Pipettor management system and method

Abstract

The invention belongs to the technical field of pipettor management, and discloses a pipettor management system and method.The method comprises the steps that pipettor performance monitoring data, piston pressure signal data, basic data and laboratory environment data are collected; performing feature extraction on the pipettor piston pressure signal data to obtain extracted feature data; preprocessing the pipettor performance monitoring data and the extracted feature data to obtain a performance data set; based on the performance data set, using a PSO-SVM model to judge whether the pipettor has a fault or not; when the fault does not occur, taking the basic data of the pipettor as input, and using an XGBoost model to predict the pipetting precision data; a real-time correction precision tolerance threshold value is set, when the pipetting precision data is larger than or equal to the real-time correction precision tolerance threshold value, early warning is carried out, and a calibration program is triggered for calibration; the operation reliability and precision of the pipettor are improved, errors caused by faults and inaccurate precision are reduced, and the accuracy of experimental results is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of pipette management, and in particular to a pipette management system and method. Background Art

[0002] The existing pipette management system has exposed many problems, which affect the intelligence, accuracy and safety of the system. First, the wear of the pipette tip cannot be monitored in real time, resulting in the inability to detect wear and aging of the equipment in time during use, thus affecting the accuracy of the experimental results. The existing system lacks comprehensive data collection and analysis of pipette performance. There is no effective real-time data to determine whether the pipette has potential faults. Many faults can only be discovered when abnormalities occur during use, which increases the risk in the experimental process. In addition, the system lacks real-time detection and judgment of pipetting accuracy, and is unable to determine whether the pipetting accuracy meets the requirements, and there is no corresponding early warning mechanism. Once the pipetting accuracy deviates from the standard, the system cannot issue a warning, nor does it trigger a calibration procedure to correct the accuracy deviation. This increases the uncertainty of the experimental results and affects the reliability of the experiment and the repeatability of the results.

[0003] In view of this, the present invention proposes a pipette management system and method to solve the above problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution, a pipette management system, comprising: a data acquisition module: collecting pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Feature extraction module: extracting features from the collected pipette piston pressure signal data to obtain extracted feature data; Data preprocessing: preprocessing the pipette performance monitoring data and extracted feature data to obtain a performance data set; Fault judgment module: Based on the performance data set, the PSO-SVM model is used to determine whether the pipette has a fault; Pipette accuracy detection module: When no failure occurs, the basic data of the pipette is used as input, and the XGBoost model is used to predict the pipette accuracy data; Early warning module: Set the real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued and the calibration program is triggered for calibration.

[0005] Further, the pipette performance monitoring data includes pipette tip wear data, pipette battery voltage and current data; The basic data of the pipette includes the pipetting volume, pipetting speed, liquid density, and liquid viscosity data; Laboratory environmental data includes laboratory temperature, humidity and air pressure data.

[0006] Furthermore, the method for acquiring the wear data of the pipette head includes: Use laser to scan the new pipette head to obtain 3D data. Use the 3D data of the new pipette head as standard data. Then use laser to scan the current pipette head and obtain the current 3D data. Use the height difference between the current 3D data and the M_L points in the standard 3D data to quantify the wear data: ,in, represents the three-dimensional data of the standard data point q, Represents the three-dimensional data of the current data point q, represents the total number of three-dimensional data points, q represents the index of the three-dimensional data point, Indicates the wear data of the pipette tip.

[0007] Furthermore, the specific method of extracting feature data includes: The pressure signal data when the pipette piston moves is obtained, and the instantaneous skewness characteristic data, instantaneous kurtosis characteristic data and instantaneous amplitude characteristic data of the pressure signal are calculated. The formulas for the instantaneous skewness characteristic data and the instantaneous kurtosis characteristic data are: ; ,in, and Represents the instantaneous skewness characteristic data and instantaneous kurtosis characteristic data at time t, Indicates the pressure signal Pressure value, Indicates the total number of pressure values ​​in the pressure signal. Indicates the average value of the pressure signal; The formula for instantaneous amplitude characteristic data is: , represents the instantaneous amplitude characteristic data at time t, represents the original pressure value at time t, Represents the Hibert transform pressure value at time t.

