A centrifugal detection system

Abstract

This application relates to the field of centrifugation detection technology and provides a centrifugation detection system. The system acquires spectral data, image data, and pressure data through a detection module, extracts spectral features, layer edge sharpness, and pressure change features, and fuses these features using an attention mechanism to obtain centrifugation fusion features, thereby analyzing an accurate centrifugation status score. The control module combines this centrifugation status score with vibration amplitude and torque changes during simulated centrifugation to determine whether the centrifugation status is abnormal. This multi-source data detection and multi-factor comprehensive judgment method can comprehensively and accurately assess the centrifugation status, avoiding the limitations of single data or single-factor judgment. It can promptly and accurately detect potential physical and biological problems during centrifugation, providing a reliable basis for subsequent adjustments and control.

Description

Technical Field

[0001] This application relates to the field of centrifugation detection technology, and in particular to a centrifugation detection system. Background Technology

[0002] In the fields of medical treatment and scientific research, the analysis and testing of samples is crucial. Sample quality directly affects the effectiveness of subsequent research and treatment. Existing centrifugation testing systems often rely on only a single type of data for analysis, which cannot comprehensively reflect the state of the sample during the simulated centrifugation process. Furthermore, when judging whether the centrifugation state is abnormal, they usually only consider a single factor and ignore the influence of other important factors on the centrifugation state, resulting in inaccurate anomaly judgments and the inability to detect potential problems in a timely manner. This affects the normal operation of the simulated centrifugation testing process and the sample quality. Moreover, the lack of targeted adjustments leads to a further decline in sample quality, and there is no solution on how to achieve centrifugation testing that improves the target sample quality. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this application provides a centrifugation detection system, comprising: a detection module and a control module;

[0004] The detection module is used to acquire spectral data, image data, and pressure data, extract spectral features, layer edge sharpness, and pressure change features, and fuse these features through an attention mechanism to obtain centrifugation fusion features. The centrifugation fusion features are then analyzed to obtain a centrifugation status score.

[0005] The control module is used to monitor the vibration amplitude and torque changes during the simulated centrifugation process, and to determine whether the centrifugation status is abnormal based on the centrifugation status score. If the centrifugation status is abnormal, the centrifugation parameters are adjusted, and it is determined whether to trigger sample distribution adjustment and inertia deviation compensation.

[0006] As an optional implementation, the logic for acquiring the spectral data includes:

[0007] Integrated dual-wavelength laser, with time-division alternating triggering of the dual-wavelength laser;

[0008] Spectral data acquisition is initiated when the target sample is within the detection window;

[0009] The spectral characteristics of the dual-wavelength laser channels in the spectral data were analyzed separately.

[0010] As an optional implementation, image data is acquired through an image sensor, and pressure data is acquired through a pressure sensor. The logic for extracting the layered edge sharpness includes:

[0011] Edge detection algorithms are used to identify edge pixels in image data, and the image data is then segmented into multiple hierarchical regions based on these edge pixels.

[0012] Configure a local window, and calculate the gradient magnitude and gradient direction of the edge pixels of each layered region within the local window;

[0013] The layered edge sharpness is obtained by comprehensively analyzing the gradient magnitude, gradient direction, and local contrast of edge pixels.

[0014] As an optional implementation, the segmentation logic of the hierarchical region includes:

[0015] An adjacency graph is generated based on the edge pixels of the image data, where each node represents an edge pixel and the edges represent the connections between adjacent edge pixels.

[0016] Calculate the similarity between nodes and configure the similarity threshold;

[0017] Clustering algorithms are used to group nodes with similarity greater than a similarity threshold into the same hierarchical region.

[0018] As an optional implementation, the logic for feature fusion includes:

[0019] Construct a spatiotemporal matrix of spectral features, layer edge sharpness, and pressure change features, and calculate spatial attention weights based on the spatiotemporal matrix;

[0020] Calculate the temporal attention weights based on the spatiotemporal matrix;

[0021] The spatial attention weights and temporal attention weights are multiplied by a Hadamard product to obtain the centrifugal fusion features.

[0022] As an optional implementation, the logic for determining whether the centrifugation state is abnormal includes:

[0023] The physical anomaly state is determined based on the changes in vibration amplitude and torque during the simulated centrifugation test.

[0024] Configure a scoring threshold, and compare the centrifugation status score with the scoring threshold to obtain the abnormal biological status;

[0025] The centrifugation status is judged based on a combination of physical and biological abnormalities.

[0026] As an optional implementation, the judgment logic for triggering sample distribution adjustment includes:

[0027] Real-time monitoring of pressure data is used to determine the adjustment trigger conditions for sample distribution adjustment;

[0028] An annular pressure gradient map is generated based on pressure data to mark pressure distribution areas, including areas of sudden pressure increase and areas of sudden pressure decrease, and the direction of imbalance is determined.

[0029] The airflow type is determined based on the pressure distribution area, and the airflow intensity is determined based on the direction of imbalance.

[0030] The annular pressure gradient map is updated in real time to verify the adjustment effect.

[0031] As an optional implementation, the direction of imbalance is determined based on the distribution pattern of the pressure surge area and the pressure drop area. If the pressure surge area and the pressure drop area are symmetrically distributed, it is determined to be a static imbalance.

