Infusion pump motor fault detection method based on adaptive weight SVM model

Abstract

The invention discloses an infusion pump motor fault detection method based on an adaptive weight SVM model, and relates to the field of intelligent diagnosis of motor faults, and the method comprises the following steps: obtaining a sample set; calculating initial weights of different types of faults based on the fault frequency and the fault severity score; constructing a motor fault detection SVM model, and defining a target function based on the initial weight; training a motor fault detection SVM model by using the sample set; updating the weight of the corresponding type of fault based on misclassification loss; on the basis of the updated weight optimization objective function, constructing a Lagrange function and solving the Lagrange function; and inputting operation characteristics of a to-be-detected infusion pump motor into the motor fault detection SVM model, and outputting a detection result. According to the method, different initial weights are set for different types of faults, so that the classification accuracy of the model on key faults can be improved. A self-adaptive weight mechanism is adopted, the weight is adjusted according to misclassification loss, and the generalization ability of the model is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent diagnosis of motor faults, and in particular to a method for detecting motor faults of an infusion pump based on an adaptive weighted SVM model. Background Art

[0002] An infusion pump is an instrument that accurately controls the number of infusion drops or flow rate, ensuring that the drug enters the patient's body evenly and accurately, ensuring that it is safe and effective. It is often used in situations where strict control of infusion volume and drug dosage is required, such as when administering pressor drugs, antiarrhythmic drugs, and intravenous infusions or intravenous anesthesia for infants. One of the core components of an infusion pump is the stepper motor that drives the pump system. The stepper motor precisely controls the infusion rate and volume by controlling its rotational speed and number of rotations. When an infusion pump motor fails, it must be detected and addressed immediately. However, existing fault detection methods primarily rely on threshold judgment or simple statistical analysis, which have limitations when detecting complex faults. Prior art CN111474476A discloses a motor fault prediction method that primarily relies on a support vector machine (SVM) model for motor fault diagnosis. However, this method does not account for the imbalance in the sample dataset for motor faults, where some types of faults have far more samples than others, and sets the same penalty for all fault types. This can lead to the classifier tending to classify the test samples as belonging to the fault type with a large number of samples. This is because during the training process, the model is more influenced by the majority class samples and ignores the characteristics of the minority class samples. This is especially difficult for the model to accurately identify faults that occur infrequently (with a small number of samples) but are very serious. Summary of the Invention

[0003] The purpose of the present invention is to improve the accuracy of motor fault detection SVM model for infusion pump motor fault detection.

[0004] To achieve the above object, the present invention provides an infusion pump motor fault detection method based on an adaptive weighted SVM model, the method comprising the following steps:

[0005] Step 1: Obtain historical data of the infusion pump motor operating characteristics and the corresponding infusion pump motor operating status to obtain a sample set;

[0006] Step 2: Calculate the initial weights of different types of faults based on fault frequency and fault severity scores;

[0007] Step 3: Build a motor fault detection SVM model and define an objective function based on the initial weights;

[0008] Step 4: Using the sample set to train the motor fault detection SVM model, and calculating the misclassification loss of fault detection;

[0009] Step 5: Update the weight of the corresponding type of fault based on the misclassification loss;

[0010] Step 6: Optimize the objective function based on the updated weights, construct and solve the Lagrangian function, and obtain the classification hyperplane parameters;

[0011] Step 7: Input the operating characteristics of the infusion pump motor to be detected into the motor fault detection SVM model, and output the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters.

