Intelligent control system of high-speed centrifugal machine

Abstract

The invention relates to the technical field of centrifugal machine intelligent control, in particular to an intelligent control system of a high-speed centrifugal machine, and the system comprises a sensing module which is used for collecting operation state parameters of the centrifugal machine in real time, carrying out the synchronous sampling of the operation state parameters, and generating multi-mode operation data; the dynamic modeling module is used for correcting parameters of the dynamic physical model in real time so as to generate a characteristic parameter set reflecting the overall dynamic state of the centrifugal machine; the control execution module is used for adjusting the rotating speed and the running time of the centrifugal machine driving motor in real time to realize closed-loop control; the automatic control module is used for adaptively correcting control strategy parameters according to a learning result so as to generate a dynamically updated centrifugal control curve and sending the centrifugal control curve to the control execution module so as to execute real-time control; and the sample characteristic data module is used for storing the sample characteristic data. And the operation stability, the control precision and the self-adaptive capacity of the centrifugal machine can be improved.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent control of centrifuges, and more specifically to an intelligent control system for a high-speed centrifuge. Background Technology

[0002] High-speed centrifuges, as precision separation equipment commonly used in laboratories and industrial production, are widely applied in fields such as biopharmaceuticals, chemical analysis, environmental monitoring, and new material preparation. They primarily utilize the powerful centrifugal force generated by high-speed rotation to effectively separate components of different densities, particle sizes, or phases within a sample. With increasing demands for separation precision and the diversification of sample types, centrifuge control systems are gradually evolving from traditional mechanical and open-loop control towards intelligent, data-driven approaches.

[0003] Traditional high-speed centrifuge control systems typically rely on preset control curves and fixed operating parameters. During operation, the centrifugation process is mainly regulated by simply setting the rotation speed or time. This method can complete basic separation operations under standardized samples and constant environmental conditions. However, in practical applications, the dynamic characteristics of the centrifuge can change significantly due to factors such as sample density distribution, temperature fluctuations, airflow disturbances, and equipment aging. Fixed parameter control struggles to respond to these nonlinear disturbances in real time, often leading to decreased rotation speed stability, uneven sample stratification, or fluctuations in separation accuracy. In severe cases, it can even cause abnormal chamber vibration, increased energy consumption, and fatigue of mechanical components.

[0004] In existing technologies, to improve the operational accuracy of centrifuges, some systems have begun to introduce sensors to monitor their operating status. For example, operating signals are collected using speed sensors, temperature sensors, or vibration sensors to enable a certain degree of feedback regulation within the control system. However, most solutions only monitor a single parameter, lacking simultaneous acquisition and feature fusion of multi-source signals, thus failing to form a complete characterization of the operating status. For instance, temperature, pressure, and vibration signals exhibit sampling differences in the time dimension and have different signal feature dimensions, making it impossible to directly align and fuse the data for time-series modeling, resulting in insufficient physical consistency of the model.

[0005] Meanwhile, complex dynamic coupling relationships exist between multiple physical fields such as energy flow, mass flow, and mechanical vibration within the centrifuge chamber. Traditional control models based on empirical equations or static fitting are insufficient to reflect the dynamic behavior evolution of the centrifuge under different loads and speed ranges. Therefore, the control accuracy of existing systems relies on manual parameter tuning or repeated experiments, lacking adaptive modeling capabilities for changes in operating conditions. Summary of the Invention

[0006] The purpose of this invention is to address the aforementioned shortcomings by proposing an intelligent control system for a high-speed centrifuge.

[0007] The present invention adopts the following technical solution: An intelligent control system for a high-speed centrifuge, the system comprising: The sensing module is used to collect the operating status parameters of the centrifuge in real time, and to synchronously sample the operating status parameters and generate multimodal operating data; The dynamic modeling module establishes a dynamic physical model of the centrifuge chamber based on multimodal operating data and performs real-time correction on the parameters of the dynamic physical model to generate a set of characteristic parameters that reflect the overall dynamic state of the centrifuge. The control execution module adjusts the speed and running time of the centrifuge drive motor in real time based on the characteristic parameter set, operating status parameters and centrifuge control curve output by the automation control module, to achieve closed-loop control. The automated control module includes a recurrent neural network structure that performs self-learning based on the centrifuge's historical operating data, feature parameter set, and sample feature data. It adaptively corrects the control strategy parameters according to the learning results to generate dynamically updated centrifugation control curves and sends the centrifugation control curves to the control execution module for real-time control. The sample characteristic data module is used to store sample characteristic data, which are characteristic parameters used to characterize the physical or chemical properties of the sample.

[0008] Optionally, the dynamic modeling module includes: The constraint modeling unit constructs a dynamic physical model based on multimodal operating data. The dynamic physical model is constructed using a cross-domain coupled constraint graph structure, in which the constraint graph structure is centered on energy flow nodes, mass flow nodes, and vibration mode nodes, and is used to characterize the interaction relationships between multiple physical fields within the centrifuge cavity. The parameter update unit updates and coordinates the parameters of the dynamic physical model in real time based on the node association weights in the constraint graph structure.

[0009] Optionally, the sensing module includes: The signal acquisition unit is used to acquire the operating status parameters of the centrifuge during operation. The synchronous sampling unit performs time-synchronized sampling of the raw signals of the operating status parameters to maintain the consistency of multiple types of sensor signals on the time scale. The multimodal fusion unit performs feature normalization and multidimensional combination on the synchronously sampled operating state parameters to generate multimodal operating data.

[0010] Optionally, the control execution module includes: The target analysis unit analyzes the target rotation speed and target running time based on the feature parameter set. The instruction synthesis unit generates initial speed instructions and initial running time instructions based on the target speed and target running time, using the centrifugal control curve output by the automatic control module as a feedforward reference. The feedback acquisition unit is used to acquire the actual speed signal and running time status information in the centrifuge's operating status parameters in real time, and compare the acquisition results with the target speed and target running time; The calculation unit is adjusted to calculate the speed deviation and time deviation based on the comparison results, and generate the corresponding correction amount. The execution control unit synthesizes the initial speed command, the initial running time command, and the correction amount, and outputs them to the centrifuge drive motor to achieve closed-loop control.

[0011] Optionally, the automation control module includes: The input preprocessing unit receives historical operating data of the centrifuge, a set of characteristic parameters, and sample characteristic data, and performs time alignment and feature normalization on the input data. The recursive learning unit uses a recursive neural network structure to extract temporal features and recognize patterns from the preprocessed input data to form a dynamic state vector that reflects the operating trend of the centrifuge. The strategy correction unit calculates the correction amount of the control strategy parameters based on the dynamic state vector and adaptively updates the control strategy parameters. The curve generation unit generates a centrifugal control curve based on the updated control strategy parameters and sends the centrifugal control curve to the control execution module to perform real-time control.

[0012] Optionally, the sample feature data module includes: Data storage unit, used to store sample characteristic data; The feature indexing unit establishes a multi-dimensional index structure according to sample type, experimental conditions, and target separation accuracy to enable the retrieval and retrieval of sample feature data; The association mapping unit establishes a one-to-one association mapping relationship between sample characteristic data and centrifuge historical operation data and characteristic parameter set based on the experiment number information, so as to achieve traceability; During the startup phase of the automation control module, the data retrieval unit extracts the corresponding sample feature data from the data storage unit based on the input sample identification information or experimental conditions, and sends the sample feature data to the automation control module for calculation.

[0013] Optionally, the system also includes an intelligent health management module, which constructs an operational health model of the centrifuge based on a set of feature parameters and operating status parameters, and identifies potential abnormal states of the centrifuge by analyzing vibration signals, temperature change trends and energy consumption characteristics during multi-cycle operation. When an abnormal trend or operational deviation is detected, the automatic control module is triggered to execute an adaptive learning process to correct the control strategy parameters and dynamically adjust the centrifugal control curve, thereby achieving predictive maintenance of the system's health status and long-term stable operation.