[0032] If the regions of sudden pressure increase and sudden pressure decrease are asymmetrically distributed, it is judged as a dynamic imbalance.

[0033] As an optional implementation, the judgment logic for triggering inertia deviation compensation includes:

[0034] Real-time monitoring of layer edge sharpness to determine the compensation trigger conditions for inertia deviation compensation;

[0035] Monitor torque changes to determine the authenticity of compensation trigger conditions;

[0036] If the compensation trigger condition is true, the rotational speed is adjusted in segments to compensate for the inertia deviation.

[0037] Identify resonance states during segmented speed adjustment processes to implement avoidance mechanisms;

[0038] Continuously monitor changes in the sharpness of the layered edges to verify the compensation effect.

[0039] As an optional implementation, the sub-logic for segmented speed adjustment includes:

[0040] When the compensation trigger condition is true, reduce the speed;

[0041] During the speed reduction process, the decrease in torque is monitored. If the decrease in torque exceeds the threshold, the speed is restored to the target speed.

[0042] If the decrease in torque is less than or equal to the amplitude threshold, a resonance state is identified, and an avoidance mechanism is executed.

[0043] Compared with existing technologies, the beneficial effects of this application are as follows: Spectral data, image data, and pressure data are acquired through the detection module, and spectral features, layer edge sharpness, and pressure change features are extracted. These features are then fused through an attention mechanism to obtain centrifugation fusion features, thereby analyzing an accurate centrifugation status score. The control module combines this centrifugation status score with vibration amplitude and torque changes during simulated centrifugation detection to determine whether the centrifugation status is abnormal. This multi-source data detection and multi-factor comprehensive judgment method can comprehensively and accurately assess the centrifugation status, avoiding the limitations of single data or single-factor judgment. It can timely and accurately detect potential physical and biological problems during centrifugation, providing a reliable basis for subsequent adjustments and control.

[0044] The detection module can acquire spectral data, image data, and pressure data, and extract spectral features, layer edge sharpness, and pressure change features from them. This provides a comprehensive and detailed understanding of the sample's state changes during centrifugation, offering rich information for accurate assessment of the centrifugation status. Simultaneously, by employing an attention mechanism, it fuses spectral features, layer edge sharpness, and pressure change features to obtain centrifugation fusion features. This fully leverages the advantages of multi-source features, highlighting key information and analyzing it to obtain a centrifugation status score. It comprehensively considers the influence of multiple factors on the centrifugation status, helping operators to promptly understand the operating status of the centrifugation detection system and the centrifugation effect of the samples, providing a strong basis for subsequent operations and decisions.

[0045] The control module monitors vibration amplitude and torque changes during the simulated centrifugation process and combines this with a centrifugation status score to determine if any abnormalities have occurred. Considering both physical and biological factors, it can more comprehensively and accurately detect anomalies during centrifugation, preventing centrifuge malfunctions and sample quality degradation. When an abnormality is detected, the control module determines whether to trigger sample distribution adjustment and inertia deviation compensation. By monitoring pressure data and layer edge sharpness in real time, it can promptly detect uneven sample distribution and inertia deviation, and take corresponding adjustment and compensation measures. The sample distribution adjustment mechanism utilizes air pressure gradients to regulate sample distribution, making it more uniform. The inertia deviation compensation mechanism compensates for inertia deviation by adjusting rotation speed in segments, preventing resonance and ensuring stable operation of the centrifugation system and high-quality sample preparation. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0047] Figure 1 This is a system flowchart of a centrifugation detection system provided in an embodiment of this application;

[0048] Figure 2 This application provides a segmentation logic diagram of a layered region in a centrifugal detection system.

[0049] Figure 3 This is a feature fusion logic diagram of a centrifugation detection system provided in an embodiment of this application;

[0050] Figure 4 This is a logic diagram for determining the segmented speed adjustment of a centrifugal detection system provided in an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application more apparent and understandable, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0052] like Figure 1 The diagram shown is a system flowchart of a centrifugal detection system provided in this application embodiment. The centrifugal detection system includes a detection module and a control module.

[0053] In this embodiment, the target sample is exemplified as a PRP sample.

[0054] The detection module is used to acquire spectral data, image data, and pressure data, extract spectral features, layer edge sharpness, and pressure change features, and use an attention mechanism to fuse these features to obtain centrifugation fusion features. The centrifugation fusion features are then analyzed to obtain a centrifugation status score.

[0055] Specifically, the logic for acquiring spectral data includes:

[0056] Integrated dual-wavelength laser, with time-division alternating triggering of the dual-wavelength laser;

[0057] Spectral data acquisition is initiated when the PRP sample is in the detection window;

[0058] The spectral characteristics of the dual-wavelength laser channels in the spectral data were analyzed separately.

[0059] To better detect PRP samples, it is necessary to acquire spectral data, image data, and pressure data. Image data is obtained by capturing images of the layered interface inside the PRP sample container using an image sensor (such as a camera). Pressure data is obtained by monitoring pressure changes inside the PRP sample container using a pressure sensor. This section focuses on describing the logic for acquiring spectral data; the logic for acquiring other data will not be elaborated upon.

