 Hz, forming a downstream infusion sensing feature dataset.

[0121] Spatiotemporal deviation correlation analysis was performed based on downstream infusion sensing characteristic data and data of risk sources to be tracked and quantified. First, a risk source feature template was constructed, including typical infrared light transmittance waveforms, pressure disturbance characteristic curves, and ultrasonic characteristic parameters. For the sensing data sequence of each downstream node, a dynamic time warping algorithm was used for template matching, calculating a similarity score S = [1 - ∑|x1(i) - x2(i)| / n] / ∑|x1(i)| / n, where x1 is the template sequence, x2 is the measured sequence, and n is the sequence length. A similarity threshold Sth = 0.75 was set; when the similarity S between the downstream node's feature data and the risk source feature template is greater than Sth, it is determined to be a disturbance deviation recurrence. Based on the recurrence rate, events were classified as follows: when more than 50% of downstream nodes detected deviation recurrence, it was marked as a continuous propagation event; when the recurrence rate was less than 50% or only the nearest 1-2 downstream nodes detected a brief recurrence followed by disappearance, it was marked as an isolated disturbance event. An abnormal propagation event type data table was generated, containing the risk source ID, a list of downstream nodes, the recurrence status of each node, and the event propagation type.

[0122] Different processing methods are applied based on the types of abnormal propagation events. When the abnormal propagation event type is a continuous propagation event, an accurate propagation trajectory prediction model is established based on the spatial topology data of the infusion path. A particle filtering algorithm is used to initialize 100 particles, representing the location distribution of the risk source in the pipeline. Each particle contains three state variables: position, propagation velocity, and disturbance intensity. The motion trajectory of the particles along the pipeline is predicted using a fluid dynamics model: X(t+Δt)=X(t)+v(t)×Δt+0.5×a×Δt 2Where X(t) is the position at time t, v(t) is the flow velocity, and a is the acceleration (considering the effects of pipe bending and gravity). The particle weights are updated based on the actual detected downstream node data, retaining the top 50% of particles, and resampling to generate a new particle set. After 10 iterations, the distribution density of the particle set forms the downstream propagation path data of the risk source, including the coordinates of key points along the path, the estimated arrival time, and the predicted value of disturbance intensity attenuation. When the abnormal propagation event type is an isolated disturbance event, a risk warning message is sent through the medical staff's terminal device, including the risk source type, location, occurrence time, and impact assessment, and the event is marked as "processed" and removed from the activity monitoring queue.

[0123] Based on downstream propagation path data, a risk event data chain processing method is applied to verified disturbance event data. First, an event chain data structure is constructed, including a head node (original risk source), body nodes (sensing nodes along the propagation path), and tail node (final arrival or dissipation point). A time-series graph database is used to store the event chain, where nodes represent disturbance events, edges represent propagation relationships, and edge attributes include propagation delay time, propagation distance, and signal attenuation rate. For each continuous propagation event, all observation data along the propagation path are extracted, and the disturbance intensity attenuation curve is calculated: I(s) = I0 × e^(-βs), where I(s) is the disturbance intensity at a distance s from the head node, I0 is the original intensity, and β is the attenuation coefficient. Simultaneously, the changing trends of disturbance characteristic parameters (such as bubble size and particle concentration) along the path are recorded. Event chain nodes are linked through causal relationship identifiers to ensure data integrity and traceability. Finally, a structured infusion risk data chain is formed, containing basic information about the risk source, propagation trajectory, impact range, duration, and intensity change curve, supporting risk tracing and early warning.

[0124] Of particular importance is the risk event data chain processing of verified disturbance event data based on downstream propagation path data, which includes:

[0125] Based on the downstream propagation path data, an expected spatiotemporal window for the response of a downstream node is set, thus obtaining the expected spatiotemporal window for the downstream event.

[0126] By using verified disturbance event data, detect whether a positive pulse higher than 20 Pascals appears in the pressure deviation of the downstream adjacent nodes within the expected spatiotemporal window of the downstream event. If detected, generate a matched downstream positive pressure response point.

[0127] Extract the spatiotemporal coordinates of the event's starting point from the data of the risk sources to be tracked and quantified;

[0128] Extract the three-dimensional spatial coordinates of the matched downstream positive pressure response points and record them as the spatiotemporal coordinates of the event endpoint;

[0129] Calculate the spatiotemporal coordinates of the event's origin and the linear spatial distance and time difference between the event's origin spatiotemporal coordinates to obtain the spatiotemporal span of the event propagation.

[0130] Physical effectiveness correlation analysis is performed based on the spatiotemporal span of event propagation, and spatiotemporal sensing monitoring signal data is processed into a data chain to generate an infusion risk data chain.

