Abnormality diagnosis method, device and system for infusion pump and storage medium

Abstract

The invention discloses an infusion pump abnormity diagnosis method, device and system and a storage medium, and relates to the technical field of medical apparatuses and instruments, and the method comprises the following steps: obtaining operation parameters of an infusion pump; performing time sequence prediction on the critical characteristic value of the operation parameter to obtain a prediction value, and calculating to obtain a trend prediction characteristic value according to the prediction value; inputting the critical characteristic value and the trend prediction characteristic value into a pre-trained fault classification model, and outputting a classification confidence coefficient corresponding to each fault classification result; based on the classification confidence coefficient and the trend prediction feature value, a comprehensive score is obtained through calculation, and the operation state of the infusion pump is determined according to the comprehensive score. The infusion pump abnormity diagnosis efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical devices, in particular to an abnormality diagnosis method of an infusion pump, an abnormality diagnosis device of an infusion pump, an abnormality diagnosis system of an infusion pump and a storage medium. BACKGROUND

[0002] The infusion pump is one of the important medical devices in the hospital, which is mainly used for accurately controlling the infusion of drugs or liquids to ensure the treatment process of patients. However, the existing infusion pump monitoring system mostly relies on manual inspection and on-site troubleshooting, and cannot continuously track the running state of the infusion pump. The abnormal state of the infusion pump is only determined based on a threshold value, so that the abnormality of the infusion pump cannot be accurately found and diagnosed, resulting in delayed fault response and prolonged equipment downtime, which seriously affects the treatment safety of patients and the operation efficiency of the hospital. In summary, the existing abnormality diagnosis process of the infusion pump has the problem of low abnormality diagnosis efficiency, which is difficult to support the fine operation and maintenance management requirements of high reliability and high availability medical devices in the modern medical scene.

[0003] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0004] The main purpose of the present application is to provide an abnormality diagnosis method of an infusion pump, an abnormality diagnosis device of an infusion pump, an abnormality diagnosis system of an infusion pump and a storage medium, which aims to solve the technical problem of low abnormality diagnosis efficiency in the existing abnormality diagnosis process of the infusion pump.

[0005] To achieve the above purpose, the present application provides an abnormality diagnosis method of an infusion pump, which comprises:

[0006] obtaining the running parameters of the infusion pump;

[0007] performing time series prediction on the critical characteristic value of the running parameters to obtain a predicted value, and calculating a trend prediction characteristic value according to the predicted value;

[0008] inputting the critical characteristic value and the trend prediction characteristic value into a pre-trained fault classification model to output a classification confidence corresponding to each fault classification result;

[0009] calculating a comprehensive score based on the classification confidence and the trend prediction characteristic value, and determining the running state of the infusion pump according to the comprehensive score.

[0010] In an embodiment, the trend prediction feature value comprises a predicted time feature value, a residual feature value and a predicted change rate feature value; the step of performing time series prediction on the critical feature value of the operating parameter to obtain a predicted value, and calculating the trend prediction feature value according to the predicted value, comprises:

[0011] performing time series prediction on the critical feature value of the operating parameter to obtain a first predicted value at a future time point, and a second predicted value at each time point from the current time point to the future time point;

[0012] determining a first time point at which the second predicted value is first greater than or equal to a first preset threshold, and calculating a predicted time from the current time point to the first time point as the predicted time feature value;

[0013] calculating a difference between the first predicted value and the critical feature value as the residual feature value;

[0014] calculating a change rate of the critical feature value in a period from the current time point to the future time point as the predicted change rate feature value;

[0015] weighting and fusing the predicted time feature value, the residual feature value and the predicted change rate feature value to obtain a trend prediction feature value.

[0016] In an embodiment, before the step of inputting the critical feature value and the trend prediction feature value into the pre-trained fault classification model, the method comprises:

[0017] obtaining historical data with historical fault type labels, wherein the historical data comprises historical critical feature values and historical trend prediction features corresponding to the historical critical feature values;

[0018] training an initial model by taking the historical critical feature values and the historical trend prediction feature values as input features and taking corresponding historical fault type labels as output targets, until the accuracy of a fault classification result output by the initial model reaches a preset accuracy threshold, to obtain the fault classification model.

[0019] In an embodiment, after the step of obtaining the operating parameter of the infusion pump, the method comprises:

[0020] comparing the operating parameter with a preset safety threshold, wherein the preset safety threshold comprises a first preset threshold and a second preset threshold, and the first preset threshold is greater than the second preset threshold;

[0021] if the operating parameter is less than the second preset threshold, determining that the infusion pump is normal;

[0022] If the operation parameter is greater than or equal to the second preset threshold but less than the first preset threshold, a time series prediction is performed on the critical feature value of the operation parameter to obtain a predicted value, and a trend prediction feature value is calculated according to the predicted value;

[0023] If the operation parameter is greater than or equal to the first preset threshold, it is determined that the infusion pump is in a serious abnormal state, and an alarm signal is generated.

[0024] In an embodiment, the step of calculating a comprehensive score based on the classification confidence and the trend prediction feature value comprises:

[0025] The residual feature value in the trend prediction feature value is standardized to obtain a first score;

[0026] The classification confidence is taken as a second score;

[0027] The first score and the second score are weighted and fused to obtain a comprehensive score.

[0028] In an embodiment, the step of calculating a comprehensive score based on the classification confidence and the trend prediction feature value comprises:

[0029] If the comprehensive score reaches a preset score threshold, an operation parameter calibration instruction is generated according to the fault classification result corresponding to the highest classification confidence;

[0030] The operation parameter calibration instruction is sent to the infusion pump.

[0031] In an embodiment, the method comprises:

[0032] The fault classification model is incrementally trained at a first preset period to obtain an updated fault classification model;

[0033] The time series prediction model is re-fitted at a second preset period to obtain an updated time series prediction model, wherein the time series prediction model is used to perform time series prediction on the critical feature value of the operation parameter;

[0034] An abnormality diagnosis of the infusion pump is performed through the updated time series prediction model and the updated fault classification model.

[0035] Further, in order to achieve the above object, the present application also provides an abnormality diagnosis device of an infusion pump, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the abnormality diagnosis method of the infusion pump as described above.

[0036] Further, in order to achieve the above object, the present application also provides an abnormality diagnosis system of an infusion pump, comprising an infusion pump and the abnormality diagnosis device of the infusion pump.