[0014] Optionally, after the automatic control module performs adaptive learning, the intelligent health management module tracks the operating status of the centrifuge in real time and calculates the operating stability index based on the corrected control strategy parameters. When the operating stability index does not meet the preset threshold condition, the automatic control module is triggered to perform the adaptive learning process again.

[0015] The beneficial effects achieved by this invention are: Through the collaborative work of the sensing module, dynamic modeling module, control execution module, automated control module, and sample feature data module, a closed-loop intelligent control system with real-time perception, dynamic modeling, adaptive control, and knowledge self-evolution capabilities is constructed. This helps to improve the operational stability, control accuracy, and adaptive capability of the centrifuge, reduces the impact of human intervention on system performance, and realizes autonomous control of the entire process from data-driven to intelligent decision-making. Through the sensing module, the operating status parameters of the centrifuge can be collected in real time, and multimodal operating data can be generated by synchronous sampling and multimodal fusion, thereby ensuring the consistency of different types of sensor signals on the time scale and providing a unified and accurate data foundation for subsequent dynamic modeling and intelligent control. The dynamic modeling module uses multimodal operating data to establish a dynamic physical model of the centrifuge chamber. A cross-domain coupled constraint graph structure is used to realize the interrelation between energy flow, mass flow and vibration mode. This enables real-time correction of model parameters, accurately reflects the overall dynamic state of the centrifuge, and improves the physical interpretability and prediction accuracy of the model. By controlling the execution module, based on the characteristic parameter set, operating status parameters and the centrifugal control curve output by the automation control module, the speed and running time of the centrifuge drive motor can be adjusted in real time, forming a closed-loop control loop, thereby effectively improving the control response speed and operating stability. Through the automated control module, based on the recurrent neural network structure, the centrifuge's historical operating data, feature parameter set, and sample feature data are self-learned. This enables adaptive correction of control strategy parameters, generation of dynamically updated centrifugation control curves, and continuous optimization of control strategy and improvement of environmental adaptability. Through the sample feature data module, a multi-dimensional index structure is established for sample type, experimental conditions and target separation accuracy, and a correlation mapping relationship is established with the centrifuge's historical operating data and feature parameter set. This enables efficient retrieval, traceable recall and linkage of sample feature data with control strategies, allowing the system to achieve differentiated control for different sample characteristics.

[0016] To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall structure of Embodiment 1 of the present invention; Figure 2 This is a graph showing the relationship between the actual measured value of the nth physical quantity and the overall operational stability index in Embodiment 1 of the present invention. Figure 3 This is a diagram illustrating the relationship between rotational speed and temperature in Embodiment 1 of the present invention. Figure 4 This is a schematic diagram of the communication coordination module in Embodiment 2 of the present invention. Detailed Implementation

[0018] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated in advance. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0019] This embodiment provides an intelligent control system for a high-speed centrifuge, combined with... Figures 1 to 3 As shown.

[0020] An intelligent control system for a high-speed centrifuge, the system comprising: The sensing module is used to collect the operating status parameters of the centrifuge in real time, and to synchronously sample the operating status parameters and generate multimodal operating data; The dynamic modeling module establishes a dynamic physical model of the centrifuge chamber based on multimodal operating data and performs real-time correction on the parameters of the dynamic physical model to generate a set of characteristic parameters that reflect the overall dynamic state of the centrifuge. The control execution module adjusts the speed and running time of the centrifuge drive motor in real time based on the characteristic parameter set, operating status parameters and centrifuge control curve output by the automation control module, to achieve closed-loop control. The automated control module includes a recurrent neural network structure that performs self-learning based on the centrifuge's historical operating data, feature parameter set, and sample feature data. It adaptively corrects the control strategy parameters according to the learning results to generate dynamically updated centrifugation control curves and sends the centrifugation control curves to the control execution module for real-time control. The sample characteristic data module is used to store sample characteristic data, which are characteristic parameters used to characterize the physical or chemical properties of the sample.

[0021] Specifically, its overall structure includes a sensing module, a dynamic modeling module, a control execution module, an automated control module, and a sample feature data module. The system adopts a modular design to achieve end-to-end adaptive control from multi-source signal acquisition, state modeling, strategy optimization to closed-loop execution. Its core idea lies in constructing a digital mapping of the centrifuge's operating state through multimodal sensing and dynamic modeling, combined with a self-learning control strategy, to achieve high-precision centrifugation control under different sample types and complex operating conditions. During system operation, the modules interact with each other via a data bus or internal communication protocol, forming a complete intelligent control closed loop.

[0022] In this embodiment, the sensing module is used to collect the centrifuge's operating status parameters in real time and generate multimodal operating data after time-synchronized sampling of these parameters. The operating status parameters include, but are not limited to, drive motor speed, chamber temperature, air pressure, vibration acceleration, and power signals, with signal sources from various types of sensors installed in the chamber structure and motor control unit. Since the sampling frequencies, delays, and signal scales of each sensor differ, this module ensures that signals are collected under the same time reference through a unified sampling clock and synchronous triggering mechanism, thereby achieving time consistency. The synchronized multimodal signals undergo normalization and noise suppression processing within the module, and a fusion algorithm generates a multimodal operating dataset with a unified time identifier. This module's design enables the centrifuge to extract continuous and comparable physical state information from multi-source heterogeneous signals, providing a realistic and complete data foundation for subsequent modeling and control.

[0023] The dynamic modeling module receives multimodal operational data from the sensing module and builds a dynamic physical model of the centrifuge chamber based on this data. The dynamic physical model refers to the model structure that describes the coupling mechanism of energy flow, mass flow, and vibration modes of the centrifuge under different operating states through mathematical and physical constraints. In its implementation, the model is constructed using a cross-domain constraint graph structure, where nodes represent energy flow nodes, mass flow nodes, and vibration mode nodes within the system, and directed edges between nodes represent coupling paths between various physical fields. Real-time correction of model parameters is achieved through minimum deviation iteration and Bayesian weighted update mechanisms. Specifically, within each sampling period, the system compares the model's predicted output with the sensor's measured data, calculates the prediction residual, and adjusts the association weights of each node based on the residual vector. If the relative error between energy flow and vibration response exceeds a threshold, model parameter re-estimation is triggered, and correction is completed through dynamic gain adjustment and recursive least squares method. This real-time correction process enables the model to adaptively reflect the dynamic characteristic shifts of the centrifuge caused by load changes, temperature drift, or mechanical wear during operation, ensuring that the feature parameter set has stable physical interpretability. The set of characteristic parameters obtained after model calculation comprehensively characterizes the overall dynamic state of the centrifuge, including key indicators such as the degree of speed fluctuation, energy consumption balance offset, and structural vibration mode change rate, providing quantifiable dynamic state input for subsequent intelligent control.

[0024] The control execution module is used to adjust the speed and running time of the drive motor in real time based on the characteristic parameter set, operating status parameters, and the centrifugal control curve output by the automation control module, achieving high-precision closed-loop control. The centrifugal control curve is a dynamic trajectory curve generated by the automation control module, containing a three-dimensional correspondence between time, speed, and acceleration, reflecting the optimal speed change path of the centrifuge throughout its entire operating cycle. During operation, the system first calculates the target operating point for the current centrifugation stage based on the characteristic parameter set, and uses this target as a local anchor point on the control curve, generating a continuous control trajectory through curve interpolation. The control execution module outputs the target speed and running time commands for the current moment based on the control curve and inputs them to the drive controller. During the execution phase, the control execution module collects feedback signals in real time, compares the error with the target value at the corresponding position on the control curve, and when a deviation is detected, the adjustment calculation unit within the module calculates the correction amount based on the deviation trend, and corrects the output signal in real time through a proportional-integral-derivative adjustment algorithm or a model predictive control mechanism. The corrected signal is synthesized into a final control signal and input to the motor drive module, realizing dynamic adjustment of speed and running time, thus forming a parallel and coordinated closed-loop control system of "feedforward-feedback". This mechanism enables centrifuges to achieve dynamic and stable operation under complex conditions, and significantly improves the accuracy of control response and system robustness.