[0008] Furthermore, the specific method of obtaining the performance data set includes: Based on the pipette performance monitoring data and extracted feature data, the interpolation method is used to process missing values, the Z_Score method is used to process outliers, and normalization is performed to obtain a normalized data set. The dynamic weighted interpolation method is used to unify the timestamps to obtain the basic data set.

[0009] Furthermore, the specific method of using the dynamic weighted interpolation method to unify the timestamp includes: Unify the formats of the collected pipette performance monitoring data and the timestamps corresponding to the extracted feature data; Set target timestamp , by the timestamp before the target timestamp and the next timestamp , to set the basic weights, , Indicates timestamp The basic weight of Indicates timestamp The basic weight of Calculate the previous timestamp and the next timestamp The rate of change of the upper normalized data value: , adjust the basic weight according to the change rate of the previous and next timestamps, and calculate the dynamic weight of the previous and next timestamps: in, Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , represents the adjustment coefficient, Indicates the timestamp The rate of change on Indicates the timestamp The rate of change of The normalized data value is the specific value of a data in the normalized data set; According to the calculated dynamic weight, the target timestamp is calculated using weighted interpolation Normalized data values ​​on : .

[0010] Furthermore, the specific method of using the PSO-SVM model to determine whether the pipette fails includes: The input sample set includes L_o groups of samples, each of which includes a basic data set and corresponding fault data; Initialize the particle swarm parameters, including the number of particles, maximum number of iterations, speed and position; each particle position corresponds to a hyperplane parameter weight vector and the bias term ,in represents the current position of the i-th particle, represents the weight of the pipette head wear data, represents the weight of the pipette battery voltage data, represents the weight of the pipette battery current data, represents the instantaneous skewness characteristic data, Represents the instantaneous kurtosis characteristic data and Represents instantaneous amplitude characteristic data; Fitness value calculation: Calculate the fitness value for each particle. The fitness function is based on the objective function of SVM, which is a weighted combination of maximizing the classification interval and minimizing the penalty for misclassification, that is: ,in, represents the number of samples, represents the sample index, represents the term that maximizes the classification interval, represents the slack variable of the jth sample, indicating the degree of misclassification allowed, represents the penalty factor; Update the local optimal position and global optimal position of the particle. In each iteration, if the fitness value of the current particle is better than the historical optimal fitness value of the particle, the individual optimal position is updated; for the entire particle group, when the fitness value of a particle is better than the current global optimal fitness value, the global optimal position is updated; Particle swarm update rule: Each particle will update its position according to the current speed, local optimal position, that is, the best solution in the history of the particle, and global optimal position, that is, the best solution among all particles. The speed update formula of each particle is: ,in, and Represent the weight factors of the speed of the inertia part, the individual learning part and the global learning part, and The random numbers representing the individual learning part and the global learning part are used to introduce randomness, and the range is , represents the optimal position of the ith particle, represents the global optimal position of the particle swarm, represents the current velocity of the ith particle, Represents the current position of the i-th particle, that is, the current hyperplane parameter weight vector and the bias term ; The position update formula for each particle is: ; When the number of iterations reaches the set maximum value, the optimization is stopped and the global optimal position, that is, the optimal hyperplane parameters, is output. and , build a decision function , represents the transpose of the optimal weight vector, Represents a new input sample, when When , it indicates that the pipette is not faulty; when , it indicates that the pipette is malfunctioning.