[0037] Further, in order to achieve the above object, the present application also provides a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the abnormality diagnosis method of the infusion pump as described above.

[0038] Further, in order to achieve the above object, the present application also provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps of the abnormality diagnosis method of the infusion pump as described above.

[0039] The one or more technical solutions provided by the present application have at least the following technical effects:

[0040] Since the existing abnormality diagnosis process of the infusion pump has the problem of low efficiency, after obtaining the running parameters of the infusion pump, the present application performs time series prediction on the critical characteristic values extracted from the running parameters to obtain predicted values, and calculates trend prediction characteristic values based on the predicted values, thereby realizing trend analysis of the running state of the infusion pump, identifying potential abnormalities before failure occurs, and overcoming the lag of relying on manual inspection and on-site troubleshooting.

[0041] Further, the critical characteristic values and the trend prediction characteristic values are input into a pre-trained fault classification model, and the model calculates the classification confidence corresponding to the fault classification result with a probability significance, thereby realizing accurate identification and quantitative evaluation of the fault type; improving the accuracy and interpretability of the diagnosis process, and overcoming the limitation of inaccurate traditional single threshold judgment.

[0042] Finally, a comprehensive score is calculated based on the classification confidence and the trend prediction characteristic values, and the running state of the infusion pump is determined according to the comprehensive score, thereby realizing objective and quantitative evaluation of the abnormal state of the infusion pump. The present application overcomes the lag of relying on manual diagnosis and the limitation of inaccurate traditional single threshold judgment through the above technical means, realizes the accuracy and timeliness of the abnormality diagnosis of the infusion pump, and significantly improves the efficiency of the abnormality diagnosis of the infusion pump. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate preferred embodiments of the present application and, together with the description, serve to explain the principles of the application.

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings can also provide other drawings for those ordinarily skilled in the art without any creative effort.

[0045] Figure 1 A flowchart provided by an embodiment of the abnormality diagnosis method of the infusion pump of the present application;

[0046] Figure 2 A hierarchical processing schematic diagram provided by the abnormality diagnosis method of the infusion pump of the present application;

[0047] Figure 3 A brief flowchart provided by the abnormality diagnosis method of the infusion pump of the present application;

[0048] Figure 4 A schematic diagram of the abnormality diagnosis system of the infusion pump of the present application;

[0049] Figure 5 A device structure schematic diagram of the hardware running environment involved in the abnormality diagnosis method of the infusion pump of the present application.

[0050] The object implementation, functional features and advantages of the present application will be further explained with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and not to limit the present application.

[0052] In order to better understand the technical solutions of the present application, the following will be described in detail in combination with the drawings of the specification and the specific embodiments.

[0053] It should be noted that the execution subject of the present embodiment can be an abnormality diagnosis device of an infusion pump, or a computing service device with data processing, network communication and program running functions, such as a cloud platform, a server, a tablet computer, a personal computer, a mobile phone, etc., or an electronic device or a processor capable of realizing the above functions. The present embodiment and the following embodiments will be described below taking the cloud platform as an example.

[0054] Based on this, the present embodiment provides an abnormality diagnosis method of an infusion pump, which will be described in detail with reference to Figure 1 , Figure 1 A flowchart of an embodiment of the abnormality diagnosis method of the infusion pump of the present application.

[0055] In this embodiment, the abnormality diagnosis method of the infusion pump comprises steps S10-S40:

[0056] In step S10, the operating parameters of the infusion pump are acquired.

[0057] It should be noted that the types of operating parameters of the infusion pump include battery capacity, infusion rate, cumulative amount, temperature, pressure, bubble detection, etc. The acquisition of the operating parameters of the infusion pump can be real-time collection by the built-in sensor of the infusion pump, or periodic collection. Specifically, the infusion rate can be collected by a Hall sensor, the pressure can be collected by a pressure sensor, the bubble detection can be performed by ultrasonic waves, the battery capacity can be collected by a battery management chip every 60 seconds, the temperature can be collected by an NTC thermistor every 20 seconds, and the cumulative amount can be obtained by a counter; and then uploaded to the cloud platform. In an embodiment, a communication module can be integrated in the infusion pump, and the communication module is used to upload to the cloud platform. In another embodiment, the cloud platform can also be uploaded to the cloud platform by a wireless method such as a Bluetooth method.

[0058] Exemplarily, the communication module can be a communication module supporting multiple 4G protocols [FDD-LTE (Frequency Division Duplexing-Long Term Evolution) or TD-LTE (Time Division Duplexing-Long Term Evolution) and the like], compatible with 2G or 3G protocol fallback, which can ensure that the infusion pump maintains stable communication connection in different environments, so as to ensure that the operating parameters of the infusion pump can be stably and real-timely uploaded, facilitating remote monitoring and fault analysis. The hardware of the communication module can be an industrial-grade 4G Cat.1 (Category 1) DTU (Data Transfer Unit) module, the interface can be an MCU (Microcontroller Unit) such as an STM32 series, the asynchronous transceiver can be configured to have a transmission rate of 115200 bps, transmit 8 data bits, request to send or clear to send, the power supply can be an independent DC-DC voltage regulator, support low-power mode, the antenna can be a dual-antenna diversity receiver, which can effectively avoid metal shielding area, the firmware configuration can be that the MCU sends an MQTT connection parameter frame after power-on, sets a Topic (in the MQTT protocol, refers to the address or classification label of the message), queries the network state, and the data interaction protocol can be a self-defined protocol frame structure, supporting data upload, remote instruction issuing, and state reporting.

[0059] It also needs to be explained that after collecting the operating parameters, the infusion pump will preprocess the operating parameters, such as filtering processing using a sliding window mean and abnormal value elimination processing; then according to the frame length determined by the self-defined protocol in the communication module in the infusion pump, data packaging processing is performed; then using an encryption algorithm such as the AES encryption algorithm, the data packet is encrypted, and uploaded to the cloud platform through the communication module. Exemplarily, it can be real-time uploading, batch uploading, and using cache retransmission when the line is disconnected. How to upload can be set according to the actual situation.