[0060] PRP samples are complex, and different components have different absorption characteristics for different wavelengths of light. Dual-wavelength lasers can cover more component information, and time-sharing alternating triggering can avoid interference and improve data accuracy. Two specific wavelength laser generators sensitive to key components of the PRP sample are selected, such as wavelengths chosen to target the absorption peaks of characteristic substances in platelets and plasma. Assuming the wavelengths are respectively... and During the spectral data acquisition phase, dual-wavelength lasers are triggered alternately in a time-division manner according to a preset triggering timing logic. For example, during the first half of a preset rotation cycle, the trigger wavelength is... The laser generator; in the latter half, the trigger wavelength is The laser generator is designed to avoid interference between different wavelengths of laser light and achieve time-division alternation, thereby obtaining purer and more accurate spectral data that reflects the composition of PRP samples. This provides a high-quality data foundation for subsequent analysis of spectral features, and accurate spectral data makes the extracted spectral features more representative.

[0061] To ensure that only spectral data of the PRP sample inside the PRP sample container is acquired, avoiding the acquisition of invalid background information and improving the effectiveness of the spectral data, the detection window can be set at a 30° angle to the axis of the PRP sample container. A photoelectric sensor is placed at the detection window location. The position sensor and the photoelectric sensor work together; when the PRP sample enters the detection window, it blocks the light, causing the photoelectric sensor to generate a level change signal. By analyzing the signals from the position sensor and the photoelectric sensor, the moment when the PRP sample is within the detection window is determined, and spectral data acquisition is initiated. This reduces the acquisition of invalid data, improves acquisition efficiency, and enhances data relevance.

[0062] Spectral data from different wavelength laser channels reflect different aspects of PRP samples. Analyzing them separately helps to comprehensively understand the composition and state of PRP samples. By processing the spectral data of each wavelength channel through Fourier transform, the time-domain spectral data is converted into frequency-domain data, thereby extracting spectral feature parameters such as characteristic peak positions, intensities, and full width at half maximum (FWHM). For complex spectral data, wavelet analysis is combined to decompose the spectral data into different scale spaces, further mining the detailed information of the spectral data, enhancing the ability to identify weak characteristic peaks, and thus obtaining detailed sample composition feature information, providing rich spectral feature dimensions for subsequent feature fusion.

[0063] After acquiring spectral data, image data, and pressure data, it is necessary to extract the features of the data, including spectral features, layer edge sharpness, and pressure change features. Here, we will focus on describing the extraction logic of layer edge sharpness; the extraction logic of other features will not be elaborated on.

[0064] Specifically, the logic for extracting the edge sharpness of layers includes:

[0065] Edge algorithms are used to identify edge pixels in image data, and the image data is then segmented into several hierarchical regions based on these edge pixels.

[0066] Configure a local window, and calculate the gradient magnitude and gradient direction of the edge pixels of each layered region within the local window;

[0067] The layered edge sharpness is obtained by comprehensively analyzing the gradient magnitude, gradient direction, and local contrast of edge pixels.

[0068] After centrifugation, PRP samples are stratified, and edge features can reflect the stratification. Segmenting the stratified regions facilitates targeted analysis of the characteristics of each layer. The Canny edge detection algorithm is used to identify edge pixels in the image data. First, Gaussian filtering is applied to the acquired image data to remove noise interference. Then, the gradient magnitude and direction of each pixel in the image data are calculated. Next, non-maximum suppression is used to refine the edges. Finally, double threshold detection and edge tracking are used to determine the final edge pixels. Based on the spatial distribution and connectivity of edge pixels, the image data is segmented into different stratified regions using an adjacency graph, thus clearly delineating the stratified structure of the samples and providing accurate region segmentation for subsequent calculation of stratified edge sharpness.

[0069] Local analysis of edge pixel characteristics can more meticulously reflect the detailed information of layered edges, providing basic data for comprehensive analysis. Based on the image data resolution and the size of the layered regions, appropriately sized square or rectangular local windows are set. For example, for an image with a resolution of 1024×768, if the layered region size is between tens and hundreds of pixels, the window size can be set to 3×3, 5×5, or 7×7. Within each local window, the Sobel operator is used to perform convolution operations on the edge pixels to obtain the gradient magnitude and gradient direction. The Sobel operator uses two 3×3 convolution kernels to perform horizontal and vertical convolutions on the image data, and then calculates the gradient magnitude and gradient direction according to relevant formulas. This allows for the acquisition of detailed gradient information of the edge pixels in each layered region, reflecting the changing trends and intensity of the edge pixels.

[0070] A single feature cannot fully reflect the clarity and stability of the layered edges. Combining multiple features can more accurately quantify the sharpness of the layered edges. Therefore, different weights are assigned to gradient magnitude, gradient direction, and local contrast. The sharpness of the layered edges is calculated using a weighted summation formula. The weights are optimized and determined through machine learning algorithms, such as support vector regression. The gradient magnitude, gradient direction, and local contrast of a large number of samples with known layered quality are used as input features, and the manually labeled layered edge sharpness is used as the output label. The support vector regression model is trained to obtain the optimal weight allocation, thereby obtaining a quantitative indicator that can accurately reflect the layered quality of PRP samples, providing an important basis for judging the centrifugal state.

