Intelligent monitoring method for ultracentrifuge tube based on sensor network

By deploying sensor networks and edge computing systems in ultracentrifuges, combining multi-source data fusion and closed-loop control, the shortcomings of existing monitoring methods are solved, precise monitoring and adaptive management of the centrifuge tube status are achieved, and equipment safety and experimental reliability are improved.

CN120346921AActive Publication Date: 2025-07-22KIRGEN BIOSCIENCE (SHANGHAI) CO LTD

Patent Information

Application Number
CN202510837479.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-23
Publication Date
2025-07-22
Estimated Expiration
2045-06-23

AI Technical Summary

Technical Problem

The monitoring methods of existing ultracentrifuges rely on a single sensor and cannot fully capture the changes in the state of the centrifuge tube, resulting in a high false alarm rate and lack of adaptive adjustment capabilities, making it difficult to meet the needs of high-precision experiments.

Method used

Deploy sensor networks, including three-axis acceleration sensors, temperature sensors, pressure sensors and liquid level sensors, data acquisition and preprocessing through edge computing gateways, build centrifugal force-stress and thermal conduction models, use extended Kalman filtering and random forest classifiers for multi-source data fusion and abnormal identification, and combine with PID controller for closed-loop feedback control.

Benefits of technology

It realizes multi-dimensional accurate monitoring, improves the accuracy and adaptability of abnormal detection, reduces the false alarm rate, ensures equipment safety and sample stability, and extends the service life of the equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120346921A_ABST
    Figure CN120346921A_ABST
Patent Text Reader

Abstract

The invention provides an ultracentrifuge tube intelligent monitoring method based on a sensor network, and belongs to the technical field of ultracentrifuge tubes, the method comprises the following steps: S1, deploying and initializing the sensor network; s2, carrying out real-time data acquisition and preprocessing; s3, building a centrifugal tube physical model and calibrating parameters; s4, multi-source data fusion and feature extraction; s5, abnormal state identification and early warning decision making; and S6, closed-loop feedback control and parameter self-adaption are carried out. By acquiring multi-dimensional data such as vibration, temperature, internal pressure and liquid level change of the centrifugal tube, key state parameters in the operation process of the centrifugal tube are comprehensively covered, and compared with traditional single sensor monitoring, the integrity of monitoring information is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of ultracentrifuge tubes, and in particular to an intelligent monitoring method for ultracentrifuge tubes based on a sensor network. Background Art

[0002] In the fields of biomedicine, chemical analysis, etc., an ultracentrifuge is a key device for separating and purifying samples such as biomolecules and cell components. As a core consumable, an ultracentrifuge tube needs to maintain structural integrity and sample stability under high-speed rotation (the rotation speed can reach tens of thousands of revolutions per minute). However, there are significant deficiencies in existing monitoring means: on the one hand, traditional monitoring mostly relies on a single type of sensor (such as only monitoring the rotation speed or temperature), and it is impossible to comprehensively capture the state changes of the centrifuge tube under complex working conditions. For example, situations such as liquid leakage and abnormal vibration caused by imbalance are difficult to be detected in time; on the other hand, existing monitoring methods mostly adopt simple logics such as threshold judgment, lacking in-depth fusion and analysis of multi-source data, unable to accurately identify potential risks, and prone to false alarms or missed alarms, resulting in equipment failures and even sample losses. In addition, the monitoring systems in the existing technologies usually do not have the ability of adaptive adjustment, unable to dynamically optimize operation parameters according to the real-time state of the centrifuge tube, and it is difficult to meet the requirements of high-precision experiments and production. Summary of the Invention

[0003] The present invention provides an intelligent monitoring method for ultracentrifuge tubes based on a sensor network to solve one or more of the above problems.

[0004] To achieve the above object, the present invention adopts the following technical solutions: An intelligent monitoring method for ultracentrifuge tubes based on a sensor network, comprising: S1. Deploy a sensor network and initialize it; deploy multi-type sensors on the fixed base, rotating shaft, inner wall of the centrifugation chamber and the surface of each centrifuge tube of the centrifugation device to form a three-dimensional distributed monitoring network. The multi-type sensors include triaxial acceleration sensors, temperature sensors, pressure sensors and liquid level sensors; connect all sensors to the edge computing gateway by wired or wireless means; S2. Perform real-time data collection and preprocessing; the edge computing gateway synchronously collects all sensor data, uses a low-pass filter to remove high-frequency noise from the original acceleration signal, eliminates random interference from the original temperature signal through moving average filtering, and eliminates outliers from the original pressure and liquid level signals through threshold verification; S3. Construct a physical model of the centrifuge tube and parameter calibration; establish a centrifugal force-stress model and a heat conduction model according to the specification parameters of the centrifuge tube, calibrate the model parameters through historical data, fit the vibration transfer function of the bracket through no-load operation data, and determine the relationship between liquid sloshing and liquid level through load experiments; S4. Multi-source data fusion and feature extraction: Spatially and temporally align the preprocessed sensor data with the output of the centrifugal force-stress model, use the extended Kalman filter to fuse multi-source information, and estimate the real-time state vector of the centrifuge tube; extract feature parameters based on the real-time state vector of the centrifuge tube, including the rotational speed change rate, temperature rise rate, pressure fluctuation coefficient, vibration energy ratio, liquid level change gradient, and centrifugal force deviation rate; S5. Abnormal state identification and early warning decision-making: Input the feature parameters into the random forest classifier to identify the state categories as normal, slight vibration, overheating, overload, and liquid leakage; if it is determined to be abnormal, trigger a first-level warning, a second-level warning, or a third-level warning according to the preset rules; S6. Closed-loop feedback control and parameter adaptation: If a first-level warning or a second-level warning is detected, the edge computing gateway sends a control instruction to the centrifuge equipment to adjust the motor speed through the PID controller; at the same time, dynamically update the parameters of the support vibration transfer function, the centrifugal force-stress model, or the heat conduction model according to the real-time monitoring data.