[0011] Furthermore, when no failure occurs, the basic data of the pipette is used as input, and the specific method of using the XGBoost model to predict the pipetting accuracy data includes: The input accuracy sample set includes K_F group samples, each group of samples includes a basic data set and corresponding pipetting accuracy data; the data in the basic data set is the characteristic variable, and the pipetting accuracy data is the target variable; Initialize hyperparameters: Use grid search method to initialize the hyperparameters of XGBoost model; Define the final objective function. On the basis of using the mean square error as the objective function, introduce a regularization term to determine the final objective function. Initialize the prediction value, take B_P precision data to calculate the mean, and define it as the initial prediction value: ,in, represents the initial predicted value of pipetting accuracy, Indicates Pipetting accuracy data values, Indicates the index of pipetting precision data, and B_P indicates the total number of precision data; Calculate the residual. In each iteration, calculate the residual between the current predicted value and the true value: ,in, Indicates The residual of the wheel, Indicates Predicted value of pipetting accuracy of the wheel; In each round, a decision tree is constructed to fit the residual, a feature variable is selected, the gain of each data value in the feature variable is calculated, and the data value with the largest gain is selected as the current optimal partition point, which is divided into two sub-nodes, the left sub-tree and the right sub-tree. Each sub-tree is then partitioned using the gain, and the partitioning stops when the maximum depth of the set tree is reached. In each round of iteration, the prediction value of the model is updated through the newly generated tree until the set number of iterations is reached and the iteration is stopped.

[0012] Furthermore, the specific method of setting the real-time correction accuracy tolerance threshold includes: Use historical pipetting accuracy to set the initial accuracy tolerance threshold: ,in, Represents the average value of historical pipetting accuracy, represents the standard deviation of historical pipetting accuracy, represents the adjustment factor; Get real-time comprehensive environmental data: ,in, and Respectively represent the real-time laboratory temperature, humidity and air pressure data, and Respectively represent the weights of the temperature, humidity and air pressure data of the laboratory, Represents real-time comprehensive environmental data; Use real-time comprehensive environmental data to correct the initial accuracy tolerance threshold and obtain the real-time corrected accuracy tolerance threshold: ; When the predicted pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, the calibration procedure is triggered, an early warning is given, and the operation is repeated; When the predicted pipetting accuracy data is less than the real-time corrected accuracy tolerance threshold, the operation continues.

[0013] A pipette management method, which is implemented based on the pipette management system, comprises: Step S1, collecting pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Step S2, performing feature extraction on the collected pipette piston pressure signal data to obtain extracted feature data; Step S3, preprocessing the pipette performance monitoring data and the extracted feature data to obtain a performance data set; Step S4: Based on the performance data set, use the PSO-SVM model to determine whether the pipette fails; Step S5: When no failure occurs, the basic data of the pipette is used as input, and the XGBoost model is used to predict the pipetting accuracy data; Step S6, setting a real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than the real-time correction accuracy tolerance threshold, an early warning is issued and a calibration program is triggered to perform calibration.

[0014] The technical effects and advantages of a pipette management system and method of the present invention are as follows: The present invention realizes accurate monitoring and dynamic adjustment of the pipette state through comprehensive data collection and real-time monitoring, combined with feature extraction, fault diagnosis, pipetting accuracy prediction and early warning mechanism. It uses the PSO-SVM model to efficiently judge faults, the XGBoost model to accurately predict pipetting accuracy, and combines real-time environmental data to correct the accuracy tolerance threshold to ensure that the pipette always operates within the optimal accuracy range, providing an efficient and intelligent equipment management and maintenance solution. The main advantages of the present invention lie in its intelligent and automated features. Through real-time data collection and analysis, equipment failures and precision anomalies can be discovered in a timely manner, human intervention can be reduced, and the stability and work efficiency of the equipment can be improved. At the same time, the system can dynamically adjust the precision tolerance according to environmental changes, avoiding precision errors caused by environmental fluctuations and ensuring the high precision of pipetting operations. In addition, the early warning and trigger calibration functions can effectively reduce the downtime and maintenance costs of the equipment, extend the service life of the equipment, and improve the overall work efficiency of the laboratory. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a pipette management system of the present invention; Figure 2 This is a schematic diagram of a pipette management method of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Embodiment 1:

[0017] See also Figure 1 As shown, a pipette management system described in this embodiment includes: Data acquisition module: collects pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Pipette performance monitoring data includes pipette tip wear data, pipette battery voltage and current data; The basic data of the pipette includes the pipetting volume, pipetting speed, liquid density, and liquid viscosity data; Laboratory environmental data include laboratory temperature, humidity and air pressure data; Among them, an embedded microcontroller is used to obtain the battery voltage and current data of the pipette; the built-in flow sensor of the pipette is used to monitor the pipette transfer volume data in real time; and a speed sensor is equipped to monitor the speed of the liquid aspiration and discharge process in real time, and the average speed is used as the transfer speed; a density meter is used to directly measure the liquid density data in real time; a rotational viscometer is used to measure the liquid viscosity data in real time; a temperature sensor, a humidity sensor and an air pressure sensor are installed in the laboratory to measure the laboratory temperature, humidity and air pressure data in real time; a pressure sensor is installed on the piston of the pipette or a part related to the piston, and the sensor can monitor and record the pressure signal changes generated by the piston during the working process in real time; Feature extraction module: extracting features from the collected pipette piston pressure signal data to obtain extracted feature data; Data preprocessing: preprocessing the pipette performance monitoring data and extracted feature data to obtain a performance data set; Fault judgment module: Based on the performance data set, the PSO-SVM model is used to determine whether the pipette has a fault; Pipette accuracy detection module: When no failure occurs, the basic data of the pipette is used as input, and the XGBoost model is used to predict the pipette accuracy data; Early warning module: Set the real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued and the calibration program is triggered for calibration.

[0018] The methods for obtaining the wear data of the pipette tip include: Use laser to scan the new pipette head to obtain 3D data. Use the 3D data of the new pipette head as standard data. Then use laser to scan the current pipette head and obtain the current 3D data. Use the height difference between the current 3D data and the M_L points in the standard 3D data to quantify the wear data: ,in, represents the three-dimensional data of the standard data point q, Represents the three-dimensional data of the current data point q, represents the total number of three-dimensional data points, q represents the index of the three-dimensional data point, Indicates the wear data of the pipette tip; Using laser scanning to measure pipette tip wear data provides a non-contact, accurate and efficient measurement method. The three-dimensional data obtained by laser scanning can accurately record the geometry of the pipette tip, including surface details and tiny wear changes. This method can quantify the degree of wear and avoid the errors or damage that may be caused by contact measurement. In addition, laser scanning technology can efficiently collect large amounts of data, which is suitable for large-scale testing, and can provide high-precision wear analysis in a short time, providing a reliable basis for pipette maintenance and quality control.

[0019] Specific methods for extracting feature data include: The pressure signal data when the pipette piston moves is obtained, and the instantaneous skewness characteristic data, instantaneous kurtosis characteristic data and instantaneous amplitude characteristic data of the pressure signal are calculated. The formulas for the instantaneous skewness characteristic data and the instantaneous kurtosis characteristic data are: ,in, and Represents the instantaneous skewness characteristic data and instantaneous kurtosis characteristic data at time t, Indicates the pressure signal Pressure value, Indicates the total number of pressure values ​​in the pressure signal. Indicates the average value of the pressure signal; The formula for instantaneous amplitude characteristic data is: represents the instantaneous amplitude characteristic data at time t, represents the original pressure value at time t, represents the Hibert transform pressure value at time t; The instantaneous skewness and instantaneous kurtosis reflect the degree of deviation and volatility of the pressure signal, and can capture abnormal fluctuations caused by internal component wear or other mechanical failures. The instantaneous amplitude characteristic data provides the instantaneous amplitude of the pressure signal through Hilbert transform, which can reveal the intensity change of the pressure signal, and further reveal whether the piston movement is smooth, whether there are signs of faults such as excessive vibration or resistance; Hilbert transform is a mathematical operation usually used in signal processing. It can convert a real-valued signal into a complex-valued signal. By performing a 90-degree phase shift on the signal, an imaginary signal with the same frequency but phase shift as the original signal is generated, thus forming a complex signal.

[0020] The specific methods for obtaining the performance data set include: Based on the pipette performance monitoring data and extracted feature data, the interpolation method is used to process missing values, the Z_Score method is used to process outliers, and normalization is performed to obtain a normalized data set. The dynamic weighted interpolation method is used to unify the timestamps to obtain the basic data set.