[0131] In this embodiment of the invention, based on downstream propagation path data, a fluid dynamics propagation model is applied to set the expected spatiotemporal window for downstream node responses. First, the precise pipeline path length L (mm) from the risk source to each downstream node is calculated, rather than the linear distance in space. Then, the theoretical arrival time ttheoretical = L / v (seconds) is calculated based on the infusion flow rate v (mm / s). Considering the uneven flow rate caused by drug viscosity, a time window widening coefficient κ = 1.5 is set, constructing a time window as [ttheoretical - κ × τ, ttheoretical + 2 × κ × τ], where τ is the characteristic time of infusion flow rate fluctuation (usually taken as 5 seconds). Simultaneously, a spatial detection window is set within a range of ±25 mm from the node location. The time window and spatial window are combined to form a four-dimensional spatiotemporal detection window, generating a structured downstream event expected spatiotemporal window data table, containing information such as node ID, upper and lower limits of the spatial range, and the start and end points of the time window.

[0132] Based on verified disturbance event data, pressure anomaly detection is performed within the expected spatiotemporal window of downstream events. Measured hydrostatic pressure data from downstream nodes are extracted from the spatiotemporal sensing monitoring signal data, with a sampling frequency of 100 Hz. First, the average pressure Pbaseline (Pascals) for the first 10 seconds of the expected time window is calculated as the baseline value. Then, the peak value of the pressure signal is searched within the spatiotemporal window: Ppeak = max(P(t)), t∈[ttheoretical - κ×τ, ttheoretical + 2×κ×τ]. The pressure pulse amplitude ΔP is calculated as Ppeak - Pbaseline (Pascals). When a positive pulse with ΔP > 20 Pascals is detected, the peak time tpeak, duration Δt (time required for pressure to recover to the baseline value + 5 Pascals), and location coordinates are recorded. Downstream nodes meeting the criteria are marked as matched downstream positive pressure response points, generating a matching point data table containing node ID, pressure pulse amplitude, peak time, and duration.

[0133] Based on the data of the risk sources to be tracked and quantified, the precise spatiotemporal coordinates of the starting point of the risk event are extracted. The spatial location information of the risk source is read from the data table, including the X coordinate (mm, horizontal position), Y coordinate (mm, vertical position), and Z coordinate (mm, vertical height), forming a three-dimensional spatial coordinate (X0, Y0, Z0). Simultaneously, the precise occurrence time t0 of the risk source is extracted, accurate to the millisecond level, in the format "year-month-day hour:minute:second.millisecond". The precise location information of this coordinate point on the infusion pipeline is confirmed through the spatial topology data of the infusion path, including the pipeline length S0 (mm) from the infusion starting point, the pipeline segment number N0, and the local pipeline curvature κ0 (1 / mm). This information is integrated into a spatiotemporal coordinate data structure for the event starting point, containing complete location information in three dimensions: absolute time, three-dimensional spatial coordinates, and relative pipeline position.

[0134] Based on the matched downstream positive pressure response point data table, the precise three-dimensional spatial positioning information of the downstream response points is extracted. First, the sensor node number of the matched point is read, and the precise three-dimensional coordinates (X1, Y1, Z1) of its installation location are retrieved from the sensor deployment configuration table, in millimeters. The relative position of this coordinate point on the infusion pipeline is confirmed through the infusion path spatial topology data, including the pipeline length S1 (mm) from the infusion start point, the pipeline segment number N1, and the local pipeline curvature κ1 (1 / mm). Simultaneously, the precise occurrence time t1 of the pressure pulse peak is extracted, accurate to the millisecond level. The above information is integrated to construct the event endpoint spatiotemporal coordinate data structure, including absolute time t1, three-dimensional spatial coordinates (X1, Y1, Z1), and pipeline relative position parameters, forming complete event endpoint spatiotemporal coordinate information.

[0135] Based on the spatiotemporal coordinates of the event's origin and destination, the spatiotemporal span of the event propagation is calculated. First, the straight-line distance in three-dimensional space is calculated. (mm). Then calculate the pipeline friction distance L = |S1 - S0| (mm), where L represents the actual propagation distance measured along the centerline of the infusion pipeline. Calculate the time difference Δt = t1 - t0 (seconds), representing the time interval from the occurrence of the risk source to the detection of a pressure response at the downstream node. Based on these parameters, calculate the actual propagation velocity vactual = L / Δt (mm / second). Integrate the calculation results into a spatiotemporal span data structure for event propagation, including four key parameters: spatial straight-line distance, pipeline friction distance, time difference, and actual propagation velocity.