[0005] In this specification, the triaxial acceleration sensor is pasted on the central axis of the side wall of the centrifuge tube, the temperature sensor is embedded at the bottom of the centrifuge tube, the pressure sensor is integrated on the inner side of the tube cap, and the liquid level sensor is fixed at the 1 / 2 height of the tube body by an infrared pair-tube structure.

[0006] In this specification, the low-pass filter used for the original acceleration signal is a Butterworth low-pass filter.

[0007] In this specification, the centrifugal force-stress model is constructed based on the centrifugal force formula, and the heat conduction model is established according to the specific heat capacity and heat dissipation coefficient of the tube body material.

[0008] In this specification, the state transition function of the extended Kalman filter is determined according to the change law of the centrifuge tube rotational speed, and the process noise covariance matrix is obtained through sensor error statistics.

[0009] In this specification, the training samples of the random forest classifier include normal working condition data and historical data of various abnormal working conditions.

[0010] In this specification, for the first-level warning, a warning message is displayed on the equipment screen; for the second-level warning, a text message is sent to the monitoring terminal and the abnormal log is recorded; for the third-level warning, the equipment shutdown procedure is automatically triggered to cut off the power supply of the motor.

[0011] In this specification, the proportional, integral, and derivative coefficients of the PID controller are tuned by the critical ratio method.

[0012] In this specification, in S6, when the liquid level change exceeds 5%, recalculate the total mass of the centrifuge tube, including the mass of the sample liquid and the mass of the tube body.

[0013] In this specification, in S6, by comparing the residuals between the actual monitoring data and the model prediction data, when the residuals exceed a certain threshold, the model parameters are recalibrated.

[0014] In summary, the present invention has at least the following beneficial effects: Multi-dimensional precise monitoring: By deploying various sensors such as triaxial acceleration, temperature, pressure, and liquid level at key positions of the centrifugal equipment and centrifuge tubes to construct a three-dimensional distributed monitoring network, multi-dimensional data such as the vibration, temperature, internal pressure, and liquid level changes of the centrifuge tubes can be obtained in real time, comprehensively covering the key state parameters during the operation of the centrifuge tubes. Compared with traditional single-sensor monitoring, the integrity of monitoring information is improved. Intelligent anomaly recognition: By integrating algorithms such as non-negative matrix factorization (NMF), radial basis function (RBF) surface fitting, and random forest classification, dimensionality reduction, non-linear modeling, and feature fusion of multi-source data are performed, and various abnormal states such as normal, slight vibration, overheating, overload, and liquid leakage can be accurately identified. Compared with traditional threshold methods, the accuracy of anomaly detection is improved, and the false alarm and missed alarm rates are effectively reduced. Dynamic adaptive control: Based on the anomaly detection results, the system can automatically execute multi-level early warning and closed-loop control strategies. At the first-level early warning, the sampling frequency is increased to closely track the abnormal trend; at the second-level early warning, the rotation speed is accurately adjusted through a PID controller; at the third-level early warning, an emergency stop is performed to avoid equipment damage and sample loss. At the same time, the physical model parameters and algorithm models are dynamically updated according to real-time data, enabling the system to adapt to different working conditions and sample loads and shortening the control response time. Full-process data-driven: From sensor data acquisition, preprocessing, feature extraction to model construction and decision control, each link is closely related and the data is traceable, forming a complete data-driven monitoring system. It not only improves the safety and stability of the centrifugation process, but also provides rich data support for subsequent process optimization and equipment maintenance, extends the service life of the equipment, and reduces the sample loss cost. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a schematic diagram of the intelligent monitoring method for ultra-high-speed centrifuge tubes based on a sensor network involved in the present invention. Figure 2 It is a schematic diagram of the sensor deployment and initialization process involved in the present invention. Figure 3Schematic diagram of the abnormal state recognition and early warning process involved in the present invention. Detailed implementation manners

[0017] In the following, only some exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and the description are considered to be exemplary in nature rather than restrictive.

[0018] The following disclosure provides many different implementation manners or examples for implementing different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, the components and settings of specific examples are described below. Of course, they are only examples and are not intended to limit the embodiments of the present invention. In addition, the embodiments of the present invention may repeat reference numerals and / or reference letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various implementation manners and / or settings discussed.