[0012] Among them, the support vector machine (SVM) algorithm is a data mining method based on statistical learning theory. Its mechanism is to find a classification hyperplane that meets the classification requirements. In theory, it can achieve the optimal classification of linearly separable data. For infusion pump motors, there are many possible types of faults, such as rotation faults, noise faults, temperature faults, etc. The frequency and severity of these different types of faults are different. In order to enable the model to learn the boundaries of critical faults, the present invention sets different initial weights for different types of faults according to the fault frequency and fault severity score when defining the objective function of the motor fault detection SVM model, so that different types of faults have different penalty degrees, forcing the model to effectively take into account faults with a low sample size. In the subsequent training process of the model, the weights are reversely adjusted according to the misclassification loss, and the weights of faults that are continuously misclassified are increased to optimize their classification boundaries. Finally, the objective function is optimized based on the updated weights and the Lagrangian function is constructed to solve the decision boundary that can maximize the classification interval and minimize the weighted error, thereby improving the detection accuracy and adaptability of the motor fault detection SVM model.

[0013] Furthermore, in step 1, the operating characteristics of the infusion pump motor include the motor output pressure, and the motor output pressure is obtained by the following steps:

[0014] Use pressure sensors to collect the measured pressure of pipeline liquid in real time;

[0015] Obtaining a vertical height difference between the infusion pump and the infusion site of the patient, and calculating a compensation pressure based on the vertical height difference and the density of the drug solution;

[0016] The motor output pressure is calculated based on the measured pressure and the compensated pressure.

[0017] Among them, in the training process of the motor fault detection SVM model, the model distinguishes the different operating states of the motor by learning the differences in the operating characteristics of the motor when it is normal and when it fails. The motor output pressure, as a key operating feature, can reflect the health status of the motor together with other features. However, during the use of the infusion pump, the measured pressure of the pipeline liquid collected by the pressure sensor often cannot reflect the real motor output pressure. This is because: the height difference between the infusion pump and the patient's infusion site will cause the liquid in the infusion pipeline to generate a certain static pressure, and the measured pressure actually includes this part of the static pressure. The present invention obtains the vertical height difference between the infusion pump and the patient's infusion site in real time, and then uses the vertical height difference and the density of the liquid medicine to calculate the compensation pressure, and finally obtains the accurate motor output pressure, which provides more reliable data input for the training of the motor fault detection SVM model.

[0018] Furthermore, in step 1, the infusion pump motor operation characteristics further include: motor phase current effective ripple characteristics, and the motor phase current effective ripple characteristics are obtained by the following steps:

[0019] monitoring the dynamic changes of the motor output pressure, and when the motor output pressure suddenly changes, calculating the rate of change of the motor output pressure over time during the sudden change period to generate a pressure change rate sequence;

[0020] synchronously collecting current waveform data of the infusion pump motor winding during the mutation period, calculating the current fluctuation rate within each driving cycle based on the current waveform data, and generating a current fluctuation rate sequence;

[0021] A time domain correlation coefficient between the pressure change rate sequence and the current fluctuation rate sequence is calculated. If the time domain correlation coefficient exceeds a preset coefficient threshold, the current fluctuation rate sequence is marked as the motor phase current effective ripple feature.

[0022] The motor used in the infusion pump is a stepper motor. Under low speed and sudden load changes, it may experience micro-stepping failures. Micro-stepping failures of the motor can lead to inaccurate delivery of the drug solution, affecting the infusion accuracy and bringing adverse consequences. Such failures are usually ignored by traditional motor detection models. The present invention focuses on the coupling characteristics between the current fluctuation rate and the motor output pressure. By screening the effective ripple characteristics of the motor phase current caused by the liquid pipeline pressure disturbance through the time domain correlation of the pressure change rate sequence and the current fluctuation rate sequence, it uses this as the characteristic dimension of micro-stepping failures, thereby improving the sensitivity of the motor fault detection SVM model to early stepping failures.

[0023] Furthermore, in step 1, the infusion pump motor operation characteristics further include: dynamic characteristics of the patient's infusion site, and the dynamic characteristics of the patient's infusion site are obtained by the following steps:

[0024] Real-time acquisition of the displacement duration of the patient's infusion site and the three-dimensional spatial displacement acceleration of the patient's infusion site;

[0025] Obtaining the instantaneous displacement of the patient's infusion site based on the three-dimensional spatial displacement acceleration integral calculation;

[0026] If the displacement duration is lower than a preset time threshold and the instantaneous displacement exceeds a preset displacement threshold, the displacement duration and the instantaneous displacement are marked as dynamic features of the patient's infusion site.