[0136] Based on the spatiotemporal span data of event propagation, a physical validity correlation analysis is performed. First, the theoretical propagation speed vtheoretical = Q / (π×rprincip) is calculated. 2(mm / s), where Q is the infusion flow rate (m³ / s) and r is the tubing inner diameter (mm). The deviation rate η = |vactual - vtheoretical| / vtheoretical is compared between the actual propagation velocity vactual and the theoretical propagation velocity vtheoretical. When η < 0.3, it is considered a physically valid association; when η ≥ 0.3, it is considered a physically invalid association and a warning is triggered. For physically valid event pairs, all spatiotemporal sensing monitoring signal data between the risk source and the response point are extracted, including the complete data stream of all sensor nodes between the two points within the time window [t0-5, t1+5]. These data are linked in chronological order, with each data block containing a node ID, timestamp, multidimensional sensor value, and forward hash value. Finally, an infusion risk data chain containing complete information on the event start point, propagation path, and event endpoint is generated, supporting full-process traceability and risk assessment.

[0137] Preferably, step S4, which involves monitoring disturbance events using a smart IoT sensing network based on the infusion risk data chain, includes:

[0138] The propagation speed is calculated based on the position and time difference of the first and last nodes of the infusion risk data chain to obtain the abnormal event movement speed value;

[0139] Calculate the estimated time of impact on the infusion terminal based on the movement speed value of the abnormal event;

[0140] Based on the estimated time of impact at the end of infusion, the data of verified disturbance events are monitored through a smart IoT sensing network to obtain risk propagation verification data.

[0141] When the risk propagation verification data is true, a highlighted dynamic line is drawn on the visualization interface of the infusion path spatial topology data to represent the propagation path, thus obtaining the dynamic traceability map of infusion risk.

[0142] In this embodiment of the invention, the position and time information of the first node (the starting point of the risk source) and the last node (the last matched downstream positive pressure response point) are extracted based on the infusion risk data chain. First, the actual pipeline length L (millimeters) between the first and last nodes is calculated. Then, using the sequence of three-dimensional coordinate points of the pipeline centerline in the infusion path spatial topology data, a segmented accumulation method is employed. Where (xi, yi, zi) is the i-th coordinate point on the centerline of the pipeline. Then, the time difference Δt = t_end - t_begin (seconds) between the first and last nodes is calculated, accurate to the millisecond level. Applying the fluid dynamics propagation velocity calculation formula v = L / Δt (mm / s), the velocity of the abnormal event in the actual pipeline is obtained. The calculated result is compared with the theoretical infusion velocity v_theoretical = Q / (π×r). 2 A comparison is performed, where Q is the infusion flow rate (cubic millimeters / second), r is the inner diameter of the pipeline (millimeters), and the generation rate ratio η = v / v is used to assess the propagation anomaly.

[0143] Based on the abnormal event's movement speed value v (mm / s), and combined with the infusion path spatial topology data, the estimated time for the risk event to reach the infusion end (patient injection point) is calculated. First, the precise tubing length Lremaining (mm) from the last matched downstream positive pressure response point to the infusion end is determined, using the same method of segmented accumulation of the three-dimensional coordinate point sequence along the tubing centerline. The estimated remaining propagation time tremaining = Lremaining / v (seconds) is calculated. Considering the influence of tubing curvature and height changes on speed, a correction factor λ is introduced: λ = 1 + 0.02 × κaverage + 0.01 × |Δh| / 100, where κaverage is the average curvature of the remaining path (1 / mm), and Δh is the vertical height change of the remaining path (mm). Finally, the estimated time of impact at the infusion end testimated = tend + tremaining × λ (seconds) is converted to an absolute time format "year-month-day hour:minute:second.millisecond" to accurately record the estimated time of the risk event's arrival at the infusion end.

[0144] Based on the estimated time t for the impact of the infusion terminal, a smart IoT sensing network is used to perform targeted monitoring of the infusion terminal area. All sensor nodes within a 5 cm radius of the infusion terminal are set to high-frequency monitoring mode, with the sampling frequency increased to 200 Hz to ensure the capture of transient events. The monitoring time window is set to [t-estimated - 10 seconds, t-estimated + 20 seconds]. Monitoring data includes pressure fluctuations, changes in optical transmittance, and ultrasonic echo characteristics, which are matched in real time with feature templates in the verified disturbance event data to calculate a similarity score S. A verification threshold Sth = 0.7 is set. If any sensor node detects a signal feature with a similarity S > Sth, it is recorded as a successful detection; if no node detects a feature that meets the threshold within the specified time window, it is recorded as a failed detection. Risk propagation verification data containing monitoring results, detection time, highest similarity value, and trigger node ID is generated to confirm whether the risk has actually propagated to the infusion terminal.