[0019] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] As Figure 1 shown, this embodiment provides an intelligent monitoring method for ultracentrifuge tubes based on a sensor network, including the following steps: S1. Deploy the sensor network and initialize it; deploy triaxial acceleration sensors, temperature sensors, pressure sensors, and liquid level sensors inside the cavity of the ultracentrifugation device, on the centrifuge tube bracket, and on the outer wall of each centrifuge tube respectively; connect all sensors to the edge computing gateway in a wired or wireless manner, and complete the registration of hardware addresses and the initialization of communication protocols; S2. Perform real-time data acquisition and preprocessing; the edge computing gateway synchronously acquires all sensor data at a set frequency, uses a low-pass filter to remove high-frequency noise from the original acceleration signal, eliminates random interference from the original temperature signal through moving average filtering, and eliminates outliers from the original pressure and liquid level signals through threshold verification; S3. Construct a physical model of the centrifuge tube and parameter calibration; establish a centrifugal force model and a heat conduction model according to the specification parameters of the centrifuge tube, calibrate the model parameters through historical data, fit the vibration transfer function of the bracket through no-load operation data, and determine the relationship between liquid sloshing and liquid level through load experiments; among them, the centrifugal force model is constructed based on the centrifugal force formula, and the heat conduction model is established according to the specific heat capacity and heat dissipation coefficient of the tube body material; S4. Multi-source data fusion and feature extraction; align the preprocessed sensor data with the output of the centrifugal force model constructed in S3 in space and time, use the extended Kalman filter to fuse multi-source information, and estimate the real-time state vector of the centrifuge tube, where the real-time state vector includes rotational speed, temperature, pressure, liquid level, and peak vibration acceleration; Separate the vibration characteristics of the equipment and the centrifuge tube itself using the vibration transfer function of the bracket; Correct the liquid level measurement value according to the relationship between liquid sloshing and liquid level; Calculate the theoretical temperature rise rate in combination with the heat conduction model;

[0021] Extract the following characteristic parameters: Rotation speed change rate: the ratio of the difference in rotation speed between two adjacent sampling times to the sampling time interval; Temperature rise rate: the ratio of the difference in temperature between two adjacent sampling times to the sampling time interval; Pressure fluctuation coefficient: the ratio of the standard deviation of pressure data to the average pressure; Vibration energy proportion: by performing Fourier transform on the acceleration signal, calculate the ratio of the vibration energy within a specific frequency range to the total vibration energy; Liquid level change gradient: the ratio of the difference in liquid level between two adjacent sampling times to the sampling time interval; Centrifugal force deviation rate: the ratio of the difference between the actually calculated centrifugal force and the average centrifugal force during normal operation to the average centrifugal force during normal operation; S5. Abnormal state identification and early warning decision-making; input the characteristic parameters extracted in S4 into the random forest classifier to identify the state categories as normal, slight vibration, overheat, overload, and liquid leakage; the process is as Figure 3 shown; When the actually measured temperature rise rate exceeds 120% of the predicted value of the heat conduction model, it is determined that there is an abnormal heat conduction. If the abnormal heat conduction causes the temperature rise rate to exceed 50% of the critical temperature rise rate, trigger a secondary early warning, send a text message notification to the monitoring terminal and record the abnormal log; if the secondary early warning condition is not met, but it exceeds the normal range, trigger a primary early warning, and display a warning message on the equipment screen; When the proportion of the equipment vibration component in the vibration characteristics exceeds 70%, it is determined that there is an abnormal vibration of the bracket. If the abnormal vibration of the bracket causes the vibration peak acceleration to exceed 80% of the maximum tolerance value, trigger a secondary early warning, send a text message notification to the monitoring terminal and record the abnormal log; if the secondary early warning condition is not met, but it causes the pressure fluctuation coefficient to exceed the normal range by 10%-20%, trigger a primary early warning, and display a warning message on the equipment screen; When the rotation speed change rate exceeds the set threshold 1 and does not exceed threshold 2 (which can be set according to the actual situation, for example, set threshold 1 to 5% / s (the rotation speed changes by more than 5% per second) and threshold 2 to 10% / s), or the pressure fluctuation coefficient exceeds the normal range by 10%-20%, it is determined that there is a slight vibration abnormality, trigger a primary early warning, and display a warning message on the equipment screen; When the temperature rise rate exceeds 50% of the critical temperature rise rate, or the centrifugal force reaches 80%-90% of the maximum tolerable centrifugal force, it is determined as overheating or overloading abnormality, triggering a secondary warning, sending a text message notification to the monitoring terminal and recording the abnormality log; When the liquid level drops suddenly by more than 30%, or the peak vibration acceleration exceeds the maximum tolerable value, it is determined as liquid leakage or severe vibration abnormality, triggering a tertiary warning, automatically triggering the equipment shutdown procedure and cutting off the power supply of the motor; S6. Closed-loop feedback control and parameter adaptation; if adjustable abnormality (other abnormality situations except the tertiary warning) is detected, the edge computing gateway sends a control instruction to the centrifugal equipment to adjust the motor speed through the PID controller; meanwhile, the parameters of the centrifugal force model and heat conduction model in S3 are dynamically updated according to the real-time monitoring data. When the liquid level changes by more than 5%, the total mass of the centrifuge tube is recalculated, including the mass of the sample liquid and the tube body mass; When the abnormal vibration of the bracket lasts for more than 5 minutes, recalibrate the vibration transfer function of the bracket; When the heat conduction abnormality occurs continuously three times, update the heat dissipation coefficient parameter in the heat conduction model.