[0025] The automated control module, as the intelligent core of the system, is used to learn from historical experience and the current state during long-term operation, and adaptively adjust the control strategy parameters based on the learning results to generate dynamically updated centrifugation control curves. Internally, this module employs a recurrent neural network structure. Its inputs include historical operating data of the centrifuge, a feature parameter set generated by the dynamic modeling module, and sample feature data provided by the sample feature data module. The automated control module first performs time alignment and feature normalization on the input data to construct a unified input matrix, and then uses the recurrent neural network for time-series feature extraction and pattern recognition. The network structure captures the time dependency between historical and current states through hidden recurrent units, thus forming a dynamic state vector describing the system's evolution trend. The module calculates the adjustment amount of the control strategy parameters based on this dynamic state vector, including parameters such as speed gain factor, response delay compensation coefficient, and acceleration smoothing weight. Through gradient descent optimization and sliding window training mechanisms, the system can continuously reduce prediction errors and achieve adaptive parameter updates. Once the model has stabilized and converged, the curve generation submodule regenerates the centrifugation control curve based on the updated strategy parameters and transmits this curve to the control execution module for execution, thus forming an intelligent control process that is self-learning, self-correcting, and dynamically updating. The core of this process lies in the fact that the system can continuously learn the optimal control law through operating data without human intervention, enabling the centrifuge to automatically achieve the best operating trajectory under different sample and environmental conditions.

[0026] The sample characteristic data module stores, manages, and retrieves physical and chemical characteristic data related to samples, including sample density, particle size distribution, solution viscosity, and target separation accuracy. The module establishes a multi-dimensional index system using experiment numbers, sample identifiers, and operating conditions, creating a one-to-one mapping between sample characteristics and historical operating data and model characteristic parameter sets. When a centrifugation task is initiated, the system automatically retrieves the corresponding sample characteristic data based on the input sample identifier and sends this data to the automation control module as an initial reference for generating the centrifugation control curve. Through this mechanism, the control strategy can adaptively adjust to different sample characteristics, enabling the system to achieve high-precision and repeatable separation results in various task scenarios.

[0027] The system operation process in this embodiment is as follows: After the centrifuge starts, the sensing module collects operating status parameters in real time and generates multimodal operating data; the dynamic modeling module constructs and corrects the dynamic physical model based on this data and outputs a set of feature parameters; the automation control module receives historical operating data, sample feature data, and the set of feature parameters, performs a self-learning process, calculates the control strategy parameter correction amount, and generates a dynamically updated centrifugation control curve; the control execution module receives the centrifugation control curve and combines it with real-time feedback signals to perform closed-loop adjustment, implementing precise control over the speed and running time of the drive motor. Throughout the entire process, the system continuously records operating data and feeds it back to the automation control module to complete model retraining and strategy re-optimization, thereby forming an intelligent evolutionary closed loop of "perception—modeling—learning—control—relearning".

[0028] Through the solution in this embodiment, the centrifuge can identify its own dynamic state in real time during operation, automatically correct the control curve, and optimize the operating strategy, achieving high-precision separation control under complex operating conditions. This system not only improves control response speed and separation accuracy but also significantly enhances operational stability and adaptability, exhibiting intelligent operating characteristics of self-learning, self-correction, and long-term evolution.

[0029] Optional, the dynamic modeling module includes: The constraint modeling unit constructs a dynamic physical model based on multimodal operating data. The dynamic physical model is constructed using a cross-domain coupled constraint graph structure, in which the constraint graph structure is centered on energy flow nodes, mass flow nodes, and vibration mode nodes, and is used to characterize the interaction relationships between multiple physical fields within the centrifuge cavity. The parameter update unit updates and coordinates the parameters of the dynamic physical model in real time based on the node association weights in the constraint graph structure.

[0030] Specifically, the constraint modeling unit's function is to construct a dynamic physical model based on multimodal operational data obtained from the sensing module. This model employs a cross-domain coupled constraint graph structure, which simulates and characterizes the coupling relationships between multiple physics fields within the centrifuge chamber through nodes and edges. In practice, the constraint graph structure includes three main types of nodes: energy flow nodes, mass flow nodes, and vibration mode nodes. Energy flow nodes represent energy transfer and change, mass flow nodes describe the flow and distribution of matter during centrifugation, and vibration mode nodes reflect the mechanical vibration modes generated by the centrifuge system during operation and their impact on system stability. The nodes are connected by directed edges, representing the interactions between physical fields, such as the coupling between energy flow and mass flow, and the transfer of vibration and energy. Through these nodes and their associated edges, the constraint modeling unit can establish an accurate and dynamically changing centrifuge physical model in real time.

[0031] For example, suppose the data collected during operation includes signals such as rotational speed, temperature, pressure, and vibration acceleration. Using these signals, the constraint modeling unit can establish corresponding constraint relationships based on physical principles, dynamically describing the coupling and mutual influence between these physical quantities. In this way, the dynamic physical model can not only reflect the operating status of the centrifuge but also be dynamically updated in real time based on sensor data.

[0032] This dynamic physical model not only reflects the coupling relationships of various physical elements within the centrifuge chamber, but also adaptively adjusts according to changes in system operation (such as load changes, mechanical wear, temperature drift, etc.). Therefore, the establishment and calibration of the model can reflect the state of the centrifuge in real time, ensuring that the parameters during the centrifugation process remain accurate and stable.

[0033] The main function of the parameter update unit is to update and collaboratively adjust the parameters of the dynamic physical model in real time based on the node association weights in the constraint graph structure. During system operation, the dynamic modeling module continuously receives real-time data from the sensing module, including multiple physical parameters such as centrifuge speed, temperature, pressure, and vibration acceleration. The parameter update unit calculates the error by comparing the model's predicted values ​​with the actual measured values ​​(i.e., sensor data) and updates the parameters in the model based on the error. Specifically, the error is usually obtained through the calculation of the residual vector. The parameter update process utilizes algorithms such as least squares, Kalman filtering, and Bayesian optimization to dynamically adjust the node weights according to the error value, enabling the model to adapt to new operating states.

[0034] For example, in a certain operation, there might be a certain error between the energy flow value predicted by the system and the actual measured energy flow value. By calculating the residual vector, the system can identify the deviation in the model, adjust the weights of the energy flow nodes in the constraint diagram according to the deviation, minimize the error using the least squares method, and recalibrate the node parameters based on new sensor data. This adjustment ensures that the dynamic physical model can reflect the dynamic changes of the centrifuge during actual operation due to load variations, temperature fluctuations, or mechanical wear, thereby improving the model's accuracy and predictive ability.

[0035] The real-time correction process is completed through dynamic gain adjustment and recursive least squares method. During operation, if the relative error between energy flow and vibration response exceeds a set threshold, the model parameters are re-estimated, and the model is further optimized through a Bayesian weighted update mechanism. Within each sampling period, the system can continuously correct and optimize various parameters of the model in this way to ensure the best control effect of the centrifuge under different operating conditions.

[0036] Each time the model is updated, the coupling relationships between relevant physical fields are adjusted based on data feedback from sensors, and the model's prediction accuracy is further optimized based on the new data. Especially when load changes, mechanical aging occurs, or environmental conditions change, the parameter update unit can dynamically adjust the node weights in the model, enabling the model to adapt to new operating conditions.