[0071] Furthermore, such as Figure 2 As shown, the segmentation logic of the hierarchical region includes:

[0072] An adjacency graph is generated based on the edge pixels of the image data, where each node represents an edge pixel and the edges represent the connections between adjacent edge pixels.

[0073] Calculate the similarity between nodes and configure the similarity threshold;

[0074] Clustering algorithms are used to group nodes with similarity greater than a similarity threshold into the same hierarchical region.

[0075] Transforming the spatial relationships of edge pixels in image data into a graph structure facilitates region segmentation using graph theory algorithms. This involves traversing the edge pixels of the image data, creating a node for each edge pixel, determining adjacent edge pixels based on their spatial relationships, and establishing edges between nodes to construct an adjacency graph. For example, this approach is used for edge pixels in a two-dimensional image. Its adjacent pixels include , , and The number of neighboring pixels of the boundary pixels will be reduced. Edges are established between the corresponding nodes, and the adjacency graph is stored using an adjacency matrix or adjacency list to facilitate subsequent graph operations. This provides an intuitive graph structure to represent the relationship between edge pixels and provides a data structure foundation for subsequent clustering analysis.

[0076] Similarity measures the degree of similarity between nodes to determine which nodes should belong to the same region. A similarity threshold controls the tightness of clustering. The similarity between nodes is calculated using Euclidean distance based on features such as grayscale value, gradient magnitude, and gradient direction. For example, for two nodes A and B, their feature vectors are... and , This represents the grayscale value of node A. This represents the gradient magnitude of node A. Indicates the gradient direction of node A. This represents the grayscale value of node B. This represents the gradient magnitude of node B. Let represent the gradient direction of node B. Then, the formula for calculating the similarity d between node A and node B is: Cross-validation is used to determine an appropriate similarity threshold, thereby providing a quantitative basis for clustering and deciding which nodes will be merged into the same hierarchical region.

[0077] To achieve automatic grouping of edge pixels and form different hierarchical regions, the K-Means clustering algorithm is selected. Based on the calculated similarity between nodes and a set similarity threshold, nodes in the adjacency graph are clustered into different hierarchical regions. For example, K initial cluster centers are randomly selected first, and then the similarity between each node and each cluster center is calculated. The similarity can be calculated using Euclidean distance, and the node is assigned to the cluster containing the cluster center with the highest similarity. Then, the cluster center of each cluster is recalculated, and the above process is repeated until the cluster center no longer changes or the preset number of iterations is reached. This accurately delineates the hierarchical regions of the PRP sample and improves the accuracy of the hierarchical edge sharpness calculation.

[0078] Specifically, such as Figure 3 As shown, the logic for feature fusion includes:

[0079] Construct a spatiotemporal matrix of spectral features, layer edge sharpness, and pressure change features, and calculate spatial attention weights based on the spatiotemporal matrix;

[0080] Calculate the temporal attention weights based on the spatiotemporal matrix;

[0081] The spatial attention weights and temporal attention weights are multiplied by a Hadamard product to obtain the centrifugal fusion features.

[0082] Different features have varying impacts on centrifugal states in both spatial and temporal dimensions. Constructing a spatiotemporal matrix can uniformly represent multi-source features, and calculating spatial attention weights can highlight spatially important features. By arranging spectral features, layer edge sharpness, and pressure change features in temporal order, a three-dimensional spatiotemporal matrix can be constructed. For example, assuming the spectral features have a dimension of... The dimension of the layer edge sharpness is The dimensions of pressure change characteristics are The time step is Then the dimension of the spacetime matrix is The spatiotemporal matrix is ​​convolved using a convolutional neural network, and spatial attention weights are calculated using the Softmax function.

[0083] Specifically, multiple convolutional layers are designed, each containing multiple convolutional kernels. By sliding the convolutional kernels on the spatiotemporal matrix, spatial features at different scales are extracted. Then, the output of the convolutional layer is passed through a fully connected layer and normalized by the Softmax function to obtain a matrix of spatial attention weights, thereby highlighting the importance of different features in the spatial dimension and making the fused features more focused on key spatial information.

[0084] Considering the changing trends of features over time, calculating temporal attention weights can emphasize feature information at important time points. The spatiotemporal matrix is ​​processed through a Long Short-Term Memory (LSTM) network. Through gating mechanisms and weight calculations, temporal attention weights are obtained. The LSTM unit contains input gates, forget gates, and output gates. These gates control the selective memorization and updating of information in the time series. The spatiotemporal matrix is ​​expanded along the time dimension and sequentially input into the LSTM network. The LSTM network calculates the weights for each time step based on the input information and its internal state. Finally, the matrix of temporal attention weights is obtained through normalization using the Softmax function, thereby capturing the changing trends of features over time and highlighting features at key time points.

[0085] By integrating attention weights across spatial and temporal dimensions, effective fusion of multi-source features is achieved, resulting in comprehensive features that better reflect the centrifugal state. The centrifugal fusion feature matrix is ​​obtained by multiplying corresponding elements of the spatial attention weight matrix and the temporal attention weight matrix. For example, the spatial attention weight matrix is... The matrix of time attention weights is The matrix of centrifugal fusion features is ,but ,in and These represent the row and column indices of the matrix, respectively, thus effectively integrating multi-source features to generate comprehensive features that fully reflect the centrifugal state.