[0022] In some embodiments, as Figure 2 shown, during the operation of the ultracentrifugation equipment, to achieve comprehensive monitoring of the centrifuge tube state, multiple types of sensors need to be deployed at key positions. First, inside the cavity of the ultracentrifugation equipment, this area is the core space where the centrifuge tube operates and can directly reflect the overall operating environment of the equipment. Deploying sensors here can monitor environment-related parameters. Sensors are deployed on the centrifuge tube bracket because the state of the bracket directly affects the stability of the centrifuge tube. And sensors are deployed on the outer wall of each centrifuge tube to obtain the operating parameters of the centrifuge tube itself. The specifically deployed sensors include a three-axis acceleration sensor for monitoring the vibration of the centrifuge tube during operation; a temperature sensor for real-time monitoring of the temperature change of the centrifuge tube; a pressure sensor for measuring the pressure inside the centrifuge tube; and a liquid level sensor for detecting the liquid level in the centrifuge tube. After deployment, all sensors are connected to the edge computing gateway through wired (such as Ethernet) or wireless (such as ZigBee) means. After connection, hardware address registration is performed to assign a unique identification address to each sensor for subsequent data transmission and management; meanwhile, the communication protocol initialization is completed to determine parameters such as the data transmission format and frequency between the sensor and the gateway to ensure accurate and stable data transmission.

[0023] In some embodiments, when installing the triaxial acceleration sensor, it is precisely pasted at the position of the central axis of the side wall of the centrifuge tube, which can most accurately reflect the vibration state of the centrifuge tube; the temperature sensor is embedded at the bottom of the centrifuge tube to ensure that the temperature change inside the centrifuge tube can be sensed in a timely and accurate manner; the pressure sensor is integrated on the inner side of the tube cap to directly measure the internal pressure of the centrifuge tube; the liquid level sensor is fixed at the 1 / 2 height of the tube body by using an infrared pair-tube structure for real-time monitoring of the liquid level. During the hardware address registration process, the edge computing gateway sends a registration instruction to each sensor. After receiving the instruction, the sensor feeds back its own hardware address information to the gateway, and the gateway records and manages the address. When initializing the communication protocol, the gateway negotiates with the sensor to determine the sampling frequency, for example, set to 100 Hz, that is, 100 data are collected per second; it is determined that the data transmission format is JSON format, which includes fields such as sensor ID, timestamp, physical quantity type, and value, ensuring the standardization and readability of the data.

[0024] In some embodiments, the edge computing gateway synchronously collects the data of all sensors at a set frequency, such as 100 Hz. Since the raw data collected by the sensors may contain noise, interference, or outliers, preprocessing is required. For the raw acceleration signal, which may contain high-frequency noise that affects the judgment of the true vibration state of the centrifuge tube, a Butterworth low-pass filter is used to remove the high-frequency noise, making the acceleration signal smoother and accurately reflecting the vibration of the centrifuge tube. The raw temperature signal may have random interference, and it is processed by moving average filtering to eliminate these random fluctuations and obtain more stable temperature data. For the raw pressure and liquid level signals, outliers outside the normal range may occur, and these outliers are removed by threshold verification to ensure the validity of the pressure and liquid level data. After the above processing, the processed data is organized into a multi-sensor data matrix , providing a basis for subsequent data processing and analysis.

[0025] In some embodiments, when performing Butterworth low-pass filtering, the raw acceleration signal is first substituted into the digital filter formula obtained by converting the analog filter through bilinear transformation: ; where is the transfer function of the digital filter, which describes the processing ability of the digital filter for signals of different frequencies. Here, z is the variable in the discrete signal Z-transform, used to represent the frequency domain characteristics of the discrete time series, and the coefficients = 0.043, = 0.086, = 0.043, = -1.226, = 0.498 (cutoff frequency 50 Hz, second order). Calculate the acceleration data at each sampling point according to this formula to obtain the filtered acceleration signal , effectively removing high-frequency noise. : It is the coefficient of the term in the numerator of the filter transfer function, which has a direct impact on the gain of the filter at zero frequency (DC component). In a low-pass filter, it determines the retention degree of the DC component (low-frequency stable part) in the signal. For example, when is larger, it means that the filter has a stronger amplification effect on the DC component; conversely, when is smaller, the DC component will be relatively weakened. and : These two coefficients are the coefficients of the and terms in the numerator. They work together with the denominator coefficients to determine the shape of the phase-frequency characteristic and amplitude-frequency characteristic curves of the filter. By adjusting the values of and , the attenuation speed and phase delay of the filter for signals of different frequencies can be changed. In the design of Butterworth low-pass filters, they work together with the denominator coefficients to make the filter have a flat amplitude-frequency response in the passband and achieve rapid attenuation in the stopband. Taking a second-order Butterworth low-pass filter as an example, appropriate values of and can ensure that near the cutoff frequency, the transition of the signal from the passband to the stopband is relatively smooth. and : They are the coefficients of the and terms in the denominator, which play a key role in determining the filter characteristics. and cooperate with the numerator coefficients and directly affect the stability and frequency response of the filter. For Butterworth low-pass filters, the values of the denominator coefficients determine the order and cutoff frequency of the filter. In a second-order Butterworth low-pass filter, specific values of and (such as = -1.226, = 0.498, corresponding to a cut-off frequency of 50 Hz) can ensure the maximum flatness of the filter within the passband and start to attenuate high-frequency signals at the cut-off frequency at a predetermined rate, while ensuring the stable operation of the filter at all frequencies and avoiding unstable phenomena such as oscillations. The values of these coefficients are determined according to the design requirements of the filter (such as cut-off frequency, filter order, passband ripple, stopband attenuation, etc.) through mathematical calculations (such as bilinear transformation, impulse response invariance method, etc.). Taking the design of a second-order Butterworth low-pass filter by bilinear transformation as an example, first calculate the parameters of the analog filter according to the required cut-off frequency, and then convert the transfer function of the analog filter into the transfer function of the digital filter through bilinear transformation, so as to obtain the coefficient values in the above digital filter transfer function.