[0021] The specific methods of using the dynamic weighted interpolation method to unify timestamps include: Unify the formats of the collected pipette performance monitoring data and the timestamps corresponding to the extracted feature data; Set target timestamp , by the timestamp before the target timestamp and the next timestamp , to set the basic weights, , Indicates timestamp The basic weight of Indicates timestamp The basic weight of Calculate the previous timestamp and the next timestamp The rate of change of the upper normalized data value: , adjust the basic weight according to the change rate of the previous and next timestamps, and calculate the dynamic weight of the previous and next timestamps: ,in, Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , represents the adjustment coefficient, Indicates the timestamp The rate of change on Indicates the timestamp The rate of change of The normalized data value is the specific value of a data in the normalized data set; According to the calculated dynamic weight, the target timestamp is calculated using weighted interpolation Normalized data values ​​on : .

[0022] The specific ways to use the PSO-SVM model to determine whether the pipette is faulty include: The input sample set includes L_o groups of samples, each of which includes a basic data set and corresponding fault data; Initialize the particle swarm parameters, including the number of particles, maximum number of iterations, speed and position; each particle position corresponds to a hyperplane parameter weight vector and the bias term ,in represents the current position of the i-th particle, , represents the weight of the pipette head wear data, represents the weight of the pipette battery voltage data, represents the weight of the pipette battery current data, represents the instantaneous skewness characteristic data, Represents the instantaneous kurtosis characteristic data and Represents instantaneous amplitude characteristic data; Fitness value calculation: Calculate the fitness value for each particle. The fitness function is based on the objective function of SVM, which is a weighted combination of maximizing the classification interval and minimizing the penalty for misclassification, that is: ,in, represents the number of samples, j represents the sample index, represents the term that maximizes the classification interval, represents the slack variable of the jth sample, indicating the degree of misclassification allowed, represents the penalty factor; Update the local optimal position and global optimal position of the particle. In each iteration, if the fitness value of the current particle is better than the historical optimal fitness value of the particle, the individual optimal position is updated; for the entire particle group, when the fitness value of a particle is better than the current global optimal fitness value, the global optimal position is updated; Better means that the current fitness value of the particle is higher than its best fitness value in history, and the current fitness value of the particle is higher than the best fitness value in history of other particles in the particle swarm; Particle swarm update rule: Each particle will update its position according to the current speed, local optimal position (that is, the best solution in the history of the particle) and global optimal position (that is, the best solution among all particles). The speed update formula of each particle is: ,in, and Represent the weight factors of the speed of the inertia part, the individual learning part and the global learning part, and The random numbers representing the individual learning part and the global learning part are used to introduce randomness, and the range is , represents the optimal position of the ith particle, represents the global optimal position of the particle swarm, represents the current velocity of the ith particle, Represents the current position of the i-th particle, that is, the current hyperplane parameter weight vector and the bias term ; The position update formula for each particle is: ; When the number of iterations reaches the set maximum value, the optimization is stopped and the global optimal position, that is, the optimal hyperplane parameters, is output. and , build a decision function , represents the transpose of the optimal weight vector, Represents a new input sample, when When , it indicates that the pipette is not faulty; when When , it indicates that the pipette is malfunctioning; The PSO-SVM model maximizes the interval between the two types of samples by optimizing the hyperplane, so that one side of the hyperplane corresponds to the non-fault class and the other side corresponds to the fault class. Therefore, when the value of the decision function is greater than 0, it means that the sample belongs to the non-fault class, and when it is less than or equal to 0, it means that the sample belongs to the fault class; The system should automatically record detailed information about the fault event, such as the time when the fault occurred and performance data, to perform pipette inspection.