[0145] When the verification result of the risk propagation verification data is true, dynamic risk path plotting is performed on the infusion path spatial topology data visualization interface of the medical monitoring terminal. First, a three-dimensional pipeline model is constructed based on the infusion path spatial topology data, using an RGBA color model, with the default pipeline color being a semi-transparent light blue (128, 200, 255, 120). According to the position of each node in the infusion risk data chain, key points are marked on the 3D model, including the risk source starting point (red solid circle), intermediate propagation nodes (orange hollow circle), and the terminal arrival point (red and yellow alternating flashing circles). A 3-pixel wide solid line is constructed connecting all key points, with the color set to a gradient red, gradually changing from the risk source starting point (255, 0, 0, 255) to the terminal arrival point (255, 255, 0, 255). Animation effects are applied to simulate the risk propagation process, with the animation rate controlled by the actual calculated propagation speed v, updating the position every 100 milliseconds along the path. At the same time, an event information panel is generated on the right side of the interface, displaying key parameters such as risk type, occurrence time, spread speed, and scope of impact, forming a complete dynamic traceability map of infusion risk.

[0146] Of particular importance is the processing of forward-looking risk warning instructions based on the dynamic traceability map of infusion risks, including:

[0147] Based on the confirmed propagation path in the dynamic traceability map of infusion risk, the vertical height difference of the infusion system of all nodes on the path is extracted to generate the hydrostatic pressure change value, and the hydrostatic pressure sequence of the propagation path is obtained.

[0148] Trend analysis of the hydrostatic pressure sequence of the propagation path is performed. If the hydrostatic pressure shows a continuous downward trend, the path is determined to be the bubble acceleration section. If it shows a continuous upward trend, it is the bubble stagnation section. The flow evolution characteristics of the propagation path are obtained.

[0149] Based on the propagation path flow evolution characteristics, the probability of an event occurring is predicted for downstream nodes that have not yet occurred on the path. If it is a bubble acceleration section, the probability of an event occurring in the downstream node increases by 20%, and the event probability prediction data of the downstream node is obtained.

[0150] Nodes with a probability exceeding 50% in the downstream node event probability prediction data are marked to generate a list of high-risk propagation nodes;

[0151] The node information in the high-risk transmission node list is pushed to terminal devices for early warning, and is set as the highest monitoring priority to obtain forward-looking risk warning instructions.

[0152] In this embodiment of the invention, based on the dynamic traceability map of infusion risk, the vertical height data of all nodes on the confirmed propagation path are extracted. First, the Z-axis coordinate value zi (mm) of each node on the propagation path is obtained from the spatial topology data of the infusion path, with a node interval of 5 cm. The height difference Δhi = zi - zi-1 (mm) between adjacent nodes is calculated. According to the principle of hydrostatics, the hydrostatic pressure change generated by each height difference is calculated: ΔPi = ρ × g × Δhi / 1000 (Pascals), where ρ is the drug density (kg / m³), ranging from 1000 to 1050, and g is the acceleration due to gravity (9.8 m / s²). All ΔPi values ​​are arranged sequentially along the propagation path to form a propagation path hydrostatic pressure sequence table containing node number, location coordinates, height value, height difference, and hydrostatic pressure change value, containing a total of 15-25 data points, completely recording the hydrostatic pressure distribution on the risk propagation path.

[0153] Trend analysis was performed on the hydrostatic pressure sequence along the propagation path. First, a linear trend line was fitted to the hydrostatic pressure data points using the least squares method: P(s) = k × s + b, where P is the hydrostatic pressure (Pascals), s is the distance along the pipe (cm), k is the pressure gradient (Pascals / cm), and b is the intercept (Pascals). The sign and magnitude of the pressure gradient k were calculated. When k < -0.2 Pascals / cm and the linear correlation coefficient R0 is within the acceptable range... 2 When k > 0.7, it is determined to be a continuous downward trend and marked as the bubble acceleration phase; when k > 0.2 Pascals / cm and R 2 When |k| > 0.7, it is determined to be a continuous upward trend and marked as a bubble stagnation segment; when |k| ≤ 0.2 Pascals / cm or R 2 When the pressure gradient is ≤0.7, it is marked as a neutral propagation section. Simultaneously, the curvature distribution of the path is calculated to identify sharp bends with curvature greater than 0.05 / cm. The trend analysis results are combined with pipeline geometric characteristics to generate a propagation path flow evolution characteristic table containing path type, pressure gradient, correlation coefficient, and key bend locations.