[0026] In some embodiments, for the temperature signal, the temperature sequence is read sequentially in time order. For every M1 = 5 consecutive data points, the formula is used to calculate the moving average value to obtain the smoothed temperature sequence to eliminate random interference.

[0027] In some embodiments, when processing pressure and liquid level signals, the pressure and liquid level data collected in real time are compared with the preset range . If the pressure exceeds the range, that is, it does not meet , then the adjacent valid values are taken for linear interpolation, and the reasonable pressure value is calculated according to the formula . Similarly, the liquid level data is processed. Finally, the processed acceleration, temperature, pressure, and liquid level data are aligned in time order to form a multi-sensor data matrix , where M is the number of samples and n = 4 represents the number of sensor types.

[0028] In some embodiments, in order to accurately analyze the physical state of the centrifuge tube during operation, it is necessary to establish a centrifugal force-stress model and a heat conduction model according to the material density , volume , maximum tolerable rotational speed and other specification parameters of the centrifuge tube. These models can theoretically describe the physical changes of the centrifuge tube under different working conditions. In order to make the model more in line with the actual situation, through no-load running experiments, the vibration data of the equipment base and the centrifuge tube are collected, and the vibration transfer function of the bracket is fitted by the least squares method to find the relationship between the vibration of the equipment base and the centrifuge tube. Then, through load experiments, the liquid level , Monitor the liquid sloshing condition and determine the relationship between liquid sloshing and liquid level. Finally, use the data collected during the actual operation of the equipment to calibrate the parameters of the established model so that the model can more accurately predict the actual operating state of the centrifuge tube.

[0029] In some embodiments, when establishing the centrifugal force model, according to the formula , where is the centrifugal force, is the angular velocity, is the radius of rotation. The calculation of the total mass m (of the centrifuge tube and the internal sample) involves the liquid mass and the tube body mass. The liquid mass is determined by the liquid density and the liquid volume . The liquid volume , and the tube body mass is . So .

[0030] In some embodiments, the heat conduction model is constructed based on the formula , where is the rate of change of temperature T with time t, that is, the temperature rise rate of the centrifuge tube, is the frictional heat generation power, is the specific heat capacity, is the heat dissipation coefficient, is the current temperature, is the ambient temperature.

[0031] In some embodiments, when conducting the no-load running experiment, start the ultra-high speed centrifuge equipment and make it run in the no-load state. Use the sensor to collect the vibration signal of the equipment base and the vibration signal of the centrifuge tube , and substitute the Laplace-transformed and into the transfer function . Use the least squares method to calculate the parameters , , , values. For example, obtain = 0.5, = 0.2, = 0.8, = 0.1 to determine the vibration transfer function of the bracket. In the load experiment, inject different volumes of liquid into the centrifuge tube to change the liquid level , run the equipment and monitor the vibration condition of the centrifuge tube. Establish the relationship between the sloshing acceleration and the liquid level through data analysis, such as . During the actual operation, the real-time collected data such as temperature and vibration are substituted into the established model, the initial model parameter values are set, and the heat dissipation coefficient is continuously adjusted through the gradient descent method and the frictional heat generation power to minimize the error between the temperature predicted by the model and the actually measured temperature, and complete the calibration of the model parameters.

[0032] The goal of the least squares method is to minimize the sum of the squares of the errors between the observed data and the model prediction values . Let the error function be , where n is the number of data samples, is the predicted value calculated through the transfer function and . To solve for the minimum value of , partial derivatives are taken with respect to and set to 0: ; The system of equations is solved simultaneously to obtain the value of , and the vibration transfer function of the bracket is determined.

[0033] In some embodiments, after obtaining the preprocessed multi-sensor data matrix , due to the high data dimension and information redundancy, non-negative matrix factorization (NMF) is performed to extract more valuable features. Through the NMF algorithm, the high-dimensional data is mapped to a low-dimensional space to obtain the basis matrix and the coefficient matrix (i.e., the feature matrix) , thereby extracting the implicit feature patterns of the data. Then, the feature matrix obtained by NMF (including elements such as the rotational speed change rate, temperature rise rate, pressure fluctuation coefficient, vibration energy ratio, liquid level change gradient, centrifugal force deviation rate, etc.) and the physical parameters of the centrifuge tube, such as the rotational speed and the liquid level are used as inputs, and the radial basis function (RBF) surface fitting technology is adopted to establish a non-linear mapping model between the sensor data. This model can predict physical quantities such as the pressure and vibration of the centrifuge tube based on the input features and physical parameters, realizing the deep fusion and analysis of multi-source data.

[0034] In some embodiments, when performing non-negative matrix factorization, the multi-sensor data matrix is used as the input, and the preset feature dimension k1 = 3 (this value is determined through cross-validation to ensure maximum retention of data features while reducing the dimension). First, the basis matrix and the coefficient matrix , then calculate according to the iterative update formula. The iterative update formula is: ; are the sample index, feature index, and sensor dimension index respectively. In each iteration, update the and element values according to the formula, and continuously adjust the matrix until the reconstruction error converges (usually set the number of iterations to 500 times), and finally obtain the dimensionality-reduced feature matrix . When constructing the RBF surface model, , rotational speed , and liquid level are composed into the input vector , and the actually measured pressure is used as the output target . Select sample points according to the RBF model formula: , where is the weight of the th basis function, which determines the contribution degree of each basis function to the final prediction result; the number of basis functions = 2g, g is the number of samples, the center of the basis function , is the kernel width parameter, and is the linear term coefficient. Use the least squares method to solve the model parameters , that is, adjust the parameters to minimize the loss function , and complete the training of the RBF model. In the model application stage, the , , and collected in real time are composed into the input vector , input into the trained RBF model, and the predicted pressure is calculated.