[0037] Through the above process, the dynamic modeling module continuously updates and optimizes the model in each operating cycle, enabling the centrifuge's control system to reflect changes in equipment status in real time and maintain stable control performance. Through real-time correction and updates, the system can cope with various changes under complex operating conditions, ensuring that the centrifuge maintains optimal separation performance under all operating conditions.

[0038] In summary, the dynamic modeling module, through constraint modeling units and parameter update units, combined with real-time multimodal data, constraint graph structure, and node association weights, establishes and optimizes the dynamic physical model of the centrifuge, thereby providing an accurate and dynamically updated set of feature parameters for the subsequent control execution module. This module enables the centrifuge to maintain high-precision and high-stability control performance under constantly changing operating conditions.

[0039] The following example demonstrates how to perform dynamic modeling based on real-time acquired multimodal data and update the parameters of the dynamic physical model.

[0040] Suppose that during the operation of a high-speed centrifuge, the following real-time data are obtained: rotational speed (RPM): 10,000 revolutions per minute; temperature (°C): 37°C; vibration acceleration (m / s²). 2 ): 0.2m / s 2 Air pressure (Pa): 101325Pa; These data were collected synchronously by the sensing module and formed into a multimodal dataset after time-synchronized sampling.

[0041] The calculation process of the constraint modeling unit: Assume the dynamic physical model uses a constraint graph structure, including the following three types of nodes: Energy flow nodes (E): representing energy transfer and change; Mass flow nodes (M): describing the flow and distribution of matter during centrifugation; Vibration mode nodes (V): reflecting the mechanical vibration modes generated by the centrifuge.

[0042] A preliminary constraint diagram is constructed: the energy flow node reflects the energy input of the centrifuge drive motor. The mass flow node is coupled to the energy flow node through a mass-energy transfer relationship, representing the flow of matter under centrifugal force during centrifugation. The vibration mode node reflects the vibration characteristics of the centrifuge during high-speed rotation and is coupled to the energy flow node, representing the impact of vibration on energy distribution and transfer.

[0043] Suppose that the relationship between the energy flow node, the mass flow node, and the vibration mode node can be expressed by the following constraint equation: E(t)=k1*M(t)+k2*V(t), where k1 and k2 are model parameters that need to be corrected by sensor data, representing the contribution of mass flow and vibration mode to energy flow, respectively; t represents time.

[0044] The calculation process of the parameter update unit: Assuming the following data is obtained after a certain sampling period: Measured rotational speed: 10000 rpm (converted to ω = 10000 * 2π / 60 ≈ 1047.2 rad / s); Measured temperature: 37°C; Measured vibration acceleration: 0.2 m / s². 2 .

[0045] The model parameters need to be updated through the following steps: Step 1: Error Calculation First, the energy flow (Epred) is predicted using initial parameters (assuming k1=0.5 and k2=0.3): Epred = 0.5 * M(t) + 0.3 * V(t); where k1 = 0.5 and k2 = 0.3 are empirical assumptions. These initial choices are merely to ensure the model can start and perform preliminary calculations. In the dynamic modeling of the centrifuge, mass flow (k1) and vibration mode (k2) are two important factors affecting energy flow. Assuming the mass flow has a larger impact and the vibration mode has a smaller impact, a larger k1 (0.5) and a smaller k2 (0.3) are chosen.

[0046] Assume that the mass flow rate M(t) and vibration acceleration V(t) during this time period are calculated from sensor data and set as follows: M(t) = 1.2 kg / s; V(t) = 0.2 m / s 2 ; Therefore, the energy flow predicted by the model is: Epred=0.5*1.2+0.3*0.2=0.6+0.06=0.66 W; Next, the actual energy flow (Eact) is obtained as follows: Eact = 0.8 W; this value, which may be measured by a sensor, represents the operating status of the centrifuge under specific conditions.

[0047] Calculation error (residual): Residual=Eact-Epred=0.8-0.66=0.14 W; Step 2: Model parameter update Parameter updates are performed using residuals, assuming the weights of k1 and k2 are updated using least squares or Bayesian optimization. Through iterative optimization of the algorithm, new parameters are assumed to be obtained: k1=0.55; K2=0.32 Step 3: Corrected energy flow prediction Recalculate the energy flow using the new parameters: Enew=0.55*1.2+0.32*0.2=0.66+0.064=0.724 W; This correction brings the model closer to the actual energy flow (0.8W), improving prediction accuracy.

[0048] Step 3: Summary and Results Through the steps described above, the dynamic modeling module successfully updated the model using real-time acquired data. In actual operation, as the sampling cycle continues, the system can gradually correct the parameters of the physical model based on sensor data, ensuring that the dynamic physical model can adapt to changes in equipment status. For example, under the influence of factors such as load changes, ambient temperature fluctuations, or mechanical wear, the dynamic physical model always maintains a high degree of accuracy, thus providing accurate real-time data support for the centrifuge control system.

[0049] Ultimately, through this model, the centrifuge can continuously optimize its control strategy in a changing operating environment, ensuring the high efficiency, stability, and accuracy of the centrifugation process. During centrifugation control, the dynamic modeling module and the control execution module work closely together to form a complete closed-loop control system.

[0050] Optionally, the sensing module includes: The signal acquisition unit is used to acquire the operating status parameters of the centrifuge during operation. The synchronous sampling unit performs time-synchronized sampling of the raw signals of the operating status parameters to maintain the consistency of multiple types of sensor signals on the time scale. The multimodal fusion unit performs feature normalization and multidimensional combination on the synchronously sampled operating state parameters to generate multimodal operating data.

[0051] Specifically, the sensing module preferably consists of a signal acquisition unit, a synchronous sampling unit, and a multimodal fusion unit forming a data processing link in sequence to ensure high-reliability acquisition, time-consistent alignment, and integrated feature output of operating status parameters during centrifuge operation. The signal acquisition unit establishes stable data channels with sensors installed on the drive motor, centrifuge chamber, and frame structure. These sensors include a speed sensor for acquiring rotational speed, a temperature sensor for acquiring chamber temperature, a pressure sensor for acquiring chamber pressure / air pressure, a vibration acceleration sensor for acquiring structural response, and a power measurement device for acquiring energy consumption. To ensure the quality of the original signal, the signal acquisition unit performs sensor zero-point / sensitivity calibration, range self-check, and channel consistency verification before data enters the bus. Anti-aliasing filtering and overload protection are implemented on the analog input path; communication integrity verification and data packet loss detection are performed on digital inputs; all original sampling frames are accompanied by the source sensor identifier and sampling timestamp, providing a clear time reference and data source traceability for subsequent alignment and fusion. The synchronous sampling unit is used to perform time-synchronized sampling and unified time-base shaping of raw signals from different sensors, different sampling periods, and different communication delays. It distributes a synchronization reference to each acquisition front end through a unified reference clock or hardware trigger pulse. For channels that cannot be directly hard synchronized, it adopts a high-precision timestamp alignment and interpolation resampling strategy to map the signals of each channel to a unified time grid. To reduce time deviations caused by asynchronous backhaul, bus congestion, or jitter, the synchronous sampling unit sets up an input buffer queue and arrival time sorting mechanism. It uses amplitude-limiting interpolation or robust interpolation within a sliding window to fill in short-term missing samples and records the synchronization deviation estimate in the frame header to ensure strict consistency of multiple types of sensor signals on the time scale. For the frequency band difference between high-frequency vibration signals and low-frequency temperature / pressure signals, the synchronous sampling unit performs bandpass / low-pass shaping and downsampling ratio alignment strategies after alignment to enable each channel to have comparable time-frequency characteristics.The multimodal fusion unit is used to perform feature normalization and multidimensional combination of the synchronized operating state parameters to generate multimodal operating data. First, it performs scale normalization and outlier suppression on each channel based on sensor calibration factors and historical statistics, and adopts a robust denoising method to reduce interference caused by transient spikes and operating condition switching. Then, it extracts basic time-domain features (such as mean, variance, slope, peak-to-valley difference, pulsation index, etc.) and frequency domain / time-frequency domain features (such as dominant frequency, energy spectrum components, spectral kurtosis, envelope energy, etc.) related to control within a sliding time window, and forms a unified feature vector by feature splicing and timestamp alignment. To avoid scale and redundancy effects between multi-source features, the multimodal fusion unit adopts orthogonalization / correlation reduction and weight normalization strategies to shape the features, and performs downweighting or masking on abnormal channels based on preset channel confidence and data integrity evaluation. Finally, it outputs a multimodal operating data frame with a unified time base, channel label, and quality label as input for the subsequent dynamic modeling module. Through the above implementation methods, the signal acquisition unit ensures the authenticity and traceability of the original data, the synchronous sampling unit ensures the strict consistency of multi-source signals in the time dimension, and the multimodal fusion unit completes the construction of discriminable features related to control, thereby generating a physically interpretable and real-time decision-making basis for the dynamic modeling module.