[0060] In a possible implementation, after step S10, steps S11-S14 are included:

[0061] Step S11, comparing the operating parameters with the preset safety threshold, wherein the preset safety threshold includes a first preset threshold and a second preset threshold, and the first preset threshold is greater than the second preset threshold;

[0062] It needs to be explained that the obtained operating parameters are compared with the respective preset safety threshold. Exemplarily, the safety threshold range of flow rate is ±5%, the safety threshold range of pressure is 180-200 kPa, etc., and the safety threshold can be configured according to the actual application scene.

[0063] In an embodiment, the safety threshold includes a first preset threshold and a second preset threshold, wherein the first preset threshold is higher than the second preset threshold. If a certain operating parameter exceeds the second preset threshold but has not reached the first preset threshold, the cloud platform determines that the operating parameter is in a "slight out-of-limit" state, and the cloud platform performs time series prediction at this time. For example, taking the pressure parameter as an example, the first preset threshold (i.e. the serious abnormal threshold) can be set to 200 kPa, and when the pressure breaks through the first preset threshold, it indicates that there is a high risk of serious blockage of the pipeline or failure of the sensor, and the cloud platform will directly trigger the highest level alarm. The second preset threshold (i.e. the slight out-of-limit threshold) is set to 180 kPa, and if the pressure value is in the range of 180 kPa to 200 kPa, it is determined to be in a "slight out-of-limit" state, indicating that there may be early signs of partial blockage or poor liquid flow. It can be understood that this hierarchical threshold mechanism not only avoids the missed report of subtle abnormalities by the traditional single threshold, but also prevents excessive analysis of the obviously normal state, significantly optimizing the allocation efficiency of computing resources. Exemplarily, when the pressure value rises from 170 kPa to 185 kPa (breaks through the second threshold but does not reach the first threshold) within 20 seconds, the cloud platform will trigger the second level determination process for trend analysis, without waiting for the pressure to break through 200 kPa to respond, which makes the disposal window period of high-risk events one time earlier than the existing, and can better respond to failures.

[0064] Step S12, if the operating parameter is less than the second preset threshold, determining that the infusion pump is normal;

[0065] It should be noted that when the running parameter is less than the second preset threshold, the cloud platform determines that the infusion pump is in a normal running state. In the specific embodiment, this means that the running parameter fluctuates within the safe range, and there is no need to start complex analysis. Exemplarily, when the pressure parameter continuously drops below 180 kPa, the cloud platform does not need further processing. It can be understood that this screening mechanism effectively avoids unnecessary consumption of computing resources, and significantly improves the running efficiency of the cloud platform in monitoring the infusion pump.

[0066] Step S13, if the running parameter is greater than or equal to the second preset threshold but less than the first preset threshold, time series prediction of the critical characteristic value of the running parameter is performed to obtain a predicted value, and a trend prediction characteristic value is calculated according to the predicted value.

[0067] It should be noted that when the running parameter is between the second preset threshold and the first preset threshold, the cloud platform identifies potential risks and starts deep analysis. In the specific embodiment, this interval represents that the infusion pump is in a critical state, and needs to be time series predicted and calculated for the trend characteristic value. It can be understood that this prediction mechanism for the critical state realizes the transition from passive response to active early warning, and can identify potential risks before failure occurs.

[0068] Step S14, if the running parameter is greater than or equal to the first preset threshold, the infusion pump is determined to be a serious abnormality, and an alarm signal is generated.

[0069] It should be noted that when the running parameter exceeds the first preset threshold, the cloud platform determines that it is a serious abnormality and generates an alarm signal. In the specific embodiment, this indicates that the infusion pump has entered a dangerous running state, and immediate intervention measures need to be taken. It can be understood that this fast response mechanism ensures timely handling in emergency situations, effectively ensuring patient safety and infusion pump equipment reliability.

[0070] In the present embodiment, by establishing a hierarchical threshold judgment system, intelligent hierarchical processing of the running state of the infusion pump is realized. It can be understood that this hierarchical processing mechanism can take different coping strategies according to the severity of the abnormality: light monitoring for normal state, predictive analysis for critical state, and immediate alarm for serious abnormality. Specifically, this scheme solves the problem of single response mode of traditional monitoring systems to different degrees of abnormality, avoiding overreaction to slight fluctuations and ensuring fast response to serious abnormalities. At the same time, by concentrating computing resources on the critical state that really needs attention, the running efficiency and accuracy of abnormal diagnosis of the cloud platform monitoring are significantly improved, and the reliability is improved.

[0071] Exemplarily, as shown in FIG. 4, the cloud platform first determines whether the running parameter is less than the second preset threshold (step S11). If the running parameter is less than the second preset threshold, the cloud platform determines that the infusion pump is in a normal running state, and no further processing is needed (step S12). Figure 2 Figure 2 ​For the hierarchical threshold processing schematic diagram, after obtaining the running parameter, compare with the respective preset safety threshold; if the running parameter is greater than or equal to the first preset threshold, it is judged that the infusion pump is a serious abnormality; if the running parameter is less than the second preset threshold, it is judged that the infusion pump is normal, and no processing is required, which can be recorded in the log; when the running parameter is between the second preset threshold and the first preset threshold, the cloud platform identifies potential risks, performs time series prediction and calculates trend characteristic values. The hierarchical processing mechanism of the application can take different coping strategies according to the severity of the abnormality, concentrate the computing resources on the critical state that really needs attention, significantly improve the running efficiency and abnormal diagnosis accuracy of the cloud platform monitoring, and improve the reliability.