[0023] When no failure occurs, the basic data of the pipette is used as input, and the specific method of using the XGBoost model to predict the pipetting accuracy data includes: The input accuracy sample set includes K_F group samples, each group of samples includes a basic data set and corresponding pipetting accuracy data; the data in the basic data set is the characteristic variable, and the pipetting accuracy data is the target variable; Initialize hyperparameters: Use grid search method to initialize the hyperparameters of XGBoost model; Define the final objective function. On the basis of using the mean square error as the objective function, introduce a regularization term to determine the final objective function. Initialize the prediction value, take B_P precision data to calculate the mean, and define it as the initial prediction value: ,in, represents the initial predicted value of pipetting accuracy, Indicates Pipetting accuracy data values, Indicates the index of pipetting precision data, and B_P indicates the total number of precision data; Calculate the residual. In each iteration, calculate the residual between the current predicted value and the true value: ,in, Indicates The residual of the wheel, Indicates Predicted value of pipetting accuracy of the wheel; In each round, a decision tree is constructed to fit the residual, a feature variable is selected, the gain of each data value in the feature variable is calculated, and the data value with the largest gain is selected as the current optimal partition point, which is divided into two sub-nodes, the left sub-tree and the right sub-tree. Each sub-tree is then partitioned using the gain, and the partitioning stops when the maximum depth of the set tree is reached. In each round of iteration, the prediction value of the model is updated through the newly generated tree until the set number of iterations is reached and the iteration is stopped.

[0024] The specific methods for setting the tolerance threshold of real-time correction accuracy include: Use historical pipetting accuracy to set the initial accuracy tolerance threshold: ,in, Represents the average value of historical pipetting accuracy, represents the standard deviation of historical pipetting accuracy, represents the adjustment factor; Get real-time comprehensive environmental data: ,in, and Respectively represent the real-time laboratory temperature, humidity and air pressure data, and Respectively represent the weights of the temperature, humidity and air pressure data of the laboratory, Represents real-time comprehensive environmental data; Use real-time comprehensive environmental data to correct the initial accuracy tolerance threshold and obtain the real-time corrected accuracy tolerance threshold: ; When the predicted pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, it indicates that there is an error deviation or performance problem, which affects the accuracy of the experimental results. Therefore, an early warning is issued and the accuracy must be corrected through calibration to ensure the reliability of subsequent operations. When the predicted pipetting accuracy data is less than the real-time corrected accuracy tolerance threshold, it means that the predicted pipetting accuracy is within the allowable tolerance range, the performance of the pipette still meets the requirements, and the operation can continue, avoiding unnecessary calibration and waste of resources.

[0025] This embodiment realizes accurate monitoring and dynamic adjustment of the pipette state through comprehensive data collection and real-time monitoring, combined with feature extraction, fault diagnosis, pipetting accuracy prediction and early warning mechanism, uses the PSO-SVM model to efficiently judge faults, and the XGBoost model to accurately predict pipetting accuracy. It combines real-time environmental data to correct the accuracy tolerance threshold, ensuring that the pipette always operates within the optimal accuracy range, and provides an efficient and intelligent equipment management and maintenance solution; Through real-time data collection and analysis, equipment failures and accuracy anomalies can be discovered in a timely manner, reducing human intervention and improving equipment stability and work efficiency. At the same time, the system can dynamically adjust the accuracy tolerance according to environmental changes, avoiding accuracy errors caused by environmental fluctuations and ensuring high accuracy of pipetting operations. In addition, the early warning and trigger calibration functions can effectively reduce equipment downtime and maintenance costs, extend equipment service life, and improve the overall work efficiency of the laboratory.

[0026] Example 2 See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a pipette management method, including: Step S1, collecting pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Step S2, performing feature extraction on the collected pipette piston pressure signal data to obtain extracted feature data; Step S3, preprocessing the pipette performance monitoring data and the extracted feature data to obtain a performance data set; Step S4: Based on the performance data set, use the PSO-SVM model to determine whether the pipette fails; Step S5: When no failure occurs, the basic data of the pipette is used as input, and the XGBoost model is used to predict the pipetting accuracy data; Step S6, setting a real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued and a calibration program is triggered to perform calibration.

[0027] Example 3 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned pipette management system and method is implemented.

[0028] Since the electronic device introduced in this embodiment is an electronic device used to implement a pipette management system and method in the embodiment of the present application, based on the pipette management system and method introduced in the embodiment of the present application, the technical personnel of the field can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of the field implement the electronic device used in the pipette management system and method in the embodiment of the present application, it belongs to the scope of protection of this application.