[0035] In some embodiments, in order to ensure the accuracy and reliability of the constructed RBF surface model, it needs to be verified. Input the actually measured data into the trained RBF surface model, and calculate the residual between the predicted value and the true value of the model. By statistically calculating the mean and standard deviation of the residuals, evaluate the fitting accuracy of the model. If the residuals exceed a certain threshold, that is, (According to the three - standard - deviation criterion, which is commonly used in statistics to judge the abnormality of data), it is determined as an abnormal data point. For abnormal data points, the system will trigger a data re - sampling mechanism to re - sample the data at that moment; if multiple abnormal data points appear continuously, the model will be triggered to be retrained to ensure that the model can accurately reflect the actual operating state of the centrifuge tube.)

[0036] In some embodiments, when calculating the residuals, for each sampling point, the actually measured pressure value and the predicted pressure value of the RBF model are substituted into the formula to obtain the residual value . As the data is continuously collected, the mean value and standard deviation of the residuals are calculated using a sliding window method (the window size can be set according to the actual situation, such as including 10 sample points). The calculation formulas are: ; After each calculation of the residual statistics, each residual value is compared with the threshold . If , then this data point is determined as an abnormal data point, and the edge computing gateway sends an instruction to the sensor to re - collect the relevant data at that moment. If abnormal data points are detected in multiple consecutive sliding windows (such as 3 windows), it is considered that the model may have a deviation, and the re - training process of the RBF model is triggered. The model is retrained using the newly collected data to adjust the model parameters to improve the accuracy of the model.)

[0037] In some embodiments, after extracting the NMF feature matrix and calculating the surface fitting residual , the mean value of the residuals , the standard deviation of the residuals and the maximum residual value , these data are fused to form a new feature vector . This feature vector is input into an improved random forest classifier. Utilizing the powerful classification ability of the random forest classifier, the state of the centrifuge tube is identified. The output state categories include normal, slight vibration, overheating, overload, liquid leakage, and model mismatch. During the identification process, according to the residual standard deviation threshold (such as , where is the residual standard deviation of the training set) to determine whether to trigger the model adaptive update, so as to adjust the model in a timely manner to ensure the accuracy and reliability of the monitoring system.)

[0038] In some embodiments, first, the NMF feature matrix , the mean value of the residuals Residual standard deviation and the maximum residual value are combined in sequence to form a new feature vector . When training the random forest classifier, a large amount of historical feature vector data and the corresponding accurate status labels (various statuses such as normal, slight vibration, etc.) are used as training samples. Set the parameters of the random forest, and construct a classification model containing 100 decision trees. When training each decision tree, 70% of the data and 50% of the features are randomly sampled from the training samples for training, in this way to increase the diversity and generalization ability of the model. After training is completed, the feature vector generated in real time is input into the trained random forest classifier. Each decision tree in the classifier classifies the input feature vector to obtain a classification result, and finally the final classification result is determined through a voting mechanism, that is: ; where is the th decision tree, is the indicator function, represents the th classification category in the random forest classifier. In the application scenario of intelligent monitoring of ultracentrifuge tubes, these classification categories correspond to different operating states that the centrifuge tube may present, specifically: : represents the "normal" state, that is, the operating parameters of the centrifuge tube are within the normal range and no abnormal situation occurs; : represents the "slight vibration" abnormal state, which means that the vibration amplitude of the centrifuge tube exceeds the normal range but has not reached a serious level; : represents the "overheat" abnormal state, that is, the temperature of the centrifuge tube is too high, which may affect the normal operation of the equipment and the quality of the sample; : represents the "overload" abnormal state, indicating that the load borne by the centrifuge tube exceeds its designed bearing range; : represents the "liquid leakage" abnormal state, indicating that the liquid in the centrifuge tube leaks; : represents the "model mismatch" abnormal state, that is, the residual of the surface fitting model exceeds the expected range and the model cannot accurately reflect the actual situation.

[0039] When it is 1, otherwise it is 0. After obtaining the classification result, compare the current residual standard deviation with the residual standard deviation of the training set If ​, it is determined as the "model mismatch" state, triggering the subsequent model adaptive update process; if it is determined as other abnormal states (such as slight vibration, overheating, etc.), corresponding-level warnings are triggered according to the preset warning rules. For example, when slight vibration is recognized, it may be only prompted by the local indicator light of the device flashing; for relatively serious abnormalities such as overheating or overload, not only will there be an audible and visual alarm locally on the device, but detailed abnormal information will also be sent to the remote monitoring terminal, including the type of abnormality, occurrence time, relevant characteristic parameters, etc., so that the operator can take measures in time. At the same time, the system will record the results and relevant data of each abnormal recognition for subsequent data analysis and model optimization.