[0154] Based on the flow evolution characteristics of the propagation path, risk prediction is performed on downstream nodes along the path where no events have yet occurred. First, the original event type (bubble / particle) and its baseline probability P0 (typically set at 30%-40%) are extracted from verified disturbance event data. Then, the probability is adjusted according to the propagation path type: when the path is determined to be a bubble acceleration section, the probability of bubble-type events is calculated as Pi = P0 × (1 + 0.2 × D / Dmax), where D is the distance from the node to the path start point (cm), and Dmax is the total path length (cm); when the path is determined to be a bubble stagnation section, the probability of bubble-type events is calculated as Pi = P0 × (1 - 0.1 × D / Dmax); when the path is determined to be a neutral propagation section, the event probability remains at P0. For nodes within 10 cm downstream of sharp bends, an additional 10% probability is added. Taking into account various factors, a downstream node event probability prediction data table is generated, containing node ID, location, baseline probability, and adjusted probability.

[0155] Downstream node event probability prediction data is filtered and labeled. First, a high-risk probability threshold Pth = 50% is set, and the adjusted probability Pi value of all downstream nodes is scanned. When Pi > Pth, the node is marked as a high-risk node, indicated by a red pentagram. For multiple adjacent high-risk nodes, their spatial distribution density ρ = n / L (nodes / cm) is calculated, where n is the number of high-risk nodes and L is the pipeline length (cm) covered by these nodes. When ρ > 0.2 nodes / cm, this area is marked as a high-risk concentration area, highlighted with red shading. A unique risk code HR-XXXXX-Y is assigned to each high-risk node, where XXXXX is the node location code and Y is the risk type code (1 for bubble risk, 2 for particulate risk). All high-risk node information is integrated into a high-risk propagation node list, containing five key fields: node ID, spatial location, risk type, probability of occurrence, and expected occurrence time.

[0156] The list of high-risk transmission nodes is transformed into a proactive risk warning instruction. First, based on the spatial location of each high-risk node and the expected time of risk occurrence, a warning information package is constructed. The package structure is: {Risk Code, Risk Type, Location Description, Expected Time, Severity Level, Recommended Measures}. For bubble risk, the severity level is graded according to the predicted bubble size D (mm): D < 0.5 is Level 3, 0.5 ≤ D < 2 is Level 2, and D ≥ 2 is Level 1. For particulate risk, it is graded according to the predicted particulate aggregation degree C (dimensionless): C < 5 is Level 3, 5 ≤ C < 10 is Level 2, and C ≥ 10 is Level 1. An XML-formatted warning instruction file containing warning information for all high-risk nodes is constructed and pushed to nurse stations and doctor workstations via the hospital's intranet. Simultaneously, the warning information is pushed to medical staff's mobile terminals via the MQTT protocol, with a push priority set to 9 (out of 10). In the system monitoring interface, high-risk nodes are set to the highest monitoring priority, the sampling frequency is increased to 500 Hz, and sensor data waveforms are displayed in real time, forming a complete proactive risk warning instruction system.