[0029] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0030] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A pipette management system, characterized in that: include: Data acquisition module: collects pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Feature extraction module: extracting features from the collected pipette piston pressure signal data to obtain extracted feature data; Data preprocessing: preprocessing the pipette performance monitoring data and extracted feature data to obtain a performance data set; Fault judgment module: Based on the performance data set, the PSO-SVM model is used to determine whether the pipette has a fault; Pipette accuracy detection module: When no failure occurs, the basic data of the pipette is used as input, and the XGBoost model is used to predict the pipette accuracy data; Early warning module: Set the real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued and the calibration program is triggered for calibration.

2. A pipette management system according to claim 1, characterized in that: The pipette performance monitoring data includes pipette tip wear data, pipette battery voltage and current data; The basic data of the pipette includes the pipetting volume, pipetting speed, liquid density, and liquid viscosity data; Laboratory environmental data includes laboratory temperature, humidity and air pressure data.

3. A pipette management system according to claim 2, characterized in that: The method for obtaining the wear data of the liquid pipetting head includes: Use laser to scan the new pipette head to obtain 3D data. Use the 3D data of the new pipette head as standard data. Then use laser to scan the current pipette head and obtain the current 3D data. Use the height difference between the current 3D data and the M_L points in the standard 3D data to quantify the wear data: ,in, represents the three-dimensional data of the standard data point q, Represents the three-dimensional data of the current data point q, Represents the total number of three-dimensional data points, represents the index of a 3D data point, Indicates the wear data of the pipette tip.

4. A pipette management system according to claim 3, characterized in that: The specific method of extracting feature data includes: The pressure signal data when the pipette piston moves is obtained, and the instantaneous skewness characteristic data, instantaneous kurtosis characteristic data and instantaneous amplitude characteristic data of the pressure signal are calculated. The formulas for the instantaneous skewness characteristic data and the instantaneous kurtosis characteristic data are: ; ,in, Represents the instantaneous skewness characteristic data and instantaneous kurtosis characteristic data at time t, Indicates the pressure signal Pressure value, Indicates the total number of pressure values ​​in the pressure signal. Indicates the average value of the pressure signal; The formula for instantaneous amplitude characteristic data is: , represents the instantaneous amplitude characteristic data at time t, represents the original pressure value at time t, Represents the Hibert transform pressure value at time t.

5. A pipette management system according to claim 4, characterized in that: The specific method of obtaining the performance data set includes: Based on the pipette performance monitoring data and extracted feature data, the interpolation method is used to process missing values, the Z_Score method is used to process outliers, and normalization is performed to obtain a normalized data set. The dynamic weighted interpolation method is used to unify the timestamps to obtain the basic data set.

6. A pipette management system according to claim 5, characterized in that: The specific method of using the dynamic weighted interpolation method to perform unified timestamp includes: Unify the formats of the collected pipette performance monitoring data and the timestamps corresponding to the extracted feature data; Set target timestamp , by the timestamp before the target timestamp and the next timestamp , to set the basic weights, , , Indicates timestamp The basic weight of Indicates timestamp The basic weight of Calculate the previous timestamp and the next timestamp The rate of change of the upper normalized data value: , , adjust the basic weight according to the change rate of the previous and next timestamps, and calculate the dynamic weight of the previous and next timestamps: Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , Indicates the timestamp The normalized data value on , represents the adjustment coefficient, Indicates the timestamp The rate of change on Indicates the timestamp The rate of change of According to the calculated dynamic weight, the target timestamp is calculated using weighted interpolation Normalized data values ​​on : .