[0072] Step S20, time series prediction is performed on the critical characteristic value of the running parameter to obtain a predicted value, and a trend prediction characteristic value is calculated according to the predicted value;

[0073] It should be noted that the cloud platform extracts critical characteristic values from the operating parameters. Specifically, the infusion rate, i.e., the flow rate, the pressure, the battery power, the bubble detection, the temperature, and the respective characteristic values are the flow rate fluctuation rate, the pressure change rate, the battery attenuation rate, the bubble detection frequency, and the temperature deviation. The characteristic compensation is as follows: the flow rate fluctuation rate is FRvar = σ(flow) / μ(flow), where FRvar is the flow rate fluctuation rate, σ(flow) is the standard deviation of the flow rate, indicating the fluctuation degree of the flow rate data, and μ(flow) is the mean value of the flow rate, indicating the average flow rate within a period of time. The pressure change rate is PRrate = Δpressure / Δt, where PRrate is the pressure change rate, Δpressure is the pressure change (the difference between the current pressure value and the previous pressure value), and Δt is the time interval (the time difference between two sampling times, in seconds). The battery attenuation rate is Sabattery = ΔB / Δt (% / min), where ΔB is the battery power change, and Δt is the time interval. It is used to judge the battery health status and predict the remaining available time, and combined with other features to evaluate the infusion pump operating state. The bubble detection frequency is Bubble_freq = bubble event number / monitoring time, and the normal case is that the bubble detection frequency is close to 0 (with occasional individual bubbles). The abnormal case is that the bubble detection frequency continues to rise, indicating that there is continuous air entering the pipeline. The sudden bubble detection frequency increases, indicating that the infusion bottle or bag is about to run out. High frequency but no actual bubbles indicate that the ultrasonic sensor may have a fault. The temperature deviation and characteristic compensation are as follows: the current environmental temperature Tenv is collected in real time by an external temperature sensor, and the historical environmental temperature mean Tenv_history is calculated within a sliding time window. The temperature deviation is defined as: Tempdev = Tenv-Tenv_history. The temperature deviation Tempdev is used for temperature compensation of other characteristic values (such as flow rate fluctuation rate, pressure change rate, battery attenuation rate, etc.): Featurecorrected = Featureraw-k·Tempdev, where Featureraw is the uncorrected characteristic value (such as flow rate fluctuation rate, pressure change rate, battery attenuation rate, etc.), k is the temperature compensation coefficient (unit: characteristic value change / ℃), which is obtained by experimental calibration, Tempdev is the temperature deviation (℃), and Featurecorrected is the compensated characteristic value. Through compensation, the influence of environmental temperature change on sensor readings can be effectively eliminated, false positives can be reduced, and the accuracy of fault analysis can be improved.

[0074] Then, the cloud platform performs time series prediction on the critical characteristic value of the running parameter. Exemplarily, an ARIMA model is used to model and analyze the historical data sequence, and the prediction value at the future time point is calculated to capture the parameter change trend. In the specific embodiment, the trend prediction characteristic value is further calculated based on the prediction value, and the trend prediction characteristic value can effectively represent the dynamic characteristics and abnormal development trend of the parameter change. Understandably, this step realizes early identification of potential faults through trend analysis of the running state, and overcomes the limitations of the manual inspection method in terms of lag processing of sudden abnormalities and insensitivity to gradual faults.

[0075] In a feasible embodiment, the trend prediction characteristic value includes an expected time characteristic value, a residual characteristic value, and a predicted change rate characteristic value; and step S20 includes steps S21-S25:

[0076] Step S21, time series prediction is performed on the critical characteristic value of the running parameter to obtain a first prediction value at a future time point and a second prediction value at each time point from the current time point to the future time point;

[0077] It should be noted that, by performing time series prediction on the critical characteristic value of the running parameter, not only the first prediction value at the future specific time point is obtained, but also a continuous second prediction value sequence from the current time point to the future time point is obtained. In the specific embodiment, the time series prediction model such as the ARIMA (Autoregressive Integrated Moving Average) model can be used to realize this, which can predict the future development trend of the parameter according to the historical change rule of the parameter. Exemplarily, for the pressure parameter of the infusion pump, the time series prediction model can predict the pressure value every second within the next 5 minutes, the first prediction value at the future time point is the pressure value at the 5th minute, and the second prediction value at each time point is the pressure value every second from the 0th minute to the 5th minute. It should be noted that the first prediction value can be equal to the second prediction value, the first prediction value is only one, and the second prediction value can be one or more. Understandably, using the continuous prediction mode can capture the subtle features of the parameter change, capture the evolution rule of the running parameter of the infusion pump, know the possible development trend of the running parameter in advance, solve the problem that the traditional threshold judgment cannot predict the deterioration trend, and provide a time window for active intervention.

[0078] Step S22, determining a first time point at which the second prediction value is first greater than or equal to a first preset threshold, and calculating an expected time from the current time point to the first time point as an expected time characteristic value;

[0079] It should be noted that the cloud platform calculates the predicted time from the current time to the time point at which the second predicted value exceeds the first preset threshold for the first time by determining the time point at which the second predicted value exceeds the first preset threshold for the first time. It should be noted that exceeding the first preset threshold is a serious abnormality, that is, the infusion pump is in a serious abnormality, and a dangerous situation that needs to be highly concerned about may occur. In the specific embodiment, this predicted time characteristic value quantifies the remaining time for the running parameter to deteriorate to a "serious out-of-limit" state. It can be understood that the shorter the predicted time, the more urgent the situation, and the higher the priority of the cloud platform to take countermeasures, which effectively solves the problem that the traditional method cannot quantify the risk evolution speed. Illustratively, when the predicted value of the pressure parameter will exceed the first threshold for the first time after 180 seconds, the cloud platform can obtain a 180-second warning time window. It can be understood that this characteristic value can provide an emergency response time for medical personnel, greatly improving the practicality of the warning and significantly shortening the response time of a high-risk event.

[0080] Step S23, calculating the difference between the first predicted value and the critical characteristic value as a residual characteristic value;

[0081] It should be noted that the cloud platform obtains the residual characteristic value by calculating the difference between the first predicted value and the current critical characteristic value. In the specific embodiment, this residual characteristic value reflects the degree of deviation between the prediction model and the actual running state. Illustratively, a larger positive residual may indicate that the running parameter is growing faster than expected, suggesting an accelerated abnormality; and a negative residual may indicate that the running parameter change trend does not match the expectation. It can be understood that the residual characteristic value can effectively identify the mismatch of the prediction model and is the basis for updating the time series prediction model and detecting abnormalities.

[0082] Step S24, calculating the change rate of the critical characteristic value in the period from the current time point to the future time point as a predicted change rate characteristic value;

[0083] It should be noted that the cloud platform obtains the predicted change rate characteristic value by calculating the change rate of the critical characteristic value in a specified period (in the period from the current time point to the future time point). In the specific embodiment, this change rate characteristic value quantifies the degree of parameter change. Illustratively, the change rate of the pressure parameter can distinguish between normal slow rise and abnormal rapid growth. It can be understood that the change rate characteristic value can effectively identify sudden abnormalities, and compared with the traditional static threshold monitoring, this dynamic characteristic enables the cloud platform to distinguish between slow aging and sudden failure, making up for the shortcomings of single numerical judgment.