[0052] Optionally, the control execution module includes: The target analysis unit analyzes the target rotation speed and target running time based on the feature parameter set. The instruction synthesis unit generates initial speed instructions and initial running time instructions based on the target speed and target running time, using the centrifugal control curve output by the automatic control module as a feedforward reference. The feedback acquisition unit is used to acquire the actual speed signal and running time status information in the centrifuge's operating status parameters in real time, and compare the acquisition results with the target speed and target running time; The calculation unit is adjusted to calculate the speed deviation and time deviation based on the comparison results, and generate the corresponding correction amount. The execution control unit synthesizes the initial speed command, the initial running time command, and the correction amount, and outputs them to the centrifuge drive motor to achieve closed-loop control.

[0053] Specifically, the main function of the target parsing unit is to parse the target rotational speed and target running time required by the centrifuge for the current stage from the feature parameter set and operating status parameters. The feature parameter set contains the dynamic status information of the centrifuge under the current operating conditions (such as rotational speed fluctuations, energy consumption changes, temperature, etc.), while the target rotational speed and target running time are the parameters that the centrifuge needs to achieve according to the task requirements. During operation, the target parsing unit calculates the target rotational speed and target running time based on these parameters and determines the target task that the centrifuge needs to perform.

[0054] The task of the command synthesis unit is to generate initial speed commands and initial running time commands based on the target rotational speed and target running time, combined with the centrifugation control curve provided by the automated control module. First, the command synthesis unit uses the target rotational speed and target running time as reference inputs and generates a smooth rotational speed change curve and running time command through an interpolation algorithm. Next, the unit combines these target values ​​with the dynamic centrifugation control curve generated by the automated control module (this curve is adaptively corrected based on historical data, sample characteristic data, and system state) to ensure that the generated commands more accurately reflect the actual needs of the system.

[0055] The task of the feedback acquisition unit is to collect the actual operating status parameters of the centrifuge in real time (including speed signals and running time information) and compare them with the target values. The feedback acquisition unit continuously monitors the current operating status of the centrifuge through sensors and a data acquisition system. Specifically, it periodically collects the speed and actual running time. This data is compared with the target speed and target running time to calculate the deviation value, providing the necessary basis for subsequent adjustment calculations.

[0056] The adjustment calculation unit calculates the deviation from the target value (speed deviation and time deviation) based on the actual speed signal and running time status information provided by the feedback acquisition unit, and generates a corresponding correction amount. For example, assuming the target speed is 12000 rpm and the actual speed is 11850 rpm, then the speed deviation is: Speed ​​deviation = 12000 - 11850 = 150 revolutions per minute; Similarly, if the target running time is 10 minutes and the actual running time is 9 minutes, then the time deviation is: Time deviation = 10 - 9 = 1 minute; After calculating the deviations, the adjustment calculation unit will use proportional-integral-derivative (PID) algorithms, model predictive control (MMC), or other optimization algorithms to calculate corrections based on these deviation values. These corrections will then be used to adjust the control signals so that the system can achieve the target more accurately.

[0057] The task of the execution control unit is to combine the initial speed command, the initial running time command, and the correction amount to generate the final control signal, and then output it to the centrifuge drive motor to achieve closed-loop control. First, the execution control unit receives the initial speed command and running time command from the command synthesis unit, and synthesizes them with the correction amount calculated by the adjustment calculation unit to form a new control signal. This new control signal is then sent to the centrifuge's motor drive module in real time, thereby achieving dynamic adjustment of the centrifuge's speed and running time.

[0058] Through the collaboration of five sub-units—target analysis, instruction synthesis, feedback acquisition, adjustment calculation, and execution control—the control execution module can respond to changes in operating conditions in real time, maintain the stable and efficient operation of the centrifuge, and ensure the accuracy and consistency of the centrifugation process.

[0059] Optional, the automation control module includes: The input preprocessing unit receives historical operating data of the centrifuge, a set of characteristic parameters, and sample characteristic data, and performs time alignment and feature normalization on the input data. The recursive learning unit uses a recursive neural network structure to extract temporal features and recognize patterns from the preprocessed input data to form a dynamic state vector that reflects the operating trend of the centrifuge. The strategy correction unit calculates the correction amount of the control strategy parameters based on the dynamic state vector and adaptively updates the control strategy parameters. The curve generation unit generates a centrifugal control curve based on the updated control strategy parameters and sends the centrifugal control curve to the control execution module to perform real-time control.

[0060] Specifically, the input preprocessing unit is responsible for receiving historical operational data, feature parameter sets, and sample feature data from the sensing module, and performing time alignment and feature normalization on this input data. After time alignment, all data is mapped to a unified time axis using interpolation methods (such as linear interpolation or spline interpolation), ensuring temporal consistency across different data sources. Feature normalization maps different sensor signals to the same dimensional range, avoiding inaccurate comparisons caused by dimensional differences between different physical quantities (such as temperature, rotational speed, vibration acceleration, etc.). Through data normalization and standardization, the input preprocessing unit provides a unified foundation for subsequent model training and optimization.

[0061] The recursive learning unit employs a recursive neural network structure, utilizing historical operating data, feature parameter sets, and sample feature data to extract time-series features and generate a state vector describing the dynamic operating state of the high-speed centrifuge. Through multiple layers of recursive units (such as LSTM or GRU), this learning process can capture long-term dependencies in the time-series data, thereby identifying trends and patterns in the centrifuge's operation. The final output state vector effectively reflects the current operating state of the high-speed centrifuge and provides a basis for subsequent control strategy adjustments.

[0062] The strategy correction unit calculates and adaptively updates the control strategy parameters based on the dynamic state vector. These control strategy parameters include the speed gain factor, acceleration smoothing weight, and response delay compensation coefficient, which specifically depend on the operating state of the high-speed centrifuge. The correction amount of the control strategy parameters is determined based on the calculation results of the dynamic state vector. For example, if the system detects that the current speed deviates from the target value, the strategy correction unit will determine how much to adjust the speed gain factor by calculating the deviation in the state vector, thereby enabling the system to respond quickly and optimize control.

[0063] Assuming the current dynamic state vector output has a large speed deviation, the strategy correction unit will calculate the required correction amount and adjust the relevant control strategy parameters. Through optimization algorithms (such as least squares, gradient descent, or Bayesian optimization), the system can accurately calculate the correction amount, thereby making the control strategy more adaptable to the current working environment and sample characteristics.

[0064] The curve generation unit generates a centrifugal control curve based on the updated control strategy parameters. This control curve reflects the trajectory of the target speed and acceleration of the high-speed centrifuge throughout its entire operating cycle. In specific implementation, the curve generation unit generates target speed and target running time commands that conform to the centrifuge's operating state, based on the updated control strategy parameters and the current characteristic parameter set of the centrifuge. The generated control curve is then sent to the control execution module as the basis for executing real-time control.