7. A pipette management system according to claim 6, characterized in that: The specific method of using the PSO-SVM model to determine whether the pipette fails includes: The input sample set includes L_o groups of samples, each of which includes a basic data set and corresponding fault data; Initialize the particle swarm parameters, including the number of particles, maximum number of iterations, speed and position; each particle position corresponds to a hyperplane parameter weight vector and the bias term ; Fitness value calculation: Calculate the fitness value for each particle, the fitness function ,in, represents the number of samples, j represents the sample index, represents the term that maximizes the classification interval, represents the slack variable of the jth sample, represents the penalty factor; Update the local optimal position and global optimal position of the particle. In each iteration, if the fitness value of the current particle is better than the historical optimal fitness value of the particle, the individual optimal position is updated; for the entire particle group, when the fitness value of a particle is better than the current global optimal fitness value, the global optimal position is updated; Particle swarm update rules: each particle will update its position according to the current speed, local optimal position and global optimal position. The speed update formula is: ,in, and Represent the weight factors of the speed of the inertia part, the individual learning part and the global learning part, and The random numbers representing the individual learning part and the global learning part are used to introduce randomness, and the range is , represents the optimal position of the ith particle, represents the global optimal position of the particle swarm, represents the current velocity of the ith particle, Represents the current position of the i-th particle, that is, the current hyperplane parameter weight vector and the bias term ; The position update formula for each particle is: ; When the number of iterations reaches the set maximum value, the optimization is stopped and the global optimal position, that is, the optimal hyperplane parameters, is output. and , build a decision function , represents the transpose of the optimal weight vector, Represents a new input sample, when When , it indicates that the pipette is not faulty; when , it indicates that the pipette is malfunctioning.

8. A pipette management system according to claim 7, characterized in that: When no failure occurs, the basic data of the pipette is used as input, and the specific method of using the XGBoost model to predict the pipetting accuracy data includes: The input accuracy sample set includes K_F group samples, each group of samples includes a basic data set and corresponding pipetting accuracy data; the data in the basic data set is the characteristic variable, and the pipetting accuracy data is the target variable; Initialize hyperparameters: Use grid search method to initialize the hyperparameters of XGBoost model; Define the final objective function. On the basis of using the mean square error as the objective function, introduce a regularization term to determine the final objective function. Initialize the prediction value, take B_P precision data to calculate the mean, and define it as the initial prediction value: ,in, represents the initial predicted value of pipetting accuracy, Indicates Pipetting accuracy data values, Indicates the index of pipetting precision data, and B_P indicates the total number of precision data; Calculate the residual. In each iteration, calculate the residual between the current predicted value and the true value: ,in, Indicates The residual of the wheel, Indicates Predicted value of pipetting accuracy of the wheel; In each round, a decision tree is constructed to fit the residual, a feature variable is selected, the gain of each data value in the feature variable is calculated, and the data value with the largest gain is selected as the current optimal partition point, which is divided into two sub-nodes, the left sub-tree and the right sub-tree. Each sub-tree is then partitioned using the gain, and the partitioning stops when the maximum depth of the set tree is reached. In each round of iteration, the prediction value of the model is updated through the newly generated tree until the set number of iterations is reached and the iteration is stopped.

9. A pipette management system according to claim 8, characterized in that: The specific method of setting the real-time correction accuracy tolerance threshold includes: Use historical pipetting accuracy to set the initial accuracy tolerance threshold: ,in, Represents the average value of historical pipetting accuracy, represents the standard deviation of historical pipetting accuracy, represents the adjustment factor; Get real-time comprehensive environmental data: ,in, and Respectively represent the real-time laboratory temperature, humidity and air pressure data, and Respectively represent the weights of the temperature, humidity and air pressure data of the laboratory, Represents real-time comprehensive environmental data; Use real-time comprehensive environmental data to correct the initial accuracy tolerance threshold and obtain the real-time corrected accuracy tolerance threshold: ; When the predicted pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued, and the calibration procedure is triggered to perform the operation again; When the predicted pipetting accuracy data is less than the real-time corrected accuracy tolerance threshold, the operation continues.

10. A pipette management method, based on a pipette management system according to any one of claims 1 to 9, characterized in that: include: Step S1, obtaining pipette performance monitoring data, pipette piston pressure signal data, pipette basic data and laboratory environment data; Step S2, analyzing the acquired pipette piston pressure signal data to obtain extracted feature data; Step S3, preprocessing the pipette performance monitoring data and the extracted feature data to obtain a performance data set; Step S4: for the performance data set, use the PSO-SVM model to determine whether the pipette fails; Step S5: When no failure occurs, the basic data of the pipette is input into the XGBoost model to predict the pipetting accuracy data; Step S6, setting a real-time correction accuracy tolerance threshold. When the pipetting accuracy data is greater than or equal to the real-time correction accuracy tolerance threshold, an early warning is issued and a calibration program is triggered to perform calibration.

Images