[0027] Among them, sudden changes in the patient's body position during the infusion process, such as suddenly raising or letting go of hands, turning over, etc., may cause the infusion tube to be pulled more violently, resulting in slight vibrations in the infusion pump. This vibration may be misjudged as a motor failure after being detected by the vibration sensor. Therefore, the present invention identifies this type of interference by detecting the dynamic characteristics of the patient's infusion site. Specifically, when the displacement duration is lower than the preset time threshold and the instantaneous displacement exceeds the preset displacement threshold, it means that the patient's infusion site has undergone a position change in a short period of time that may cause the infusion pump to vibrate. At this time, the displacement duration and the instantaneous displacement are input as dynamic characteristics of the patient's infusion site, so that the motor fault detection SVM model can identify such vibrations and improve the accuracy of detection.

[0028] Furthermore, in step 2, the calculation of initial weights of different types of faults based on fault frequency and fault severity scores includes:

[0029] Calculate failure frequencies based on historical records;

[0030] Assessing the severity of the fault based on the threat level of the fault to patient safety, the repair cost level of the infusion pump motor, and the downtime level to obtain the fault severity score;

[0031] The initial weight of the fault is calculated according to the following formula:

[0032]

[0033] in, is the initial weight of the i-th type fault, F i is the fault frequency of type i fault, S i Score the fault severity of the i-th fault.

[0034] Three dimensions—the threat level to patient safety, the repair cost of the infusion pump motor, and the downtime—can comprehensively and accurately reflect the severity of a particular fault type. Calculating initial weights for different fault types based on fault frequency and severity scores can balance the impact of uneven dataset distribution. For example, low-frequency but high-severity faults can also receive a certain initial weight, encouraging the classifier to pay more attention to these faults during training, thereby improving recognition accuracy.

[0035] Furthermore, in step three, the objective function is:

[0036]

[0037] Among them, w is the weight coefficient, b is the bias, ξ i is the slack variable, C is the regularization parameter, n is the number of samples, x i is the characteristic vector of the i-th type fault, y i is the output value of the i-th fault, w·x i +b is the classification hyperplane equation.

[0038] The objective function consists of two parts: maximizing the margin term (the larger the margin, the stronger the model's generalization ability); and minimizing the loss term, which aims to handle noise and outliers by allowing some data points to violate the margin constraint to a certain extent. These two components are weighed against each other via a regularization parameter to ensure a balance between accuracy and generalization. Combining these two components, the SVM objective function can be formulated as a constrained optimization problem. Solving this problem can find the optimal classification hyperplane parameters w and b.

[0039] Furthermore, in step 4, the calculation formula for the misclassification loss of the fault detection is:

[0040] L i =max(0,1-y i (w·x i +b));

[0041] Among them, L i is the misclassification loss of the i-th fault.

[0042] Furthermore, in step 5, the formula for updating the weight of the corresponding type of fault based on the misclassification loss is:

[0043]

[0044] Among them, η is the learning rate, t is the number of iterations, α iis the weight of the i-th type of fault. Misclassification loss refers to the loss value assigned by the SVM model to the misclassified samples during the classification process. By minimizing the misclassification loss, the model can continuously adjust its parameters to better adapt to the training data, thereby improving the classification accuracy.

[0045] Furthermore, in step six, the constructed Lagrangian function is:

[0046]

[0047] Among them, λ i is the Lagrange multiplier. By constructing the Lagrange function, the objective function and constraints can be integrated into one function, thus simplifying the problem solving process.

[0048] Furthermore, in step seven, the fault decision classification expression for outputting the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters is:

[0049]

[0050] Among them, k(x i , x) is the kernel function of the first motor fault detection SVM model, γ is the kernel function parameter, and x is the eigenvector of the infusion pump motor to be detected.