[0086] After normalizing the centrifugation fusion features, they are input into a multilayer perceptron for classification and regression analysis. The multilayer perceptron consists of an input layer, multiple hidden layers, and an output layer. The number of nodes in the input layer is the same as the matrix dimension of the centrifugation fusion features. The hidden layers can be set to 2-3 layers, and the number of neurons in each layer is determined experimentally, for example, selected from values ​​such as 64, 128, and 256. The number of nodes in the output layer is consistent with the number of centrifugation state indicators that need to be quantified. For example, if the quantification indicators of centrifugation state are stratification uniformity, sample separation efficiency, and centrifugation stability, then the number of nodes in the output layer is 3. Then, a weight is assigned to stratification uniformity, sample separation efficiency, and centrifugation stability, and the centrifugation state score is calculated by weighted summation.

[0087] The control module is used to monitor the vibration amplitude and torque changes during the simulated centrifugation process, and to determine whether the centrifugation status is abnormal based on the centrifugation status score. If the centrifugation status is abnormal, the centrifugation parameters are adjusted, and it is determined whether to trigger sample distribution adjustment and inertia deviation compensation.

[0088] Centrifugation parameters include rotation speed, centrifugation time, and temperature;

[0089] Specifically, the logic for determining whether an abnormality has occurred in the centrifugation state includes:

[0090] The physical anomaly state is determined based on the changes in vibration amplitude and torque during the simulated centrifugation test.

[0091] Configure a scoring threshold, and compare the centrifugation status score with the scoring threshold to obtain the abnormal biological status;

[0092] The centrifugation status is judged based on a combination of physical and biological abnormalities.

[0093] During simulated centrifugation, the vibration amplitude and torque remain within a relatively stable range. When physical parameters become abnormal, such as uneven sample distribution, these abnormalities directly lead to abnormal fluctuations in vibration amplitude and torque. Monitoring these two key parameters allows for timely detection of physical anomalies during simulated centrifugation, providing a basis for subsequent appropriate measures. Specifically, an accelerometer monitors vibration amplitude, and a torque sensor monitors torque changes, acquiring the vibration amplitude and torque variations within the simulated system. The analog signals output from the accelerometer and torque sensor are amplified and filtered, and then converted into digital signals. Including vibration and torque signals, the vibration signal is analyzed by Fourier transform algorithm to extract the vibration amplitude as the vibration amplitude; at the same time, the torque signal is analyzed in the time domain to calculate the standard deviation of the torque as the torque change. The vibration amplitude and torque change are compared with preset amplitude thresholds and change thresholds, respectively. If the vibration amplitude is greater than the preset amplitude threshold or the torque change is greater than the preset change threshold, it is judged as a physical abnormality. This allows for real-time and accurate capture of abnormal changes in the physical operation of the simulated centrifugation process, providing a reliable means for timely detection of potential faults and effectively ensuring the stability of the simulated environment and the reliability of the detection results.

[0094] Centrifugation status score is obtained by integrating multi-source information such as spectral data, image data, and pressure data. It can reflect the biological characteristics and status of the sample during centrifugation. By setting a reasonable scoring threshold and comparing it with the actual centrifugation status score, it is possible to determine whether the centrifugation effect of the sample is normal from a biological perspective, providing supplementary information for a comprehensive assessment of centrifugation status. By statistically analyzing historical centrifugation status scores, a suitable scoring threshold is determined. If the centrifugation status score is less than the scoring threshold, it is judged as an abnormal biological state. This quantitative assessment of centrifugation status from a biological perspective provides a scientific basis for judging sample quality and centrifugation effect, and improves the ability to identify abnormal situations in the centrifugation process.

[0095] Neither physical nor biological anomalies alone can comprehensively and accurately assess the centrifugation status. Physical anomalies affect the separation effect of samples, while biological anomalies directly reflect the state of the samples themselves. By comprehensively considering both physical and biological anomalies, we can gain a more comprehensive understanding of the operation and centrifugation effect of the simulated centrifugation process, avoiding misjudgments caused by a single factor. Only when both physical and biological anomalies are present can the centrifugation status be determined to be abnormal, thereby improving the accuracy and reliability of the judgment of centrifugation status anomalies. This allows for a more comprehensive discovery of potential problems in the simulated centrifugation process and provides strong support for timely measures.

[0096] Rotation speed is one of the key parameters affecting centrifugation results. When abnormalities occur during centrifugation, adjusting the rotation speed appropriately can change the centrifugal force on the sample, thereby improving the separation effect and restoring the simulated centrifugation process to normal operation. If the sample stratification is unclear due to excessive rotation speed, the rotation speed should be reduced; if the separation time is too long due to excessive rotation speed, the rotation speed should be increased. This allows for flexible and precise adjustment of the centrifugation rotation speed according to the actual situation, effectively improving the centrifugation effect, solving the problem of abnormal centrifugation due to improper rotation speed, and improving the efficiency of simulated detection and the quality of sample simulation analysis. The simulated adjustment of centrifugation rotation speed is achieved by changing the frequency of the input voltage of the simulated motor, thereby precisely regulating the centrifugation rotation speed.