[0084] Step S25, based on the predicted time characteristic value, the residual characteristic value and the predicted change rate characteristic, a trend prediction characteristic value is obtained by weighted fusion.

[0085] It should be noted that the cloud platform combines multiple feature values into a comprehensive trend prediction feature value through weighted fusion. In specific embodiments, an adaptive weighting algorithm can be used to dynamically adjust the weights according to the importance of different features. For example, in an emergency, the predicted time feature can be given a higher weight, while in a stable running phase, the change rate feature can be given a higher weight. It can be understood that this multi-feature fusion method can comprehensively reflect the dynamic characteristics of the parameters, improve the accuracy of trend prediction, and the trend prediction feature value solves the defect of insufficient reliability of single parameter determination, and enriches the information dimension for fault classification.

[0086] In the present embodiment, the risk assessment is quantified by the predicted time feature value, the residual feature value and the predicted change rate feature value, and the cloud platform can identify potential faults earlier and more accurately, significantly improving the reliability and safety of the infusion pump operation. Specifically, the predicted time feature value provides a clear time reference, the residual feature value ensures the prediction reliability, and the predicted change rate feature value enhances the sensitivity of the cloud platform to sudden abnormalities, so that the cloud platform can adapt to various complex operating conditions.

[0087] Step S30, inputting the critical feature value and the trend prediction feature value into the pre-trained fault classification model, and outputting the classification confidence corresponding to each fault classification result;

[0088] It should be noted that the cloud platform inputs the critical feature value and its trend prediction feature value into the pre-trained fault classification model. In specific embodiments, the fault classification model can use a random forest model to determine the fault type based on multi-dimensional features and output the classification confidence corresponding to each fault type. It can be understood that this step realizes the leap from single parameter threshold judgment to multi-feature fusion fault recognition, and uses the classification ability of the machine learning model to accurately identify multiple common fault types such as blockage, liquid leakage, sensor failure, etc., and gives the occurrence probability of each fault type, significantly improving the diagnosis ability of complex faults. It should also be noted that the training process of the time series prediction model and the fault classification model can be in the cloud platform or in other servers.

[0089] In one possible embodiment, step S30 includes steps S31-S32:

[0090] Step S31, obtaining historical data with historical fault type labels, wherein the historical data includes historical critical feature values and historical trend prediction features corresponding to the historical critical feature values;

[0091] It should be noted that the training set of the model is constructed by acquiring historical data with historical fault type labels, the historical data including critical characteristic values (such as flow rate fluctuation rate after temperature compensation, pressure change rate) of a historical period and predicted values obtained by performing time series prediction on the critical characteristic values of the historical period, and the historical trend prediction characteristic values calculated. Exemplarily, the historical critical characteristic values can include data extracted from the operating parameters in the past year.

[0092] In step S32, the historical critical characteristic values and the historical trend prediction characteristic values are taken as input features, and the corresponding historical fault type labels are taken as output targets, and the initial model is trained until the accuracy of the fault classification result output by the initial model reaches a preset accuracy threshold, to obtain a fault classification model.

[0093] It should be noted that the initial model is trained by using historical data, the historical critical characteristic values and the historical trend prediction characteristic values are taken as input features, and the corresponding historical fault type labels are taken as output targets. In a specific embodiment, the initial model can be a random forest model, which judges the input features in multiple dimensions by multiple trees and outputs the classification result and confidence by using a voting mechanism. The training process is continuously iterated until the fault classification accuracy output by the initial model reaches a preset accuracy threshold (such as 85%), thereby obtaining the final fault classification model.

[0094] In the embodiment, the fault classification model obtained by training using the labeled historical data not only has high diagnostic accuracy and generalization ability, but also realizes the accuracy and reliability of the fault type recognition in the fault diagnosis process, effectively solving the technical problems of the conventional infusion pump fault diagnosis, such as excessive dependence on expert experience and fixed rules, thereby leading to limited diagnostic range and poor adaptability to new or complex faults.

[0095] In step S40, a comprehensive score is calculated based on the classification confidence and the trend prediction characteristic value, and the running state of the infusion pump is determined according to the comprehensive score.

[0096] It should be noted that the cloud platform calculates a comprehensive score based on the classification confidence and the trend prediction characteristic value, and determines the running state (such as normal, attention, warning or serious abnormality) of the infusion pump according to the comprehensive score. In an embodiment, the comprehensive score can be obtained by weighting and fusing the output results of multiple models. According to the level of the comprehensive score, the running state of the infusion pump is determined, the graded warning and precise response are realized, and the problems of the conventional infusion pump monitoring, such as untimely fault diagnosis and lack of predictive maintenance, are effectively solved, realizing the transformation from passive response to active warning, and finally significantly improving the safety and operation efficiency of the equipment running.

[0097] In the specific embodiment, according to the comprehensive score (ranging from 0 to 1), the running state of the infusion pump is divided: the comprehensive score is less than 0.3, which belongs to normal; the comprehensive score is greater than or equal to 0.3 and less than 0.6, which needs to be paid attention to by medical staff; the comprehensive score is greater than or equal to 0.6 and less than 0.8, which is prewarned; and the comprehensive score is greater than or equal to 0.8, which belongs to serious abnormality. Exemplarily, the boundary value can be adjusted according to historical running data and different application scenarios to balance the sensitivity and false alarm rate.

[0098] The embodiment provides an abnormality diagnosis method of an infusion pump. After obtaining running parameters of the infusion pump, a critical characteristic value extracted from the running parameters is subjected to time series prediction to obtain a predicted value, and a trend prediction characteristic value is calculated based on the predicted value, so that trend analysis of the running state of the infusion pump is realized, potential abnormalities can be identified before failure occurs, and the lag of relying on manual inspection and on-site troubleshooting is overcome.

[0099] Further, the critical characteristic value and the trend prediction characteristic value are input into a pre-trained fault classification model, a classification confidence corresponding to a fault classification result with a probability significance is calculated by the model, accurate identification and quantitative evaluation of the fault type are realized, the accuracy and interpretability of the diagnosis process are improved, and the limitation of inaccurate traditional single threshold judgment is overcome.