[0051] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0052] 1. This invention assigns different initial weights to different fault types based on fault frequency and severity scores. This allows the model to effectively handle imbalanced infusion pump motor fault samples and improves the classification accuracy of critical faults. This invention also employs an adaptive weighting mechanism, adjusting weights based on misclassification loss during model training to optimize the model and improve its generalization capabilities.

[0053] 2. To address the problem of micro-stepping caused by sudden changes in pipeline pressure when the infusion pump motor is running at low speed, the present invention innovatively adopts the effective ripple characteristics of the motor phase current as its exclusive detection dimension, thereby improving the sensitivity of the motor detection SVM model to motor micro-stepping faults.

[0054] 3. The present invention reflects the changes in the patient's body position during the infusion process by detecting the dynamic characteristics of the patient's infusion site, preventing the vibration of the infusion pump caused by the patient's inadvertent pulling of the infusion line from being diagnosed as a motor failure, thereby improving the accuracy of model detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation of the embodiments of the present invention;

[0056] Figure 1 It is a flow chart of an infusion pump motor fault detection method based on an adaptive weighted SVM model in the present invention. DETAILED DESCRIPTION

[0057] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features therein can be combined with each other without conflict.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0059] Example 1

[0060] Please refer to Figure 1 The first embodiment of the present invention provides an infusion pump motor fault detection method based on an adaptive weighted SVM model, the method comprising the following steps:

[0061] Step 1: Obtain historical data of the infusion pump motor operating characteristics and the corresponding infusion pump motor operating status to obtain a sample set;

[0062] Step 2: Calculate the initial weights of different types of faults based on fault frequency and fault severity scores;

[0063] Step 3: Build a motor fault detection SVM model and define an objective function based on the initial weights;

[0064] Step 4: Using the sample set to train the motor fault detection SVM model, and calculating the misclassification loss of fault detection;

[0065] Step 5: Update the weight of the corresponding type of fault based on the misclassification loss;

[0066] Step 6: Optimize the objective function based on the updated weights, construct and solve the Lagrangian function, and obtain the classification hyperplane parameters;

[0067] Step 7: Input the operating characteristics of the infusion pump motor to be detected into the motor fault detection SVM model, and output the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters.

[0068] Among them, in step one, the operating characteristics of the infusion pump motor may include conventional operating characteristics, such as parameters such as pipeline flow, motor voltage, current, and the like, as well as statistical characteristics of these parameters, such as mean, variance, skewness, kurtosis, and frequency domain characteristics, such as power spectral density, etc., which can be selected by those skilled in the art according to actual conditions. In this embodiment, data preprocessing is performed before feature extraction, and the data preprocessing includes filtering high-frequency noise and normalization processing. The operating status of the infusion pump motor includes a fault state and a healthy state.

[0069] In this embodiment, in step 1, the operating characteristics of the infusion pump motor include the motor output pressure, and the motor output pressure is obtained by the following steps:

[0070] Use pressure sensors to collect the measured pressure of pipeline liquid in real time;

[0071] Obtaining a vertical height difference between the infusion pump and the infusion site of the patient, and calculating a compensation pressure based on the vertical height difference and the density of the drug solution;

[0072] The motor output pressure is calculated based on the measured pressure and the compensated pressure.

[0073] The vertical height difference between the infusion pump and the patient's infusion site can be measured using existing computer vision technology, that is, first using a camera to capture images of the infusion pump and the patient's infusion site, identifying the images and obtaining their three-dimensional coordinates, and then calculating the vertical height difference between the two based on the three-dimensional coordinates. The calculation formula for the compensation pressure is as follows:

[0074] P 补偿 =ρgh;

[0075] Among them, P 补偿 is the compensation pressure, ρ is the density of the infusion liquid, g is the acceleration due to gravity, and h is the vertical height difference between the infusion pump and the patient's infusion site. The motor output pressure is obtained by subtracting the compensation pressure from the measured pressure.