[0040] In some embodiments, after the random forest classifier outputs the abnormal category, the system performs a warning trigger operation according to the preset abnormal-warning mapping table. This mapping table clearly stipulates the warning levels and processing methods corresponding to different abnormal categories. For example: Abnormal Category Warning Level Handling Method Slight Vibration Level 1 Warning <![CDATA[The local yellow indicator light of the device flashes at a frequency of 1 time per second, and the prompt message "Slight vibration anomaly" is displayed on the device operation interface for 10 minutes; the abnormal information is recorded in the local log file, including the timestamp, X new feature vector data]]> Overheat Level 2 Warning The device emits a continuous beeping sound locally, and the red indicator light is always on; send a text message and a push notification containing the content "Overheat anomaly, current temperature: [measured temperature value], it is recommended to immediately check the cooling system" to the remote monitoring terminal; generate an abnormal work order on the device management platform and assign it to the maintenance personnel Overload Level 2 Warning Through the heat treatment method, at the same time, display the detailed overload data (current centrifugal force, rotational speed, etc.) on the device operation interface Liquid Leakage Level 3 Warning The device emits a harsh alarm sound locally, and the red indicator light flashes quickly (5 times per second); automatically cut off the device power supply and stop the centrifugal operation; send an emergency notice to all relevant personnel (operators, managers, maintenance personnel), containing the information "Liquid leakage, the device has stopped, please go to deal with it immediately"; prominently display the leakage device location and relevant data on the monitoring large screen

[0041] When recording abnormal data, the system stores information such as the current time, abnormal category, feature vector , the predicted value and measured value of the RBF model, etc. in a specific format (such as JSON format) into the database, which is convenient for subsequent analysis of the abnormal occurrence pattern through data mining technology, providing a basis for further optimizing the random forest model and adjusting the warning strategy.

[0042] In some embodiments, when S5 detects an adjustable abnormality (such as slight vibration, overheating, etc., which are non-serious emergencies except for the third-level warning), the edge computing gateway adjusts the motor speed through a PID controller according to the preset control strategy, attempting to restore the normal operating state of the centrifuge tube. If it is determined as the "model mismatch" state, or it is detected that key parameters such as the liquid level h and sample density change, the system will use the RBF surface model to reverse-infer the unknown parameters of the centrifuge tube, such as the tube body mass , heat dissipation coefficient , etc. The objective function is continuously optimized through the gradient descent method to minimize the error between the model predicted value and the measured value, realizing the online calibration of the physical model parameters, and then updating the entire monitoring model to adapt to the change of the centrifuge tube operating state, forming a complete closed-loop feedback control system.

[0043] In some embodiments, the PID control process: if an adjustable abnormality such as slight vibration or overheating is detected, the system obtains the current rotational speed and the preset target rotational speed , and calculates the rotational speed error . Substitute the error into the PID control formula: ; Among them, the proportionality coefficient , the integral coefficient , and the differential coefficient are pre-tuned by the critical proportional band method, and the initial values are set to = 0.5, = 0.1, = 0.2. After calculating the control quantity , the edge computing gateway sends the control instruction to the motor driver to adjust the motor speed. For example, when > 0, the speed is increased, < 0, the speed is decreased, so that the operating state of the centrifuge tube gradually tends to be stable. During the adjustment process, the speed and other relevant parameters are continuously monitored, and the PID parameters are dynamically adjusted according to the actual situation to ensure the control effect.

[0044] In some embodiments, the parameter inversion and model update process: when "model mismatch" ( ) is triggered or the liquid level h changes by more than 5%, the sample density changes, etc., the parameter inversion process is started. Taking the heat dissipation coefficient k as an example, the objective function is set as: ; Among them is the measured temperature, is the predicted temperature of the RBF surface model including the heat dissipation coefficient . Initialize the heat dissipation coefficient = 0.2, and the learning rate is 0.01. Calculate the gradient of the objective function with respect to through automatic differentiation technology. Perform iterative calculations according to the gradient update formula . Each time an iteration is performed, calculate the current objective function value . When (the preset convergence threshold, such as ), it is considered to have reached convergence, and the updated heat dissipation coefficient is obtained. Substitute the updated into the heat conduction model in S3, and re-train the RBF surface model to optimize the model with the latest parameters and data to ensure that the model can accurately reflect the current operating state of the centrifuge tube and complete the entire closed-loop feedback control process.

[0045] The above-described embodiments are used to illustrate the present invention, not to limit the present invention. Therefore, changes in the example values or replacement of equivalent elements still belong to the scope of the present invention.

[0046] From the above detailed description, those of ordinary skill in the art can clearly understand that the present invention can indeed achieve the foregoing objectives, and it actually complies with the provisions of the Patent Law.

[0047] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention. The above description is only the preferred embodiments of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0048] It should be noted that the above description of the process is only for illustration and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0049] The basic concept has been described above. Obviously, for those of ordinary skill in the art after reading this application, the above invention disclosure is only an example and does not constitute a limitation to this application. Although not explicitly stated here, those of ordinary skill in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are proposed in this application, so such modifications, improvements, and corrections still belong to the spirit and scope of the exemplary embodiments of this application.

[0050] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment", "an embodiment", and / or "some embodiments" mean a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "an embodiment" or "one embodiment" or "an alternative embodiment" mentioned twice or more at different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application can be appropriately combined.

[0051] In addition, those of ordinary skill in the art can understand that various aspects of this application can be described and illustrated by several patentable types or situations, including any new and useful process, machine, product, or combination of substances, or any new and useful improvement thereof. Therefore, various aspects of this application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software can all be referred to as "units", "modules", or "systems". In addition, various aspects of this application can take the form of a computer program product embodied in one or more computer-readable media, in which computer-readable program code is included.