[0100] Finally, a comprehensive score is calculated based on the classification confidence and the trend prediction characteristic value, and the running state of the infusion pump is determined according to the comprehensive score, so that objective and quantitative evaluation of the abnormal state of the infusion pump is realized. The above technical means of the present application overcomes the lag of relying on manual diagnosis and the limitation of inaccurate traditional single threshold judgment, realizes the accuracy and timeliness of abnormality diagnosis of the infusion pump, and significantly improves the efficiency of abnormality diagnosis of the infusion pump.

[0101] Based on the above embodiment of the present application, in another embodiment of the present application, the same or similar contents as the above embodiment can be referred to the above introduction, and will not be described in detail. The step S40 of the abnormality diagnosis method of the infusion pump comprises steps S41-S43:

[0102] Step S41, the residual characteristic value in the trend prediction characteristic value is subjected to standardization processing to obtain a first score;

[0103] It should be noted that the cloud platform maps the residual characteristic value in the trend prediction characteristic value to a first score in the range of 0 to 1 by standardizing the residual characteristic value. In specific embodiments, the residual characteristic value reflects the degree of deviation between the prediction model and the actual operation data. For example, when there is a large difference between the predicted value and the actual observation value, it indicates that abnormal fluctuations may occur. For example, the residual characteristic value can be converted to a score value between 0 and 1 by a standardization method such as Z-score or Min-Max. It can be understood that this standardization processing makes the residual characteristic values of different parameters and different magnitudes comparable, laying a foundation for subsequent fusion.

[0104] Step S42, taking the classification confidence as a second score;

[0105] It should be noted that the classification confidence is taken as the second score. In specific embodiments, the classification confidence is the degree of certainty of the fault classification model for its judgment result. For example, when the fault classification model judges that the "pipeline blockage" fault with a confidence of 90%, the 90% value itself has quantitative significance and can be directly used as the second score. It can be understood that directly using the classification confidence as the score not only retains the original information output by the fault classification model, but also avoids unnecessary resource consumption caused by conversion.

[0106] Step S43, weighting and fusing the first score and the second score to obtain a comprehensive score.

[0107] It should be noted that the first score and the second score are combined into a comprehensive score by a weighting fusion method. In specific embodiments, a dynamic weight distribution strategy can be used to adjust the weight proportion of the two scores according to different application scenarios. For example, in the ICU scenario with extremely high safety requirements, a higher weight can be given to the first score; while in the general ward scenario, more attention can be paid to the weight of the second score. It can be understood that this weighting fusion method can comprehensively consider the accuracy of the prediction model and the reliability of the fault classification, and form a more comprehensive and accurate abnormality evaluation result.

[0108] For example, Score = w1S1 + w2S2, where Score represents the comprehensive score; w1 represents the weight coefficient of the first score, S1 represents the first score; w2 represents the weight coefficient of the second score, and S2 represents the second score.

[0109] In the embodiment, the effective integration of multi-source information is realized by converting the residual eigenvalue and the classification confidence of different dimensions into a unified scoring system and performing weighted fusion. It can be understood that this fusion scoring method solves the limitations of single index evaluation: the first score reflects the goodness of fit of the time series prediction model and the stability of the infusion pump operation, while the second score reflects the certainty and reliability of fault diagnosis. This scoring mechanism provides a quantitative basis for subsequent early warning decisions, significantly improving the accuracy and reliability of abnormal diagnosis, and also enhancing the adaptability and practicality of the infusion pump diagnosis.

[0110] In another embodiment, the critical eigenvalue and the trend prediction eigenvalue are input into the pre-trained fault classification model to also output a fault classification result, and step S40 includes:

[0111] If the comprehensive score reaches the preset score threshold, an operation parameter calibration instruction is generated according to the fault classification result corresponding to the highest classification confidence;

[0112] The operation parameter calibration instruction is issued to the infusion pump.

[0113] It should be noted that inputting the critical eigenvalue and the trend prediction eigenvalue into the pre-trained fault classification model outputs not only the fault classification result such as blockage and sensor failure, but also the confidence of the fault classification result such as blockage 0.8 and sensor failure 0.1. When the comprehensive score reaches the preset score threshold, the cloud platform automatically generates the corresponding operation parameter calibration instruction according to the fault classification result corresponding to the highest classification confidence. In the specific embodiment, the preset score threshold represents the critical point at which the cloud platform determines that active intervention is needed. For example, when the comprehensive score exceeds 0.6, it indicates that the cloud platform detects a highly suspicious abnormal state and needs to take calibration measures immediately. It can be understood that this triggering mechanism based on the score threshold ensures that the cloud platform only starts the calibration process when it is sure that there is an anomaly, avoiding unnecessary device intervention.

[0114] It should be noted that the generated operation parameter calibration instruction will be issued to the infusion pump device end through the communication network. In the specific embodiment, the operation parameter calibration instruction contains specific calibration parameters and operation steps. For example, the flow rate calibration instruction can contain target calibration value, calibration duration, and allowable error range, etc. It can be understood that the operation parameter calibration instruction issuing process adopts an encrypted transmission protocol to ensure data security, and at the same time has a retransmission mechanism to ensure reliable delivery of the instruction. After receiving the operation parameter calibration instruction, the infusion pump will automatically execute the corresponding self-calibration program, and the parameter adjustment and device state optimization can be completed without manual intervention. For example, when the cloud platform identifies that there is a drift fault in the pressure sensor, it will automatically issue a zero-point calibration instruction to restore the measurement accuracy of the infusion pump device without stopping. This automatic calibration process significantly reduces the downtime of the infusion pump device, improves the operation and maintenance efficiency, and at the same time ensures the accuracy and safety of infusion therapy through precise parameter adjustment.

[0115] In the present embodiment, by establishing a closed loop from anomaly diagnosis to automatic calibration, intelligent operation and maintenance of the infusion pump is realized. It can be understood that the present embodiment solves the problem of disconnection between diagnosis and treatment in traditional operation and maintenance: not only can the fault type be accurately identified, but also targeted calibration measures can be taken immediately. Specifically, the triggering mechanism based on comprehensive scoring ensures the timeliness and accuracy of intervention, and the principle of generating instructions according to the highest confidence ensures the pertinence of calibration operation.