[0065] In the control execution module, the received control curve will be further adjusted by comparing it with the real-time feedback signal to ensure that the centrifuge can accurately achieve the target operating state.

[0066] Through continuous adjustment and optimization of the automated control module, the centrifuge can maintain optimal operating accuracy and stability under different sample types, environmental conditions, and load variations. The system adaptively updates control strategy parameters and generates new centrifugation control curves in real time, forming a closed-loop control process of "sensing-modeling-learning-control." Through this process, the centrifuge's automated control capabilities are continuously enhanced, enabling it to self-adjust and optimize the centrifugation process.

[0067] The working principle of the strategy correction unit is explained below: Assume that a high-speed centrifuge is being used for separation operations, and the following operational data is being collected in real time via sensors: Rotational speed (RPM): 10,000 revolutions per minute (i.e., ω = 10,000 × 2π / 60 ≈ 1047.2 rad / s); Vibrational acceleration (V): 0.2 m / s² 2 Temperature (T): 37°C; Pressure (P): 101325 Pa Step 1: Input data and feature parameter set In the automation control module, the input data includes real-time data from the sensing module (such as rotation speed, vibration, temperature, etc.) as well as historical operating data of the centrifuge and sample characteristic data.

[0068] Suppose a dynamic state vector has been generated through data preprocessing (such as normalization and time alignment), and this vector consists of multiple features: S(t)=[ω(t),V(t),T(t),P(t)]; In this example, the dynamic state vector S(t) reflects the state of the centrifuge at a certain moment, including rotational speed, vibration acceleration, temperature and pressure.

[0069] Step 2: Calculate the correction amount for the control strategy parameters. The control strategy parameters include the speed gain factor (kω), acceleration smoothing factor (kV), and temperature regulation coefficient (kT). These parameters directly affect the control performance of the centrifuge.

[0070] Suppose that the control strategy corresponding to the current dynamic state vector S(t) is learned through the recurrent neural network, which should be to adjust the speed gain factor kω and the acceleration smoothing factor kV. Since the temperature has little effect on speed control, the temperature adjustment coefficient kT is relatively fixed.

[0071] Based on the current dynamic state vector S(t) = [1047.2 rad / s, 0.2 m / s] 2 [37°C, 101325 Pa], calculate the error between the model's output (predicted value) and the actual operating value: Predicted rotational speed deviation = ωpred - ωact = 1047.2 rad / s - 1050.0 rad / s = -2.8 rad / s; Predicted vibration deviation = Vpred - Vact = 0.18 m / s 2 -0.2 m / s 2 =-0.02 m / s 2 ; These errors reflect the inadequacies of the current control strategy for the centrifuge. Then, using the errors and the dynamic state vector, the correction amounts for the control strategy parameters are calculated.

[0072] Step 3: Adjust parameters using optimization algorithms In this process, the least squares method (or gradient descent, or other optimization algorithms) is used to update the control policy parameters. Assuming the least squares method is used for parameter adjustment, the goal is to adjust the control parameters by minimizing the sum of squared errors.

[0073] The objective function of the least squares method is: ; In minimizing this objective function, the optimization algorithm reduces the error between the prediction and the actual value by adjusting kω and kV.

[0074] After multiple iterations and optimizations, new control strategy parameters were obtained: kω=0.55 (increased the speed gain factor); kV=0.32 (acceleration smoothing factor added); These updated control strategy parameters reflect more precise adjustments to speed and vibration control under current operating conditions.

[0075] Step 4: Generation of updated control curves Based on the updated control strategy parameters, the curve generation unit generates a centrifugation control curve according to the new parameters. This control curve reflects the optimal speed change path throughout the centrifugation process, ensuring that the centrifuge achieves precise speed and acceleration control at different time points.

[0076] For example, based on the new control parameters kω=0.55 and kV=0.32, the new centrifugation control curve will be more accurate than the previous control curve, thus ensuring the stable operation of the centrifuge throughout the process.

[0077] Step 5: Send the control curve to the control execution module The updated centrifuge control curve is sent to the control execution module via the data bus. Upon receiving the control curve, the control execution module adjusts the motor speed and running time in real time according to the curve's instructions, achieving closed-loop control. At this point, the control execution module not only relies on the current target speed but also further refines the instructions based on real-time feedback signals, ensuring continuous optimization of the control process under different operating conditions.

[0078] Optionally, the sample feature data module includes: Data storage unit, used to store sample characteristic data; The feature indexing unit establishes a multi-dimensional index structure according to sample type, experimental conditions, and target separation accuracy to enable the retrieval and retrieval of sample feature data; The association mapping unit establishes a one-to-one association mapping relationship between sample characteristic data and centrifuge historical operation data and characteristic parameter set based on the experiment number information, so as to achieve traceability; During the startup phase of the automation control module, the data retrieval unit extracts the corresponding sample feature data from the data storage unit based on the input sample identification information or experimental conditions, and sends the sample feature data to the automation control module for calculation.

[0079] Specifically, the data storage unit stores sample-related characteristic data, including the sample's physical properties (such as density, particle size distribution, and solution viscosity) and chemical properties (such as target separation accuracy and solubility). This data is acquired in real time by experimental equipment or sensors and stored in the database along with sample identifiers (such as experiment number and batch number), ensuring the integrity and traceability of all sample characteristic data.

[0080] The feature indexing unit establishes a multi-dimensional index structure for sample feature data based on multiple dimensions such as sample type, experimental conditions, and target separation accuracy. This structure enables the system to efficiently retrieve data related to specific samples and provides rapid data support for subsequent calculations. The feature indexing unit indexes data according to dimensions such as sample type (e.g., biological samples, chemical samples), experimental conditions (e.g., temperature, rotation speed, experimental time), and target separation accuracy, allowing the system to quickly locate and extract relevant sample feature data.

[0081] The association mapping unit establishes a one-to-one correspondence between sample feature data and the centrifuge's historical operating data and feature parameter set. This mapping relationship allows sample feature data to be combined with centrifuge operating data under different experimental conditions (such as speed, time, temperature, etc.), providing support for model training and optimization. Through experiment numbers or other unique identifiers, the system can find relevant historical data for each experiment, forming a complete data chain and ensuring data traceability.

[0082] The data retrieval unit extracts the corresponding sample characteristic data from the data storage unit based on the input sample identification information or experimental conditions, and then transmits this data to the automation control module for calculation. When starting the experiment, the data retrieval unit searches for the corresponding sample characteristic data by querying the sample identification information (such as sample number, experimental conditions, etc.) and transmits it to the automation control module for self-learning and optimization of the control strategy.

[0083] For example, suppose there are two samples, sample A and sample B, with different densities, particle size distributions, and separation accuracy requirements. During operation, the system automatically retrieves the corresponding sample characteristic data using the experiment number and sample identification information. For instance, sample A may require a lower centrifugation rate and a longer running time, while sample B requires a higher centrifugation rate and a shorter running time. Through the data retrieval unit, the sample characteristic data is extracted and sent to the automation control module, helping the control system adjust its control strategy according to the characteristics of different samples.

[0084] Through the collaborative work of the data storage unit, feature indexing unit, association mapping unit, and data retrieval unit, the sample feature data module can efficiently manage and retrieve sample-related feature data, ensuring data traceability under different experimental conditions. This provides the necessary data support for the automated control module, enabling precise control and optimization of the centrifuge under different sample types and experimental conditions.