[0052] The computer program code required for the operations of various parts of this application can be written in any one or more programming languages, including object-oriented programming languages such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages such as C programming language, VisualBasic, Fortran2103, Perl, COBOL2102, PHP, ABAP, dynamic programming languages such as Python, Ruby and Groovy, or other programming languages, etc. This program code can run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (for example, through the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0053] In addition, unless clearly stated in the claims, the order of the processing elements and sequences, the use of numerical letters, or the use of other names in this application are not used to limit the order of the processes and methods of this application. Although some currently considered useful embodiments of the invention are discussed through various examples in the above disclosure, it should be understood that such details only serve the purpose of illustration, and the appended claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that conform to the essence and scope of the embodiments of this application. For example, although the implementation of the above various components can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installed on an existing server or mobile device.

[0054] Similarly, it should be noted that, in order to simplify the description of this application disclosure and thus help the understanding of one or more embodiments of the invention, in the previous description of the embodiments of this application, sometimes multiple features are merged into one embodiment, drawing or description thereof. However, this method of this application should not be construed as reflecting the intention that the claimed subject matter requires more features than those clearly recited in each claim. On the contrary, the subject matter of the invention should have fewer features than the above single embodiment.

Claims

1. An intelligent monitoring method for ultracentrifugation tubes based on a sensor network, characterized in that, Including: S1. Deploy multiple types of sensors in the centrifugal equipment to form a three-dimensional distributed monitoring network and connect to the edge computing gateway; S2. The edge computing gateway synchronously collects all sensor data and preprocesses it; S3. Establish a centrifugal force-stress model and a heat conduction model according to the centrifugal tube specification parameters, calibrate the model parameters through historical data, fit the vibration transfer function of the bracket through no-load operation data, and determine the relationship between liquid sloshing and liquid level through load experiments; S4. Align the preprocessed sensor data with the output of the centrifugal force-stress model in space and time, use the extended Kalman filter to fuse multi-source information, and estimate the real-time state vector of the centrifugal tube; Extract characteristic parameters based on the real-time state vector of the centrifugal tube; S5. Input the characteristic parameters into the random forest classifier to identify the state category; If it is determined to be abnormal, trigger a first-level warning, a second-level warning or a third-level warning according to the preset rules; S6. If a first-level warning or a second-level warning is detected, the edge computing gateway sends a control command to the centrifugal equipment and adjusts the motor speed through a PID controller; At the same time, dynamically update the parameters of the vibration transfer function of the bracket, the centrifugal force-stress model or the heat conduction model according to the real-time monitoring data.

2. The intelligent monitoring method for ultracentrifuge tubes based on a sensor network according to claim 1, characterized in that, Deploying multiple types of sensors includes: pasting a three-axis acceleration sensor on the central axis of the side wall of the centrifugal tube, embedding a temperature sensor at the bottom of the centrifugal tube, integrating a pressure sensor on the inner side of the tube cap, and fixing the liquid level sensor at the 1 / 2 height of the tube body by using an infrared pair-tube structure.

3. The intelligent monitoring method for ultracentrifuge tubes based on a sensor network according to claim 2, wherein The preprocessing in S2 includes: using a low-pass filter to remove high-frequency noise from the original acceleration signal, eliminating random interference from the original temperature signal through moving average filtering, and removing outliers from the original pressure and liquid level signals through threshold verification.

4. The intelligent monitoring method for ultracentrifuge tubes based on a sensor network according to claim 1, characterized in that The centrifugal force-stress model is constructed based on the centrifugal force formula, and the heat conduction model is established according to the specific heat capacity and heat dissipation coefficient of the tube body material.

5. The intelligent monitoring method for ultracentrifuge tubes based on a sensor network according to claim 1, wherein The state transition function of the extended Kalman filter is determined according to the change law of the centrifugal tube speed, and the process noise covariance matrix is obtained through sensor error statistics.

6. The intelligent monitoring method for ultracentrifuge tubes based on a sensor network according to claim 1, wherein, The training samples of the random forest classifier include normal condition data and historical data of various abnormal conditions.

7. The intelligent monitoring method for an ultracentrifuge tube based on a sensor network according to claim 1, characterized in that, The first-level warning displays a warning message on the device screen, the second-level warning sends a text message notification to the monitoring terminal and records the abnormal log, and the third-level warning automatically triggers the device shutdown procedure and cuts off the motor power supply.

8. The intelligent monitoring method for an ultracentrifuge tube based on a sensor network according to claim 1, characterized in that, The proportional, integral, and differential coefficients of the PID controller are tuned by the critical ratio method.

9. The intelligent monitoring method for an ultracentrifuge tube based on a sensor network according to claim 1, characterized in that, In S6, when the liquid level changes by more than 5%, recalculate the total mass of the centrifugal tube, including the mass of the sample liquid and the mass of the tube body.

10. The intelligent monitoring method for an ultracentrifuge tube based on a sensor network according to claim 1, characterized in that, In S6, by comparing the residuals between the actual monitoring data and the model prediction data, when the residuals exceed a certain threshold, recalibrate the model parameters.

Citation Information

Patent Citations

  • Intelligent centrifugal tube capable of detecting centrifugal working condition in real time and use method

    CN116273511A

  • Running state parameter acquisition and fault diagnosis method for three-phase decanter centrifuge

    CN118503882A

  • Method, system and equipment for optimizing comprehensive performance of horizontal screw centrifuge and readable storage medium

    CN118520704A

  • Medical centrifuge rotating speed calibration method and system based on base force sensor

    CN119001150A

  • Industrial-type variable-speed centrifuge

    US3738490A

Cited By

  • Centrifugal machine control method and system for soybean protein production

    CN120550951A

  • Brushing test robot system

    CN120588246A