[0116] In yet another embodiment, the anomaly diagnosis method of the infusion pump further comprises steps D10-D30:

[0117] Step D10, performing incremental training on the fault classification model according to a first preset period to obtain an updated fault classification model;

[0118] It should be noted that by performing incremental training on the fault classification model according to a first preset period (such as every month), the fault classification model can continuously learn and adapt to new operation data. In the specific embodiment, incremental training means that the newly collected fault case data is added to the training set every month to fine-tune and optimize the parameters of the fault classification model. For example, the decision rules of the classification model such as random forest can be updated. It can be understood that this periodic updating mechanism can enable the fault classification model to continuously adapt to new fault patterns caused by device aging, changes in use environment, etc., and continuously improve the accuracy of diagnosis.

[0119] Step D20, re-fitting model parameters of the time series prediction model according to a second preset period to obtain an updated time series prediction model, wherein the time series prediction model is used for time series prediction of critical characteristic values of the operation parameter;

[0120] It should be noted that the time series prediction model is refitted with the second preset period (such as every day) to ensure that the time series prediction model is consistent with the current infusion pump device running state. In the specific embodiment, the time series prediction model needs to re-estimate its autoregressive and moving average parameters according to the latest historical data. Illustratively, the time series prediction model can be retrained daily using data from the past 24 hours to capture the latest changes in infusion pump device running characteristics. It can be understood that periodic refitting can enable the time series prediction model to reflect the attenuation of the infusion pump device performance or changes in environmental conditions in a timely manner, ensuring the accuracy and timeliness of trend prediction.

[0121] Step D30, abnormal diagnosis of the infusion pump by the updated time series prediction model and the updated fault classification model.

[0122] It should be noted that by using the updated time series prediction model and the fault classification model for abnormal diagnosis, the continuous optimization of the diagnosis capability is realized. In the specific embodiment, the updated model will replace the original model and be put into actual use to process new input running parameter data. Illustratively, the incrementally trained fault classification model can identify new fault types, and the updated time series prediction model can more accurately predict parameter trend. It can be understood that this model updating mechanism ensures that the cloud platform can continuously adapt to changes in the running state of the infusion pump device.

[0123] In this embodiment, by establishing a model periodic updating mechanism, self-optimization and continuous improvement of the abnormal diagnosis device of the infusion pump are realized. It can be understood that the problem of model solidification of the traditional diagnosis system and the inability to adapt to changes in the state of the infusion pump device is solved. Specifically, the incremental training of the fault classification model can learn new fault patterns, improving the coverage and accuracy of diagnosis; the parameter refitting of the time series prediction model ensures that the trend prediction always matches the current state of the infusion pump device, enhancing the timeliness and reliability of the early warning.

[0124] Illustratively, in order to facilitate understanding of the implementation process of the abnormal diagnosis method of the infusion pump obtained by combining the above-mentioned embodiments, please refer to Figure 3 , Figure 3A brief flowchart of an abnormality diagnosis method of an infusion pump is provided, specifically: after obtaining an operation parameter of the infusion pump, a critical characteristic value is extracted from the operation parameter, time series prediction is performed to obtain a predicted value, and a trend prediction characteristic value is calculated based on the predicted value; then the critical characteristic value and the trend prediction characteristic value are input into a fault classification model to obtain a fault classification result and a classification confidence corresponding to each fault classification result; finally, the residual characteristic value in the trend prediction characteristic value is standardized to obtain a first score, and the classification confidence output by the fault classification model is used as a second score, a comprehensive score is obtained by weighted fusion, and the running state of the infusion pump is determined according to the comprehensive score; when the comprehensive score reaches a preset score threshold, a running parameter calibration instruction is generated according to the fault classification result corresponding to the highest classification confidence, and the running parameter calibration instruction is sent to the infusion pump to make the infusion pump repair the abnormality.

[0125] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the abnormality diagnosis method of the infusion pump of the present application. More forms of simple transformation based on this technical concept, such as the interaction and combination of various embodiments, are within the protection scope of the present application.

[0126] The present application also provides an abnormality diagnosis system of an infusion pump, please refer to Figure 4 The abnormality diagnosis system of the infusion pump comprises an infusion pump and an abnormality diagnosis device of the infusion pump. Illustratively, the abnormality diagnosis device of the infusion pump can be a cloud platform, that is, the infusion pump collects operation parameters and uploads the operation parameters to the cloud platform for subsequent abnormality diagnosis processing, so as to determine the running state of the infusion pump.

[0127] The abnormality diagnosis system of the infusion pump provided by the present application adopts the abnormality diagnosis method of the infusion pump in the above-mentioned embodiments, which can solve the technical problem of low abnormality diagnosis efficiency when the existing abnormality diagnosis processing is performed on the infusion pump. Compared with the prior art, the beneficial effects of the abnormality diagnosis system of the infusion pump provided by the present application are the same as those of the abnormality diagnosis method of the infusion pump provided by the above-mentioned embodiments, and the other technical features in the abnormality diagnosis system of the infusion pump are the same as those disclosed in the above-mentioned embodiments, which will not be repeated here.

[0128] The present application provides an abnormality diagnosis device of an infusion pump, which comprises at least one processor and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the abnormality diagnosis method of the infusion pump in the first embodiment.

[0129] Reference will be made to Figure 5This document illustrates a structural schematic diagram of an abnormality diagnostic device suitable for implementing the embodiments of the present application for an infusion pump. The abnormality diagnostic device for the infusion pump in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The diagnostic device for abnormalities of the infusion pump shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0130] like Figure 5 As shown, the infusion pump malfunction diagnostic device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in the read-only memory 1002 or a program loaded from the storage device 1003 into the random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the infusion pump malfunction diagnostic device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: an input device 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; a storage device 1003 including, for example, magnetic tape, hard disk, etc.; and a communication device 1009. Communication device 1009 allows the infusion pump's fault diagnosis device to communicate wirelessly or wiredly with other devices to exchange data. While the figure shows fault diagnosis devices for infusion pumps with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0131] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program codes for executing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network through a communication device, or installed from the storage device 1003, or installed from the read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiments disclosed in the present application are executed.