[0076] In this embodiment, in step 1, the infusion pump motor operation characteristics further include: motor phase current effective ripple characteristics, and the motor phase current effective ripple characteristics are obtained by the following steps:

[0077] monitoring the dynamic changes of the motor output pressure, and when the motor output pressure suddenly changes, calculating the rate of change of the motor output pressure over time during the sudden change period to generate a pressure change rate sequence;

[0078] synchronously collecting current waveform data of the infusion pump motor winding during the mutation period, calculating the current fluctuation rate within each driving cycle based on the current waveform data, and generating a current fluctuation rate sequence;

[0079] A time domain correlation coefficient between the pressure change rate sequence and the current fluctuation rate sequence is calculated. If the time domain correlation coefficient exceeds a preset coefficient threshold, the current fluctuation rate sequence is marked as the motor phase current effective ripple feature.

[0080] When the change in the motor output pressure exceeds the preset pressure change threshold within the preset mutation time threshold, it is determined that the motor output pressure has undergone a mutation. At this time, the rate of change of the motor output pressure over time is synchronously recorded. For example, if a sudden change in the motor output pressure is detected at a certain moment, the rate of change of the motor output pressure relative to a fixed time interval Δt during the mutation period is recorded to generate a pressure change rate series: Where ΔP is the pressure change within a fixed time interval Δt, N is an integer.

[0081] The current fluctuation rate is calculated as follows: Where δ is the current fluctuation rate, and ΔI is the sum of the absolute values of the differences between the adjacent current peaks and valleys of the infusion pump motor windings within each motor drive cycle. For example, three peak-valley pairs are measured in a certain drive cycle during the mutation period: 0.8A→0.2A, 0.7A→0.15A, and 0.75A→0.18A. Then ΔI==|0.8-0.2|+|0.7-0.15|+|0.75-0.18|=1.67A, I A is the average current magnitude of this cycle. The current fluctuation rate sequence during the mutation period is: Where n is the number of motor drive cycles within the mutation period.

[0082] The time domain correlation coefficient may be a Pearson correlation coefficient or a Spearman rank correlation coefficient, etc. The preset coefficient threshold may be set by those skilled in the art according to actual conditions.

[0083] In this embodiment, in step 1, the infusion pump motor operation characteristics further include: dynamic characteristics of the patient's infusion site, and the dynamic characteristics of the patient's infusion site are obtained by the following steps:

[0084] Real-time acquisition of the displacement duration of the patient's infusion site and the three-dimensional spatial displacement acceleration of the patient's infusion site;

[0085] Obtaining the instantaneous displacement of the patient's infusion site based on the three-dimensional spatial displacement acceleration integral calculation;

[0086] If the displacement duration is lower than a preset time threshold and the instantaneous displacement exceeds a preset displacement threshold, the displacement duration and the instantaneous displacement are marked as dynamic features of the patient's infusion site.

[0087] Among them, the three-dimensional spatial displacement acceleration of the patient's infusion site can be measured by using an inertial measurement unit (IMU) fixed next to the infusion site, and the displacement duration can be calculated by the timestamp attached to the inertial measurement unit (IMU). When calculating the instantaneous displacement of the patient's infusion site based on the three-dimensional spatial displacement acceleration integral, the acceleration data is first integrated once to obtain velocity data. Since the acceleration data is usually discrete, this embodiment uses a numerical integration method (such as the trapezoidal method, the Simpson method, etc.) for calculation. Then the velocity data is integrated twice to obtain the displacement data, and finally the instantaneous displacement is obtained. For the preset time threshold and the preset displacement threshold, those skilled in the art can set them according to actual conditions, and the present invention does not limit this.