[0097] The length of centrifugation time directly affects the degree of sample separation. When the centrifugation is abnormal, if it is due to insufficient or excessive centrifugation time leading to inadequate or excessive sample separation, adjusting the centrifugation time can give the sample enough time to separate, thereby improving the sample quality, ensuring that the sample completes the centrifugation process within an appropriate time, optimizing the sample separation effect, avoiding sample quality problems caused by improper centrifugation time, and improving the efficiency and reliability of simulated centrifugation detection.

[0098] Temperature has a significant impact on the activity and physicochemical properties of samples. During centrifugation, excessively high or low temperatures can lead to a decrease in sample quality. Adjusting the temperature can provide a suitable environment for the samples, ensuring their activity and stability, thereby improving centrifugation results, reducing poor centrifugation performance due to temperature issues, and enhancing the reliability of simulated centrifugation detection and the usability of sample simulation analysis.

[0099] Specifically, the logic for triggering sample distribution adjustments includes:

[0100] Real-time monitoring of pressure data is used to determine the adjustment trigger conditions for sample distribution adjustment;

[0101] An annular pressure gradient map is generated based on pressure data to mark pressure distribution areas, including areas of sudden pressure increase and areas of sudden pressure decrease, and the direction of imbalance is determined.

[0102] The airflow type is determined based on the pressure distribution area, and the airflow intensity is determined based on the direction of imbalance.

[0103] The annular pressure gradient map is updated in real time to verify the adjustment effect.

[0104] After adjusting the centrifugation parameters, pressure data is monitored according to the sampling cycle to promptly identify sample distribution problems. This provides an accurate basis for triggering sample distribution adjustments, ensuring stable operation of the simulated centrifugation test and sample quality. The standard deviation of the pressure data is calculated, and pressure changes over time are analyzed. A pressure change threshold is set. When the monitored pressure change exceeds the pressure change threshold, the adjustment trigger condition is met. This allows for real-time and accurate monitoring of pressure changes in the simulated centrifugation environment, timely detection of uneven sample distribution, and a reliable trigger signal for subsequent sample distribution adjustment operations. This improves the operational stability and sample quality of the simulated centrifugation test.

[0105] The annular pressure gradient map can intuitively display the pressure distribution in a simulated centrifugation environment. By marking pressure distribution areas and determining the direction of imbalance, it can more accurately understand the specific situation of uneven sample distribution, providing a visual basis for subsequent sample distribution adjustment. The acquired discrete pressure data is processed by converting discrete data points into continuous pressure distribution data through bilinear interpolation. Then, this pressure distribution data is mapped into an annular coordinate system to generate the annular pressure gradient map. The pressure distribution area is divided into pressure surge areas and pressure drop areas. For example, the pressure surge area refers to the area where the pressure change is greater than 1.5 times the pressure change threshold for two consecutive sampling periods, while the pressure drop area refers to the area where the pressure change is less than 0.5 times the pressure change threshold for more than two pressure sensors. At the same time, different colors or markers are used in the annular pressure gradient map to distinguish these areas for intuitive observation of the pressure distribution.

[0106] Furthermore, the direction of imbalance is determined based on the distribution pattern of the pressure surge area and the pressure drop area. If the pressure surge area and the pressure drop area are symmetrically distributed, it is judged as static imbalance.

[0107] If the regions of sudden pressure increase and sudden pressure decrease are asymmetrically distributed, it is judged as a dynamic imbalance.

[0108] The direction of imbalance is determined by the distribution pattern of the pressure surge and pressure drop regions. If the pressure surge and pressure drop regions are symmetrically distributed (180°), it is judged as static imbalance; if the pressure surge and pressure drop regions are asymmetrically distributed, such as 120°, it is judged as dynamic imbalance. This intuitively presents the pressure distribution and sample imbalance in the simulated centrifugal environment, providing a clear and visual basis for determining the airflow type and intensity, and helping to improve the accuracy and efficiency of sample distribution adjustment.

[0109] Different pressure distribution areas and imbalance directions require different airflow types and intensities to adjust sample distribution. Determining the appropriate airflow type based on the actual situation can more effectively utilize pressure gradients to regulate sample distribution and improve the problem of uneven sample distribution. For areas with sudden pressure increases, negative pressure airflow is used for adsorption to reduce the pressure in these areas. For areas with sudden pressure drops, positive pressure airflow is used for filling to increase the pressure in these areas. By controlling the direction and flow rate of the airflow, a suitable pressure gradient can be formed to promote the redistribution of samples.

[0110] The airflow intensity is determined based on the direction of imbalance and the degree of pressure change. For example, in cases of dynamic imbalance with large pressure changes, the airflow intensity is appropriately increased; in cases of static imbalance with small pressure changes, the airflow intensity is decreased. This can be achieved by generating airflow of the appropriate type and intensity through a simulated airflow device. The airflow rate is controlled by adjusting the opening of the simulated valve in the simulated airflow device to ensure that the airflow intensity meets the requirements. This targeted adjustment of the airflow type and intensity can more effectively utilize the simulated air pressure gradient to regulate the sample distribution, making the sample distribution more uniform and improving the operational stability and sample separation effect of simulated centrifugation detection.