[0157] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0158] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A digital traceability method for infusion quality and safety based on the Internet of Things, characterized in that, Includes the following steps: Step S1: Perform internal attitude scanning on the target infusion line to obtain internal attitude data of the infusion line; track the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal attitude data of the infusion line, and perform pipeline geometric feature analysis to generate infusion path spatial topology data; Step S2: Construct an intelligent IoT sensing network based on the infusion path spatial topology data; perform theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data to obtain theoretical fluid state characteristic data; wherein, the theoretical risk characteristic analysis of the target infusion pipeline using the infusion path spatial topology data in step S2 includes: Obtain information about the infusion drug; this information includes the viscosity, density, and surface tension parameters under standard conditions at 25°C. The temperature of the infusion environment is monitored using a smart Internet of Things (IoT) sensing network, generating infusion environment temperature data. The average value is calculated based on the ambient temperature data of the infusion, and then the viscosity of the infusion drug information is temperature compensated to obtain the drug compensated viscosity data. Extracting vertical height difference and curvature data of the infusion path based on spatial topology data; Based on the information of the infusion drugs and the vertical height difference and curvature data of the path, the theoretical hydrostatic pressure and additional flow resistance of the target infusion pipeline are calculated in segments to obtain the theoretical fluid resistance distribution data. The Reynolds number distribution within the microchannel is calculated using the preset infusion flow rate and the infusion drug information. Sections with Reynolds numbers greater than 1500 are marked as high-risk turbulence areas, and a theoretical Reynolds number distribution map is generated. Based on the temperature data of the infusion environment, the temperature gradient of the physical space of the infusion is processed. When the temperature gradient exceeds 2 degrees Celsius, the Marangoni effect intensity risk index caused by the surface tension gradient is calculated by the drug compensation viscosity data to obtain the thermal effect risk index sequence. The theoretical fluid resistance distribution data, theoretical Reynolds number distribution map, and thermally induced effect risk index sequence are serialized and aligned to generate theoretical characteristic data of fluid state. Based on fluid state theory characteristic data, multi-dimensional infusion sensing signals are acquired using an intelligent IoT sensing network, and time delay compensation is performed to obtain spatiotemporal sensing monitoring signal data; wherein, step S2, which involves acquiring multi-dimensional infusion sensing signals based on fluid state theory characteristic data using an intelligent IoT sensing network and performing time delay compensation, includes: Based on the characteristic data analysis of fluid state theory, the high-risk area of ​​turbulence is analyzed, and the corresponding sensor nodes in the intelligent Internet of Things sensing network are set with differentiated acquisition frequencies to generate dynamic acquisition frequency configuration data. Among them, the differentiated acquisition frequency setting sets the acquisition frequency of the corresponding sensor nodes in the high-risk area of ​​turbulence to 100 Hz, and the acquisition frequency of other nodes to 20 Hz. Based on the dynamic acquisition frequency configuration data, the intelligent IoT sensing network is dynamically monitored and configured, and the multi-dimensional sensing signal acquisition of all sensor nodes in the intelligent IoT sensing network is initiated to obtain multi-source heterogeneous real-time sensing flow; among which, the multi-source heterogeneous real-time sensing flow includes pipeline start and stop displacement sequence, measured hydrostatic pressure data, infrared light transmittance sequence and ultrasonic echo signal flow. Digital filtering is performed on multi-source heterogeneous real-time sensing streams to generate standard multi-dimensional sensing signal data; Compensation is performed on the data time delay caused by the distance between nodes in the standard multidimensional sensing signal data, and timestamps are added to obtain spatiotemporal sensing monitoring signal data. Step S3: Intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and perform cross-validation of disturbance events to obtain verified disturbance event data; Step S4: Identify and track risk sources based on verified disturbance event data, and process the risk event data chain to obtain an infusion risk data chain; use an intelligent IoT sensing network to monitor disturbance events based on the infusion risk data chain to obtain a dynamic traceability map of infusion risks; manage forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

2. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: After the target infusion line is installed, the infusion line nodes are identified according to the target infusion line to obtain the infusion line node sequence data; Step S12: Based on the infusion tubing node sequence data, use a flexible guidewire with a built-in micro inertial measurement unit array to perform attitude scanning on the inside of the target infusion tubing at a sampling frequency of 50 Hz to obtain the attitude data inside the infusion tubing. Step S13: Track and record the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point based on the internal posture data of the infusion tubing, and generate three-dimensional trajectory coordinate data of the infusion path; Step S14: Calculate a set of three-dimensional coordinate points of the infusion pipeline centerline based on the three-dimensional trajectory coordinate data of the infusion path to obtain the three-dimensional coordinate data of the pipeline centerline; Step S15: Calculate the radius of curvature and vertical height difference between each coordinate point based on the three-dimensional coordinate data of the pipeline centerline to obtain the geometric characteristic parameters of the pipeline; Step S16: Perform real-time spatial positioning based on the three-dimensional coordinate data of the pipeline centerline to generate environmental anchoring spatial coordinate data; Step S17: Map and associate the pipeline geometric feature parameters with the environmental anchoring spatial coordinate data to generate infusion path spatial topology data.

3. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 1, characterized in that, Step S2, which involves building a smart IoT sensing network based on the infusion path spatial topology data, includes: Based on the spatial topology data of the infusion path, the coordinates of the critical points of flow regime transition, sharp bends, and sections with vertical height changes exceeding 50 cm are marked to obtain preliminary coordinate data of key monitoring points. Based on the preliminary key monitoring point coordinate data, the monitoring tasks are classified as follows: when the coordinate point in the preliminary key monitoring point coordinate data is a critical point of flow regime transition, it is assigned as a gas-liquid interface monitoring task; when the coordinate point in the preliminary key monitoring point coordinate data is a sharp bend section, it is assigned as a thermal flow field disturbance monitoring task; when the coordinate point in the preliminary key monitoring point coordinate data is the midpoint of a vertical height change section, it is assigned as a hydrostatic pressure calibration task. The location configuration for the gas-liquid interface monitoring task integrates a 40 MHz high-frequency ultrasonic transducer; the location configuration for the thermal flow field disturbance monitoring task integrates a temperature sensor and an infrared transceiver pair; and the location configuration for the hydrostatic pressure calibration task integrates a piezoresistive pressure sensor and a capacitive sensor, thereby obtaining the sensing node configuration data. Inertial measurement units are integrated based on the starting and ending points of the infusion path spatial topology data, and an IoT sensing network is built based on the sensing node configuration data to obtain an intelligent IoT sensing network.

4. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Extract the theoretical reference value of the corresponding position from the fluid state theoretical characteristic data based on the node position of the spatiotemporal sensing monitoring signal data, and generate a real-time-reference data pair; Step S32: Assess the degree of deviation of the real-time-benchmark control data pair by perceptual features, and perform infusion abnormality weighting to generate infusion abnormality weight data; Step S33: Based on the infusion anomaly weight data, intelligent judgment of disturbance events is performed through spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data to generate priority micro-disturbance event codes; Step S34: Perform temporal clustering analysis based on priority micro-perturbation event codes to identify and extract individual perturbation events and generate independent perturbation event sample data; Step S35: Extract the infrared transmittance sequence and measured hydrostatic pressure data of the same node when the event to be verified occurs based on the independent disturbance event sample data to obtain the disturbance verification signal data; Step S36: Quantify the optical path disturbance intensity caused by particles or bubbles blocking the infrared light transmittance sequence in the disturbance verification signal data to generate an optical disturbance intensity value; Step S37: Calculate the instantaneous pressure change rate of the measured hydrostatic pressure data in the disturbance verification signal data based on the average instantaneous pressure change rate of the previous second to obtain the instantaneous pressure change rate. Step S38: Cross-validate the priority micro-perturbation event codes based on the optical perturbation intensity value and the instantaneous pressure change rate to obtain the validated perturbation event data.

5. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 4, characterized in that, Step S33, which involves intelligently judging disturbance events based on infusion anomaly weight data through spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, includes: Time-domain features are extracted from the ultrasonic echo signal stream in the spatiotemporal sensing monitoring signal data, and its transit time and attenuation coefficient are calculated to obtain the acoustic feature parameter sequence. Preliminary disturbance identification is performed based on infusion abnormality weight data and acoustic feature parameter sequences. When the transit time in the acoustic feature parameter sequence changes by more than 1 microsecond, it is marked as a suspected bubble event. When the attenuation coefficient in the acoustic feature parameter sequence increases continuously by more than 0.2 dB, it is marked as a suspected particle event, thus obtaining preliminary acoustic disturbance marking event data. Based on the location of the event in the preliminary acoustic disturbance marker event data, the Marangoni effect intensity risk index at that location is queried from the fluid state theory characteristic data to obtain the event environmental risk index. Based on the infrared light transmittance sequence in the spatiotemporal sensing monitoring signal data, liquid contact angle analysis was performed, and pipe wall contamination status was identified to obtain local pipe wall contamination status data. Risk-weighted judgment is performed on the events in the preliminary acoustic disturbance marker event data. If the event environmental risk index of the suspected particulate event location is greater than 0.7, the event is confirmed as a thermally induced particulate event; otherwise, it is marked as an ordinary particulate event, thus obtaining disturbance event data classified by cause. The disturbance event data are coded according to their causes. Thermally induced particulate events are assigned high-priority codes, while ordinary particulate events are assigned regular codes. The causes of disturbance events are correlated based on local pipe wall contamination data to generate priority micro-disturbance event codes.

6. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 1, characterized in that, Step S4 involves identifying and tracing risk sources based on verified disturbance event data, and performing risk event data chain processing, including: Based on the verified disturbance event data, the disturbance event type, quantification size, occurrence location and occurrence time are extracted to obtain the data of the quantified risk sources to be tracked; Identify downstream sensing nodes based on the location of occurrence in the data of the quantified risk sources to be tracked, and extract downstream infusion sensing feature data; Based on downstream infusion sensing characteristic data and data on risk sources to be tracked and quantified, spatiotemporal deviations are correlated to assess whether deviations are reproduced at upstream and downstream nodes, and abnormal propagation event type data is generated; among them, abnormal propagation event type data includes isolated disturbance events and continuous propagation events. When the abnormal propagation event type data is a continuous propagation event, the downstream propagation path data is determined based on the infusion path spatial topology data; when the abnormal propagation event type data is an isolated disturbance event, the data of the risk source to be tracked and quantified is sent to the terminal device for potential risk warning and event removal processing. Based on downstream propagation path data, the verified disturbance event data is processed into a risk event data chain to obtain the infusion risk data chain.

7. The method for digital traceability of infusion quality and safety based on the Internet of Things according to claim 1, characterized in that, Step S4, which involves monitoring disturbance events using a smart IoT sensing network based on the infusion risk data chain, includes: The propagation speed is calculated based on the position and time difference of the first and last nodes of the infusion risk data chain to obtain the abnormal event movement speed value; Calculate the estimated time of impact on the infusion terminal based on the movement speed value of the abnormal event; Based on the estimated time of impact at the end of infusion, the data of verified disturbance events are monitored through a smart IoT sensing network to obtain risk propagation verification data. When the risk propagation verification data is true, a highlighted dynamic line is drawn on the visualization interface of the infusion path spatial topology data to represent the propagation path, thus obtaining the dynamic traceability map of infusion risk.