[0088] In step 2, the calculation of the initial weights of different types of faults based on the fault frequency and fault severity scores includes:

[0089] Calculate failure frequencies based on historical records;

[0090] Assessing the severity of the fault based on the threat level of the fault to patient safety, the repair cost level of the infusion pump motor, and the downtime level to obtain the fault severity score;

[0091] The initial weight of the fault is calculated according to the following formula:

[0092]

[0093] in, is the initial weight of the i-th type fault, F i is the fault frequency of type i fault, S i Score the fault severity of the i-th fault.

[0094] The fault frequency can be calculated by dividing the number of faults by the total running time.

[0095] The threat level of a failure to patient safety can be graded from Level 1 (no threat) to Level 5 (extreme threat) based on factors such as the degree of patient harm that the failure may cause and the severity of treatment interruption. Risk assessment tools and methods such as FMEA and HAZOP can be used to assist in the assessment.

[0096] The maintenance cost of the infusion pump motor can be analyzed from the cost of parts and labor required for maintenance, combined with factors such as the model and manufacturer of the infusion pump.

[0097] Finally, the fault severity score is calculated by weighted average method.

[0098] For example, assume that there are two types of faults. The first type of fault has a failure frequency of 4 times / 1000 hours, and the second type of fault has a failure frequency of 10 times / 1000 hours.

[0099] After assessment, the threat level to patient safety, the maintenance cost level and the downtime level of the infusion pump motor of the first type of failure are 5, 2 and 4 respectively, and the second type of failure are 3, 1 and 2 respectively.

[0100] First, calculate the fault severity score. Assuming that the threat level of the fault to patient safety is weighted at 0.5, and the maintenance cost level and downtime level of the infusion pump motor are weighted at 0.25 respectively, then the weighted calculated fault severity score for the first type of fault is 4, and the fault severity score for the second type of fault is 2.25.

[0101] Then, the initial weights of the first and second type of faults are calculated by combining the fault frequency and fault severity scores to be 0.16 and 0.225, respectively. After normalization, they are 0.42 and 0.58, respectively.

[0102] Among them, in step 3, the objective function is:

[0103]

[0104] Among them, w is the weight coefficient, b is the bias, ξ i is the slack variable, C is the regularization parameter, n is the number of samples, x i is the characteristic vector of the i-th type fault, y i is the output value of the i-th fault, w·x i +b is the classification hyperplane equation.

[0105] In step 4, the calculation formula for the misclassification loss of the fault detection is:

[0106] L i =max(0,1-y i (w·x i +b));

[0107] Among them, L i is the misclassification loss of the i-th fault.

[0108] Among them, in step 5, the formula for updating the weight of the corresponding type of fault based on the misclassification loss is:

[0109]

[0110] Among them, η is the learning rate, t is the number of iterations, α i is the weight of the i-th type of fault.

[0111] Among them, in step 6, the Lagrangian function constructed is:

[0112]

[0113] Among them, λ i is the Lagrange multiplier.

[0114] Among them, in step seven, the fault decision classification expression for outputting the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters is:

[0115]

[0116] Among them, k(x i , x) is the kernel function of the first motor fault detection SVM model, γ is the kernel function parameter, which is obtained through cross-validation in this embodiment. x is the feature vector of the infusion pump motor to be tested. When f(x) = 0, the infusion pump motor is currently in a healthy state. When f(x) = 1, it indicates that the infusion pump motor is currently in a faulty state.

[0117] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0118] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for detecting infusion pump motor faults based on an adaptive weighted SVM model, characterized in that: The method comprises the following steps: Step 1: Obtain historical data of the infusion pump motor operating characteristics and the corresponding infusion pump motor operating status to obtain a sample set; Step 2: Calculate the initial weights of different types of faults based on fault frequency and fault severity scores; Step 3: Build a motor fault detection SVM model and define an objective function based on the initial weights; Step 4: Using the sample set to train the motor fault detection SVM model, and calculating the misclassification loss of fault detection; Step 5: Update the weight of the corresponding type of fault based on the misclassification loss; Step 6: Optimize the objective function based on the updated weights, construct and solve the Lagrangian function, and obtain the classification hyperplane parameters; Step 7: Input the operating characteristics of the infusion pump motor to be detected into the motor fault detection SVM model, and output the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters.

2. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 1 is characterized in that: In step 1, the infusion pump motor operating characteristics include the motor output pressure, and the motor output pressure is obtained by the following steps: Use pressure sensors to collect the measured pressure of pipeline liquid in real time; Obtaining a vertical height difference between the infusion pump and the infusion site of the patient, and calculating a compensation pressure based on the vertical height difference and the density of the drug solution; The motor output pressure is calculated based on the measured pressure and the compensated pressure.

3. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 2 is characterized in that: In step 1, the infusion pump motor operation characteristics further include: motor phase current effective ripple characteristics, and the motor phase current effective ripple characteristics are obtained by the following steps: monitoring the dynamic changes of the motor output pressure, and when the motor output pressure suddenly changes, calculating the rate of change of the motor output pressure over time during the sudden change period to generate a pressure change rate sequence; synchronously collecting current waveform data of the infusion pump motor winding during the mutation period, calculating the current fluctuation rate within each driving cycle based on the current waveform data, and generating a current fluctuation rate sequence; A time domain correlation coefficient between the pressure change rate sequence and the current fluctuation rate sequence is calculated. If the time domain correlation coefficient exceeds a preset coefficient threshold, the current fluctuation rate sequence is marked as the motor phase current effective ripple feature.

4. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 2 is characterized in that: In step 1, the infusion pump motor operation characteristics further include: dynamic characteristics of the patient's infusion site, and the dynamic characteristics of the patient's infusion site are obtained by the following steps: Real-time acquisition of the displacement duration of the patient's infusion site and the three-dimensional spatial displacement acceleration of the patient's infusion site; Obtaining an instantaneous displacement of the patient's infusion site based on the three-dimensional spatial displacement acceleration calculation; If the displacement duration is lower than a preset time threshold and the instantaneous displacement exceeds a preset displacement threshold, the displacement duration and the instantaneous displacement are marked as dynamic features of the patient's infusion site.

5. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 1 is characterized in that: In step 2, the initial weights of different types of faults are calculated based on the fault frequency and fault severity scores, including: Calculate failure frequencies based on historical records; Assessing the severity of the fault based on the threat level of the fault to patient safety, the repair cost level of the infusion pump motor, and the downtime level to obtain the fault severity score; The initial weight of the fault is calculated according to the following formula: in, is the initial weight of the i-th type fault, F i is the fault frequency of type i fault, S i Score the fault severity of the i-th fault.

6. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 5 is characterized in that: In step three, the objective function is: style i (w x i +b)≥1-ξ i ,x i ≥0 Among them, w is the weight coefficient, b is the bias, ξ i is the slack variable, C is the regularization parameter, n is the number of samples, x i is the characteristic vector of the i-th type fault, y i is the output value of the i-th fault, w·x i +b is the classification hyperplane equation.

7. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 6 is characterized in that: In step 4, the calculation formula for the misclassification loss of the fault detection is: L i =max(0,1-y i (w·x i +b)); Among them, L i is the misclassification loss of the i-th fault.

8. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 6 is characterized in that: In step 5, the formula for updating the weight of the corresponding type of fault based on the misclassification loss is: Among them, η is the learning rate, t is the number of iterations, α i is the weight of the i-th type of fault.

9. The method for detecting motor faults of an infusion pump based on an adaptive weighted SVM model according to claim 8, characterized in that: In step 6, the constructed Lagrangian function is: Among them, λ i is the Lagrange multiplier.

10. The infusion pump motor fault detection method based on the adaptive weighted SVM model according to claim 9, characterized in that: In step seven, the fault decision classification expression for outputting the fault detection result of the infusion pump motor to be detected according to the classification hyperplane parameters is: Among them, k(x i , x) is the kernel function of the first motor fault detection SVM model, γ is the kernel function parameter, and x is the eigenvector of the infusion pump motor to be detected.

Images