[0111] During sample distribution adjustment, real-time updates to the annular pressure gradient map allow for intuitive observation of pressure distribution changes, thus verifying the adjustment effect and determining whether further adjustments to airflow type and intensity are needed. This ensures the sample distribution adjustment achieves the expected goals. Continuous acquisition of pressure sensor data updates the annular pressure gradient map in real time. Comparing the annular pressure gradient maps before and after adjustment allows for observation of changes in pressure distribution areas and imbalance directions. The uniformity index of pressure distribution, such as the variance of pressure data, is calculated to evaluate the adjustment effect. If the pressure distribution gradually becomes more uniform and the imbalance improves, the adjustment is considered effective. Conversely, further adjustments to airflow type or intensity are needed. This timely feedback on the sample distribution adjustment provides a basis for optimization, ensuring the adjustment achieves the expected goals and improving the operational efficiency and sample quality of simulated centrifugation testing.

[0112] Specifically, the logic for triggering inertia deviation compensation includes:

[0113] Real-time monitoring of layer edge sharpness to determine the compensation trigger conditions for inertia deviation compensation;

[0114] Monitor torque changes to determine the authenticity of compensation trigger conditions;

[0115] If the compensation trigger condition is true, the rotational speed is adjusted in segments to compensate for the inertia deviation.

[0116] Identify resonance states during segmented speed adjustment processes to implement avoidance mechanisms;

[0117] Continuously monitor changes in the sharpness of the layered edges to verify the compensation effect.

[0118] If PRP samples are unevenly distributed during centrifugation, their stratification will be abnormal, reflected in the sharpness of the stratification edges. By continuously monitoring the sharpness of the stratification edges, signals of abnormal sample distribution can be captured in a timely manner, thus providing a basis for triggering inertia deviation compensation. A sharpness threshold needs to be set, such as 85%. When the sharpness of the stratification edges is greater than the sharpness threshold, it is initially determined that the compensation triggering condition is met. This allows for real-time and accurate monitoring of the sample stratification status, timely detection of potential inertia deviation problems, and a reliable trigger signal for subsequent compensation operations, ensuring the stable operation of simulated centrifugation and the quality of the samples.

[0119] The abnormality in the sharpness of the layered edges is not solely caused by inertia deviation; other interfering factors exist. By comprehensively judging the simulated torque changes, it is possible to determine whether an inertia deviation truly exists. This improves the accuracy of compensation operations. The above judgment indicates that the torque change is greater than the change threshold, thus identifying an abnormal torque change. In other words, when both the layered edge sharpness and the torque change exceed the change threshold, the compensation trigger condition is true. This improves the accuracy of inertia deviation judgment, avoids unnecessary compensation operations due to misjudgment, saves energy and time, and ensures the stability of the simulated centrifugation test operation.

[0120] Furthermore, such as Figure 4 As shown, the sub-logic for segmented speed adjustment includes:

[0121] When the compensation trigger condition is true, reduce the speed;

[0122] During the speed reduction process, the decrease in torque is monitored. If the decrease in torque exceeds the threshold, the speed is restored to the target speed.

[0123] If the decrease in torque is less than or equal to the amplitude threshold, a resonance state is identified, and an avoidance mechanism is executed.

[0124] Inertia deviation can subject the simulated centrifugation process to significant unbalanced forces, causing instability in the centrifugation system. Reducing the rotational speed can mitigate the impact of these unbalanced forces on the simulated centrifugation process and facilitate observation of torque changes, providing a buffer for subsequent adjustments. By altering the frequency of the simulated motor's input voltage, its rotational speed can be reduced, thereby correspondingly lowering the centrifugation speed. During this reduction, the simulated motor's speed is monitored in real time to ensure a smooth decrease to the set value. This effectively reduces the impact of inertia deviation on the simulated centrifugation process, ensuring stable operation of the centrifugation system and the quality of sample simulation analysis. It also provides a basis for further adjustments based on torque changes.

[0125] During the reduction of the simulated centrifugal speed, the decrease in torque is monitored in real time. If the decrease in torque is significant, it indicates that the reduction in speed has effectively improved the inertia deviation problem. At this point, restoring the speed to the target speed ensures the working efficiency of the simulated centrifugal detection and avoids affecting the production schedule due to prolonged low-speed operation. The decrease in torque is calculated in real time, which is the difference between the current torque and the torque before the speed reduction, divided by the time taken to reduce the speed. An amplitude threshold is configured. When the calculated decrease in torque exceeds the amplitude threshold, the frequency of the input voltage of the simulated motor is gradually increased to restore the speed to the target speed. This solves the inertia deviation problem while ensuring the working efficiency of the simulated centrifugal detection and improving the stable operation and service life of the centrifugal detection system.