8. A digital traceability system for infusion quality and safety based on the Internet of Things, characterized in that, For executing the IoT-based digital traceability method for infusion quality and safety as described in claim 1, the IoT-based digital traceability system for infusion quality and safety includes: The pipeline topology construction module is used to perform internal attitude scanning on the target infusion pipeline to obtain internal attitude data of the infusion pipeline; based on the internal attitude data of the infusion pipeline, it tracks the movement trajectory of the infusion device from the pharmaceutical preparation area to the administration point, and performs pipeline geometric feature analysis to generate infusion path spatial topology data; The IoT sensing and monitoring module is used to build an intelligent IoT sensing network based on the infusion path spatial topology data; it performs theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data to obtain theoretical fluid state characteristic data; wherein, step S2, performing theoretical risk characteristic analysis on the target infusion pipeline using the infusion path spatial topology data, includes: Obtain information about the infusion drug; this information includes the viscosity, density, and surface tension parameters under standard conditions at 25°C. The temperature of the infusion environment is monitored using a smart Internet of Things (IoT) sensing network, generating infusion environment temperature data. The average value is calculated based on the ambient temperature data of the infusion, and then the viscosity of the infusion drug information is temperature compensated to obtain the drug compensated viscosity data. Extracting vertical height difference and curvature data of the infusion path based on spatial topology data; Based on the information of the infusion drugs and the vertical height difference and curvature data of the path, the theoretical hydrostatic pressure and additional flow resistance of the target infusion pipeline are calculated in segments to obtain the theoretical fluid resistance distribution data. The Reynolds number distribution within the microchannel is calculated using the preset infusion flow rate and the infusion drug information. Sections with Reynolds numbers greater than 1500 are marked as high-risk turbulence areas, and a theoretical Reynolds number distribution map is generated. Based on the temperature data of the infusion environment, the temperature gradient of the physical space of the infusion is processed. When the temperature gradient exceeds 2 degrees Celsius, the Marangoni effect intensity risk index caused by the surface tension gradient is calculated by the drug compensation viscosity data to obtain the thermal effect risk index sequence. The theoretical fluid resistance distribution data, theoretical Reynolds number distribution map, and thermally induced effect risk index sequence are serialized and aligned to generate theoretical characteristic data of fluid state. Based on fluid state theory characteristic data, multi-dimensional infusion sensing signals are acquired using an intelligent IoT sensing network, and time delay compensation is performed to obtain spatiotemporal sensing monitoring signal data; wherein, step S2, which involves acquiring multi-dimensional infusion sensing signals based on fluid state theory characteristic data using an intelligent IoT sensing network and performing time delay compensation, includes: Based on the characteristic data analysis of fluid state theory, the high-risk area of ​​turbulence is analyzed, and the corresponding sensor nodes in the intelligent Internet of Things sensing network are set with differentiated acquisition frequencies to generate dynamic acquisition frequency configuration data. Among them, the differentiated acquisition frequency setting sets the acquisition frequency of the corresponding sensor nodes in the high-risk area of ​​turbulence to 100 Hz, and the acquisition frequency of other nodes to 20 Hz. Based on the dynamic acquisition frequency configuration data, the intelligent IoT sensing network is dynamically monitored and configured, and the multi-dimensional sensing signal acquisition of all sensor nodes in the intelligent IoT sensing network is initiated to obtain multi-source heterogeneous real-time sensing flow; among which, the multi-source heterogeneous real-time sensing flow includes pipeline start and stop displacement sequence, measured hydrostatic pressure data, infrared light transmittance sequence and ultrasonic echo signal flow. Digital filtering is performed on multi-source heterogeneous real-time sensing streams to generate standard multi-dimensional sensing signal data; Compensation is performed on the data time delay caused by the distance between nodes in the standard multidimensional sensing signal data, and timestamps are added to obtain spatiotemporal sensing monitoring signal data. The disturbance intelligent analysis module is used to intelligently analyze infusion disturbance events based on spatiotemporal sensing monitoring signal data and fluid state theoretical characteristic data, and to perform cross-validation of disturbance events to obtain verified disturbance event data. The risk traceability and early warning module is used to identify and track risk sources based on verified disturbance event data, and to process the risk event data chain to obtain the infusion risk data chain; based on the infusion risk data chain, it uses an intelligent Internet of Things sensing network to monitor disturbance events to obtain a dynamic traceability map of infusion risks; and it manages forward-looking risk warning instructions based on the dynamic traceability map of infusion risks to achieve digital traceability of infusion quality.

Images