[0085] Optionally, the system also includes an intelligent health management module, which constructs an operational health model of the centrifuge based on a set of feature parameters and operating status parameters, and identifies potential abnormal states of the centrifuge by analyzing vibration signals, temperature change trends and energy consumption characteristics during multi-cycle operation. When an abnormal trend or operational deviation is detected, the automatic control module is triggered to execute an adaptive learning process to correct the control strategy parameters and dynamically adjust the centrifugal control curve, thereby achieving predictive maintenance of the system's health status and long-term stable operation.

[0086] Optionally, after the automatic control module performs adaptive learning, the intelligent health management module tracks the centrifuge's operating status in real time and calculates the operating stability index based on the corrected control strategy parameters. When the operating stability index fails to meet the preset threshold conditions, the automatic control module is triggered to perform the adaptive learning process again.

[0087] Specifically, operational stability indicators This is a key indicator for evaluating whether the current operating status of the centrifuge meets the preset stability standard. This indicator is calculated based on the modified control strategy parameters, taking into account the errors between multiple physical quantities and the impact of each physical quantity on system stability.

[0088] The formula for the stability index is as follows: ; in, For the first The actual measured value of a physical quantity; This is the target value for the physical quantity; This is the standard deviation of the physical quantity; These are dynamic weighting coefficients, reflecting the importance of the physical quantity to the system stability at the current moment; This is the overall operational stability index, representing the overall stability of the system. The unit can be relative error or a dimensionless value.

[0089] Calculation of dynamic weights: Introducing dynamic weights This allows the importance of different physical quantities in stability assessment to change over time. Weighting coefficients The weight of speed is dynamically determined by the system's operating status. For example, when the rotational speed is unstable, the weight of rotational speed will increase; conversely, if the vibration is small, the weight of vibration can be reduced.

[0090] The formula for calculating dynamic weights is as follows: ; in, For physical quantities A function relating the actual measured values. For each physical quantity, the weights are adjusted based on its sensitivity to the centrifuge stability.

[0091] ; in, ; This represents the standard deviation of the physical quantity.

[0092] Specifically, the standard deviation of this physical quantity is the number of samples collected within the statistical window. After obtaining N actual measurements of a physical quantity, the average value of this set of measurements can be calculated first and used as a benchmark to reflect the overall level of the physical quantity. Then, the deviations between the actual measurements and the average value at each sampling time are squared and summed, divided by a normalization factor of sample size minus 1, and finally the square root of the result is taken to obtain the Nth value. The standard deviation of a physical quantity. The standard deviation obtained by this calculation method can reflect the fluctuation intensity of the physical quantity within the time window, so that the errors between different physical quantities in the subsequent stability index calculation can be compared on a uniform scale.

[0093] Calculation of the error transfer function: To introduce more innovation and complexity, an error propagation function was incorporated into the stability assessment. The system transfer function takes into account the interactions between various physical quantities. Specifically, fluctuations in rotational speed may cause an increase in vibration, which in turn may affect other physical quantities such as temperature. This transfer function captures these nonlinear effects, improving the accuracy of system evaluation.

[0094] Error propagation function As shown below: ; in, For the first Error of a physical quantity; physical quantity and physical quantities The coupling coefficient between them reflects their interactive influence. Through... The system can assess the error propagation between different physical quantities, thereby making a more accurate stability analysis.

[0095] about The coupling coefficient reflects the degree of mutual influence between two physical quantities. In other words, the coupling coefficient indicates how one physical quantity is affected when the other changes, and the strength of that effect. Different physical quantities often interact, especially in complex physical systems, such as high-speed centrifuges, whose operating state is affected by the interaction of multiple factors (such as rotational speed, temperature, pressure, vibration, etc.).

[0096] ; in, and For the two physical quantities being measured; This is the error term, representing the deviation between the actual measurement and the theoretical model.

[0097] First, it is necessary to measure and record multiple sets of data through experiments, including at least... and A one-to-one correspondence. For example, the following data can be collected: (Experiment No. 1: The speed is 5000. (Temperature: 35) (Experiment No. 2:) The rotation speed is 10000. (Temperature: 37) (Experiment No. 3:) The speed is 15000. (Temperature: 40) (Experiment No. 4:) The speed is 20,000. (Temperature: 42°C) These data represent the rotational speed under different experimental conditions. and temperature The relationship.

[0098] Construct a regression model and use the least squares method to fit the linear model, minimizing the sum of squares of the error term ϵ. Assume the error term ϵ follows a normal distribution, and... These are the optimal parameters that need to be obtained through regression analysis.

[0099] To minimize the error, in order to find the most suitable Minimize the sum of squared errors for each data point, i.e.: ; This objective function represents the predicted value for all sample points. Compared with actual value The sum of the deviations between them. Next, solve using the least squares method. .

[0100] Expand the objective function and... Find the partial derivative, then set it equal to 0 to obtain the formula for the solution: ; The formula is derived from a weighted average of the data points. .

[0101] This second embodiment provides an intelligent control system for a high-speed centrifuge, combined with... Figure 4 As shown; Example 2 includes all the contents of Example 1.

[0102] An intelligent control system for a high-speed centrifuge includes a communication coordination module for establishing a unified data communication channel among a sensing module, a dynamic modeling module, a control execution module, an automation control module, a sample characteristic data module, and an intelligent health management module. This enables data synchronization, timing coordination, and fault tolerance among the multiple modules, ensuring the real-time performance and stability of the system.

[0103] Optionally, the communication coordination module includes: The data bus unit is used to establish a data interaction channel between the sensing module, dynamic modeling module, control execution module, automation control module, sample feature data module, and intelligent health management module, so as to realize data interconnection and synchronous calling between multiple modules; The anomaly monitoring unit is used to detect data loss, delay, or abnormal interruption during communication, and to trigger a fault-tolerant retransmission mechanism when an anomaly is detected to ensure the data integrity and stable operation of the system.

[0104] Specifically, data loss refers to the loss of data packets during data transmission between modules due to network problems or other technical factors. The anomaly monitoring unit monitors the integrity of data transmission in real time by assigning a unique identifier (such as a timestamp or sequence number) to each data packet. If a data packet is not received in time, the system marks it as lost and triggers a retransmission mechanism. In this way, the anomaly monitoring unit ensures that all data transmitted between modules is received accurately.

[0105] During data transmission, latency can impact the system's real-time performance. The anomaly monitoring unit monitors the transmission time of each data packet and compares it to a preset maximum latency threshold. If the packet transmission time exceeds the threshold, the system identifies it as a latency anomaly. The monitoring unit can automatically adjust the data transmission path, optimize network resource usage, or use caching mechanisms to reduce data transmission latency, thereby ensuring that the centrifuge system's real-time response is unaffected.

[0106] Communication interruption refers to the sudden loss of connection between two modules, causing data transmission to cease. The anomaly monitoring unit periodically checks the system's network connectivity. When a communication link is detected as broken, the system automatically initiates a reconnection mechanism to attempt to restore the network connection. If the connection cannot be restored within a set time, the monitoring unit automatically switches to a backup communication path to ensure continuous data transmission. In some cases, the system will also activate a local control mode, allowing it to perform basic tasks and maintain stable operation even without a network connection.

[0107] When anomalies such as data loss, delay, or communication interruption are detected, the anomaly monitoring unit immediately triggers the fault-tolerant retransmission mechanism. Specifically: When data is lost, the system requests the lost data packets to be retransmitted to ensure that the system receives complete data. When data is delayed, the system optimizes the transmission path or enables a caching mechanism to mitigate the delay, ensuring that data can be delivered to the target module as soon as possible. When communication is interrupted, the system attempts to re-establish the communication connection or automatically switches to a backup communication link to ensure the continuity of the data flow.

[0108] These operations ensure that the system can respond quickly to communication anomalies, restore data transmission, and continue to operate stably, thereby guaranteeing the high availability and real-time performance of the centrifuge intelligent control system.

[0109] Through the above steps, the anomaly monitoring unit can effectively identify and repair data loss, delay, or interruption problems during communication, ensuring the data integrity and stable operation of the centrifuge system.