[0132] The abnormality diagnosis device of the infusion pump provided by the present application adopts the abnormality diagnosis method of the infusion pump in the above-mentioned embodiments, and can solve the technical problem of low abnormality diagnosis efficiency in the prior art when the abnormality diagnosis of the infusion pump is processed. Compared with the prior art, the beneficial effects of the abnormality diagnosis device of the infusion pump provided by the present application are the same as those of the abnormality diagnosis method of the infusion pump provided by the above-mentioned embodiments, and other technical features in the abnormality diagnosis device of the infusion pump are the same as those disclosed in the above-mentioned embodiments, which will not be repeated here.

[0133] It should be understood that various parts of the present application can be realized by hardware, software, firmware or a combination thereof. In the description of the above-mentioned embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0134] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0135] The present application provides a computer readable storage medium having stored thereon computer readable program instructions (i.e. computer program) for executing the abnormality diagnosis method of the infusion pump in the above-mentioned embodiments.

[0136] The computer readable storage medium provided in the present application may, for example, be a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system or device, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM or flash memory), an optical fiber, a portable compact disk read only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present embodiment, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer readable storage medium can be transmitted in any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), and the like, or any suitable combination of the above.

[0137] The above computer readable storage medium can be contained in the abnormality diagnosis device of the infusion pump, or can exist separately without being assembled into the abnormality diagnosis device of the infusion pump.

[0138] The computer program code for performing the operations of the present application can be written in one or more programming languages or combinations of languages including object oriented programming languages such as Java, Smalltalk, C++ or conventional procedural programming languages such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0139] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0140] The modules involved in the embodiments of the present application can be implemented in the form of software or in the form of hardware. In some cases, the name of the module does not constitute a limitation on the module itself.

[0141] The computer readable storage medium provided by the present application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer program) for executing the above-mentioned infusion pump abnormality diagnosis method, and can solve the technical problem of low abnormality diagnosis efficiency in the prior art. Compared with the prior art, the computer readable storage medium provided by the present application has the same beneficial effects as the infusion pump abnormality diagnosis method provided by the above-mentioned embodiments, and will not be described here.

[0142] The above only describes some embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structural transformation made by referring to the content of the present application and the drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. An abnormality diagnosis method of an infusion pump, characterized by, The abnormality diagnosis method of the infusion pump comprises: obtaining an operating parameter of the infusion pump; time series prediction is performed on a critical characteristic value of the operating parameter to obtain a predicted value, and a trend prediction characteristic value is calculated based on the predicted value; the critical characteristic value and the trend prediction characteristic value are input into a pre-trained fault classification model to output a classification confidence corresponding to each fault classification result; based on the classification confidence and the trend prediction characteristic value, a comprehensive score is calculated, and the operating state of the infusion pump is determined according to the comprehensive score.

2. The method of claim 1, wherein, The trend prediction characteristic value comprises a predicted time characteristic value, a residual characteristic value and a predicted change rate characteristic value; the step of performing time series prediction on the critical characteristic value of the operating parameter to obtain a predicted value, and calculating a trend prediction characteristic value based on the predicted value, comprises: time series prediction is performed on the critical characteristic value of the operating parameter to obtain a first predicted value at a future time point, and a second predicted value at each time point between the current time point and the future time point; a first time point at which the second predicted value is first greater than or equal to a first preset threshold is determined, and a predicted time from the current time point to the first time point is calculated as the predicted time characteristic value; a difference between the first predicted value and the critical characteristic value is calculated as the residual characteristic value; a change rate of the critical characteristic value within a period from the current time point to the future time point is calculated as the predicted change rate characteristic value; the predicted time characteristic value, the residual characteristic value and the predicted change rate characteristic value are weighted and fused to obtain a trend prediction characteristic value.

3. The method of claim 1, wherein, Before the step of inputting the critical characteristic value and the trend prediction characteristic value into the pre-trained fault classification model, the following steps are included: obtaining historical data with historical fault type labels, wherein the historical data comprises historical critical characteristic values and historical trend prediction characteristics corresponding to the historical critical characteristic values; training an initial model by taking the historical critical characteristic values and the historical trend prediction characteristic values as input features and taking corresponding historical fault type labels as output targets until the accuracy of the fault classification results output by the initial model reaches a preset accuracy threshold to obtain the fault classification model.

4. The method of claim 1, wherein, After the step of obtaining the operating parameter of the infusion pump, the following steps are included: comparing the operating parameter with a preset safety threshold, wherein the preset safety threshold comprises a first preset threshold and a second preset threshold, and the first preset threshold is greater than the second preset threshold; if the operating parameter is less than the second preset threshold, it is determined that the infusion pump is normal; if the operating parameter is greater than or equal to the second preset threshold but less than the first preset threshold, the step of performing time series prediction on the critical characteristic value of the operating parameter to obtain a predicted value, and calculating a trend prediction characteristic value based on the predicted value, is executed; if the operating parameter is greater than or equal to the first preset threshold, it is determined that the infusion pump is severely abnormal, and an alarm signal is generated.

5. The method of claim 1, wherein, The step of calculating a comprehensive score based on the classification confidence and the trend prediction feature value comprises: standardizing a residual feature value in the trend prediction feature value to obtain a first score; taking the classification confidence as a second score; weighting and fusing the first score and the second score to obtain a comprehensive score.

6. The method of claim 1, wherein, The step of calculating a comprehensive score based on the classification confidence and the trend prediction feature value is followed by: if the comprehensive score reaches a preset score threshold, generating a running parameter calibration instruction according to the fault classification result corresponding to the highest classification confidence; issuing the running parameter calibration instruction to the infusion pump.

7. The method of claim 1, wherein, The method comprises: performing incremental training on the fault classification model at a first preset period to obtain an updated fault classification model; re-fitting model parameters of the time series prediction model at a second preset period to obtain an updated time series prediction model, wherein the time series prediction model is used for time series prediction on the critical feature value of the running parameter; performing abnormal diagnosis of the infusion pump through the updated time series prediction model and the updated fault classification model.

8. An abnormality diagnosing apparatus of an infusion pump, characterized by comprising: The device comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the abnormal diagnosis method of the infusion pump according to any one of claims 1 to 7.

9. An abnormality diagnosis system of an infusion pump characterized by comprising: The system comprises an infusion pump and an abnormal diagnosis device of the infusion pump according to claim 8.

10. A storage medium, characterized by The storage medium is a computer-readable storage medium, and the storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the abnormal diagnosis method of the infusion pump according to any one of claims 1 to 7.

Images