[0126] When the decrease in torque is small or insignificant, it indicates the presence of resonance. Resonance can cause severe vibration in the centrifuge detection system, further damaging it. Timely identification and avoidance of resonance can prevent damage and ensure the safe operation of the centrifuge detection system. By converting the vibration signal from the time domain to the frequency domain using Fourier transform and analyzing its frequency components, a resonance state is identified when the vibration frequency is close to the natural frequency of the simulated centrifuge detection system and the torque change is abnormal. Once resonance is identified, the speed of the simulated motor is quickly adjusted to avoid the resonance frequency through rapid acceleration or deceleration, such as increasing or decreasing the speed by a certain percentage (e.g., 5% to 10%) within a very short time. Subsequent verification and adaptive adjustments are then made based on data such as torque change and layer edge sharpness. This effectively prevents damage to the centrifuge detection system due to resonance, ensuring stable operation and service life, while also improving the inertia deviation compensation effect.

Claims

1. A centrifugal detection system, characterized in that, include: Detection module and control module; The detection module is used to acquire spectral data, image data, and pressure data, extract spectral features, layer edge sharpness, and pressure change features, and fuse these features through an attention mechanism to obtain centrifugation fusion features. The centrifugation fusion features are then analyzed to obtain a centrifugation status score. The logic for feature fusion includes: Construct a spatiotemporal matrix of spectral features, layer edge sharpness, and pressure change features, and calculate spatial attention weights based on the spatiotemporal matrix; Calculate the temporal attention weights based on the spatiotemporal matrix; The spatial attention weights and temporal attention weights are multiplied by a Hadamard product to obtain the centrifugal fusion features. The control module is used to monitor the vibration amplitude and torque changes during the simulated centrifugation process, and to determine whether the centrifugation status is abnormal based on the centrifugation status score. If the centrifugation status is abnormal, the centrifugation parameters are adjusted, and it is determined whether to trigger sample distribution adjustment and inertia deviation compensation.

2. The centrifugation detection system as described in claim 1, characterized in that, The logic for acquiring the spectral data includes: Integrated dual-wavelength laser, with time-division alternating triggering of the dual-wavelength laser; Spectral data acquisition is initiated when the target sample is within the detection window; The spectral characteristics of the dual-wavelength laser channels in the spectral data were analyzed separately.

3. The centrifugation detection system as described in claim 2, characterized in that, Image data is acquired through an image sensor, and pressure data is acquired through a pressure sensor. The logic for extracting the layered edge sharpness includes: Edge detection algorithms are used to identify edge pixels in image data, and the image data is then segmented into multiple hierarchical regions based on these edge pixels. Configure a local window, and calculate the gradient magnitude and gradient direction of the edge pixels of each layered region within the local window; The layered edge sharpness is obtained by comprehensively analyzing the gradient magnitude, gradient direction, and local contrast of edge pixels.

4. The centrifugation detection system as described in claim 3, characterized in that, The segmentation logic of the hierarchical region includes: An adjacency graph is generated based on the edge pixels of the image data, where each node represents an edge pixel and the edges represent the connections between adjacent edge pixels. Calculate the similarity between nodes and configure the similarity threshold; Clustering algorithms are used to group nodes with similarity greater than a similarity threshold into the same hierarchical region.

5. A centrifugation detection system as described in claim 4, characterized in that, The logic for determining whether the centrifugation state is abnormal includes: The physical anomaly state is determined based on the changes in vibration amplitude and torque during the simulated centrifugation test. Configure a scoring threshold, and compare the centrifugation status score with the scoring threshold to obtain the abnormal biological status; The centrifugation status is judged based on a combination of physical and biological abnormalities.

6. The centrifugation detection system as described in claim 5, characterized in that, The judgment logic for triggering sample distribution adjustment includes: Real-time monitoring of pressure data is used to determine the adjustment trigger conditions for sample distribution adjustment; An annular pressure gradient map is generated based on pressure data to mark pressure distribution areas, including areas of sudden pressure increase and areas of sudden pressure decrease, and the direction of imbalance is determined. The airflow type is determined based on the pressure distribution area, and the airflow intensity is determined based on the direction of imbalance. The annular pressure gradient map is updated in real time to verify the adjustment effect.

7. A centrifugation detection system as described in claim 6, characterized in that, The direction of imbalance is determined based on the distribution of the pressure surge area and the pressure drop area. If the pressure surge area and the pressure drop area are symmetrically distributed, it is judged as static imbalance. If the regions of sudden pressure increase and sudden pressure decrease are asymmetrically distributed, it is judged as a dynamic imbalance.

8. The centrifugation detection system as described in claim 7, characterized in that, The logic for triggering inertia deviation compensation includes: Real-time monitoring of layer edge sharpness to determine the compensation trigger conditions for inertia deviation compensation; Monitor torque changes to determine the authenticity of compensation trigger conditions; If the compensation trigger condition is true, the rotational speed is adjusted in segments to compensate for the inertia deviation. Identify resonance states during segmented speed adjustment processes to implement avoidance mechanisms; Continuously monitor changes in the sharpness of the layered edges to verify the compensation effect.

9. A centrifugation detection system as described in claim 8, characterized in that, The sub-logic for segmented speed adjustment includes: When the compensation trigger condition is true, reduce the speed; During the speed reduction process, the decrease in torque is monitored. If the decrease in torque exceeds the threshold, the speed is restored to the target speed. If the decrease in torque is less than or equal to the amplitude threshold, a resonance state is identified, and an avoidance mechanism is executed.

Images