[0110] The following is a practical example of data loss. During the operation of a high-speed centrifuge, the sensing module collects data in real time through various sensors, including speed, temperature, and vibration. All collected data is transmitted via network to the dynamic modeling module for analysis. If data loss occurs during transmission, the dynamic modeling module may be unable to accurately obtain necessary operating status parameters, thus affecting the performance of the entire control system.

[0111] Data loss occurs as follows: Assume the system collects data every second, and each data packet contains a unique sequence number or timestamp (e.g., the first data packet has a sequence number of 1, the second data packet has a sequence number of 2, and so on). Under normal circumstances, the sensing module will send data packets to the dynamic modeling module in sequence.

[0112] However, due to network fluctuations or other reasons, some data packets may fail to reach the target module. For example, at timestamp T2, the data packet with sequence number 2 may not be received by the dynamic modeling module due to network latency or packet loss.

[0113] Regarding data loss detection, the dynamic modeling module performs sequence number checks based on the sequence numbers in the received data packets. Normally, the sequence numbers of received data packets should increase sequentially. For example, the sequence number of a data packet received in the first second is 1, in the second second it's 2, and so on. If the dynamic modeling module receives data packets with sequence numbers 1, 3, and 4, it will immediately detect the loss of the data packet with sequence number 2 (i.e., no data was received at time T2). Time synchronization detection: In addition to sequence number checks, the dynamic modeling module also detects data loss based on timestamps. If the timestamp sequence of data packets is discontinuous, the system can infer data packet loss. For example, if no data packets are received between timestamps T1 and T3, this indicates that a data packet was lost at time T2.

[0114] Regarding how to handle data loss and trigger a retransmission mechanism: When the anomaly monitoring unit detects data loss (for example, detecting the loss of data packet with sequence number 2), it immediately sends a request to the sensing module to retransmit the lost data packet. Specifically, the anomaly monitoring unit notifies the sensing module to retransmit the data packet with sequence number 2 by sending a "retransmission request" message. Retransmission: After receiving the retransmission request, the sensing module checks its data cache. If the lost data packet still exists (data is usually retained in the cache for a certain period), it retransmits the lost data packet with sequence number 2. Resumption of normal operation: After receiving the retransmitted data packet, the dynamic modeling module continues to process the data, ensuring the integrity of data transmission. At this point, the dynamic modeling module resumes normal operation, continuing model updates and control command generation.

[0115] In this example, the anomaly monitoring unit successfully detected the data packet loss by checking the sequence number and timestamp, and triggered a retransmission mechanism to ensure that the lost data could be recovered in a timely manner. This mechanism effectively ensures that the centrifuge's intelligent control system can still operate stably in the event of data loss, avoiding control errors and system instability.

[0116] The content disclosed above is only a preferred and feasible embodiment of the present invention, and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent technical changes made based on the content of the present invention specification and drawings are included within the scope of protection of the present invention. Furthermore, the elements therein can be updated as technology develops.

Claims

1. An intelligent control system for a high-speed centrifuge, characterized in that, The system includes: The sensing module is used to collect the operating status parameters of the centrifuge in real time, and to synchronously sample the operating status parameters and generate multimodal operating data; The dynamic modeling module establishes a dynamic physical model of the centrifuge chamber based on multimodal operating data and performs real-time correction on the parameters of the dynamic physical model to generate a set of characteristic parameters that reflect the overall dynamic state of the centrifuge. The control execution module adjusts the speed and running time of the centrifuge drive motor in real time based on the characteristic parameter set, operating status parameters and centrifuge control curve output by the automation control module, to achieve closed-loop control. The automated control module includes a recurrent neural network structure that performs self-learning based on the centrifuge's historical operating data, feature parameter set, and sample feature data. It adaptively corrects the control strategy parameters according to the learning results to generate dynamically updated centrifugation control curves and sends the centrifugation control curves to the control execution module for real-time control. The sample characteristic data module is used to store sample characteristic data, which are characteristic parameters used to characterize the physical or chemical properties of the sample.

2. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The dynamic modeling module includes: The constraint modeling unit constructs a dynamic physical model based on multimodal operating data. The dynamic physical model is constructed using a cross-domain coupled constraint graph structure, in which the constraint graph structure is centered on energy flow nodes, mass flow nodes, and vibration mode nodes, and is used to characterize the interaction relationships between multiple physical fields within the centrifuge cavity. The parameter update unit updates and coordinates the parameters of the dynamic physical model in real time based on the node association weights in the constraint graph structure.

3. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The sensing module includes: The signal acquisition unit is used to acquire the operating status parameters of the centrifuge during operation. The synchronous sampling unit performs time-synchronized sampling of the raw signals of the operating status parameters to maintain the consistency of multiple types of sensor signals on the time scale. The multimodal fusion unit performs feature normalization and multidimensional combination on the synchronously sampled operating state parameters to generate multimodal operating data.

4. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The control execution module includes: The target analysis unit analyzes the target rotation speed and target running time based on the feature parameter set. The instruction synthesis unit generates initial speed instructions and initial running time instructions based on the target speed and target running time, using the centrifugal control curve output by the automatic control module as a feedforward reference. The feedback acquisition unit is used to acquire the actual speed signal and running time status information in the centrifuge's operating status parameters in real time, and compare the acquisition results with the target speed and target running time; The calculation unit is adjusted to calculate the speed deviation and time deviation based on the comparison results, and generate the corresponding correction amount. The execution control unit synthesizes the initial speed command, the initial running time command, and the correction amount, and outputs them to the centrifuge drive motor to achieve closed-loop control.

5. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The automated control module includes: The input preprocessing unit receives historical operating data of the centrifuge, a set of characteristic parameters, and sample characteristic data, and performs time alignment and feature normalization on the input data. The recursive learning unit uses a recursive neural network structure to extract temporal features and recognize patterns from the preprocessed input data to form a dynamic state vector that reflects the operating trend of the centrifuge. The strategy correction unit calculates the correction amount of the control strategy parameters based on the dynamic state vector and adaptively updates the control strategy parameters. The curve generation unit generates a centrifugal control curve based on the updated control strategy parameters and sends the centrifugal control curve to the control execution module to perform real-time control.

6. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The sample feature data module includes: Data storage unit, used to store sample characteristic data; The feature indexing unit establishes a multi-dimensional index structure according to sample type, experimental conditions, and target separation accuracy to enable the retrieval and retrieval of sample feature data; The association mapping unit establishes a one-to-one association mapping relationship between sample characteristic data and centrifuge historical operation data and characteristic parameter set based on the experiment number information, so as to achieve traceability; During the startup phase of the automation control module, the data retrieval unit extracts the corresponding sample feature data from the data storage unit based on the input sample identification information or experimental conditions, and sends the sample feature data to the automation control module for calculation.

7. The intelligent control system for a high-speed centrifuge as described in claim 1, characterized in that, The system also includes an intelligent health management module, which constructs an operational health model of the centrifuge based on a set of feature parameters and operating status parameters, and identifies potential abnormal states of the centrifuge by analyzing vibration signals, temperature change trends and energy consumption characteristics during multi-cycle operation. When an abnormal trend or operational deviation is detected, the automatic control module is triggered to execute an adaptive learning process to correct the control strategy parameters and dynamically adjust the centrifugal control curve, thereby achieving predictive maintenance of the system's health status and long-term stable operation.

8. The intelligent control system for a high-speed centrifuge as described in claim 7, characterized in that, After the automatic control module performs adaptive learning, the intelligent health management module tracks the centrifuge's operating status in real time and calculates the operating stability index based on the corrected control strategy parameters. When the operating stability index fails to meet the preset threshold conditions, the automatic control module is triggered to perform the adaptive learning process again.

Images