Method and device for determining the end point of fluidized bed drying
By collecting and analyzing historical parameters during fluidized bed drying, a drying endpoint judgment model was established. The drying endpoint was controlled in real time using the Hotling T-square value, which solved the problem of difficulty in achieving immediate release and accurate moisture prediction during fluidized bed drying, and achieved immediate release at the drying endpoint and accurate moisture prediction results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU KANION PHARMA CO LTD
- Filing Date
- 2023-12-20
- Publication Date
- 2026-06-05
AI Technical Summary
In existing technologies, it is difficult to achieve immediate release at the drying endpoint during fluidized bed drying, and the accuracy of moisture prediction results is difficult to guarantee. Especially when the components of traditional Chinese medicine preparations are complex, it is difficult to ensure the consistency of drug quality by fixing the drying time and temperature standards.
By collecting key historical parameters from multiple batches during the fluidized bed drying process, performing dimensionality reduction and inter-group mean difference processing, a drying endpoint judgment model is established. The drying endpoint is controlled in real time using the Hotling T-square value. Combined with material spectra and process parameters, a secondary dimensionality reduction analysis is performed to obtain the drying endpoint judgment threshold, enabling immediate release.
This technology enables immediate release at the end of fluidized bed drying, ensuring the accuracy of moisture prediction results and improving the control precision of the drying process.
Smart Images
Figure CN117760193B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of fluidized bed technology, specifically to a method and apparatus for determining the endpoint of fluidized bed drying. Background Technology
[0002] Fluidized bed drying technology is a novel drying technique. The process involves placing bulk materials on an orifice plate and supplying gas from below. This causes the material particles to move on the gas distribution plate, remaining suspended in the airflow. This creates a mixed bottom layer of material particles and gas, where the particles are in full contact with the gas, facilitating heat and moisture transfer between the material and the gas. Fluidized bed drying technology is widely used in the pharmaceutical and chemical industries.
[0003] In fluidized bed drying, the drying process ends when all moisture is removed from the material. To accurately determine the drying endpoint, most traditional Chinese medicine (TCM) manufacturers currently use a fixed drying time, a fixed material temperature, and offline moisture detection as the release standard for fluidized bed drying. However, due to the complexity of TCM preparations and the influence of various factors on their quality attributes, using a fixed drying time and a fixed material temperature makes it difficult to ensure consistent drug quality. Therefore, offline moisture analysis is lagging and cannot achieve immediate release. On the other hand, online moisture analysis requires the establishment of a quantitative analysis model, which needs regular maintenance, and the accuracy of moisture prediction results is difficult to guarantee.
[0004] Therefore, there is an urgent need for a method to determine the endpoint of fluidized bed drying in order to enable immediate release at the drying endpoint and ensure the accuracy of moisture prediction results. Summary of the Invention
[0005] This application provides a method and apparatus for determining the endpoint of fluidized bed drying, which can realize the immediate release at the drying endpoint and ensure the accuracy of moisture prediction results. The technical solution is as follows.
[0006] In a first aspect, this application provides a method for determining the endpoint of fluidized bed drying, the method comprising:
[0007] Historical key relevant parameters of multiple batches during the fluidized bed drying process were collected; the historical key relevant parameters include inlet air temperature, material temperature, and outlet air temperature.
[0008] The historical key parameters are subjected to dimensionality reduction processing, and a drying endpoint judgment model is established;
[0009] The historical key parameters after dimensionality reduction are processed to obtain the threshold for judging the drying endpoint by inter-group mean difference processing.
[0010] Based on the drying endpoint determination model, the drying dimension reduction data of the material to be dried is obtained;
[0011] The drying dimensionality reduction data is monitored based on the drying endpoint determination threshold to control the drying endpoint of the material to be dried in real time.
[0012] Based on the aforementioned technical means, this application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during the drying process to perform secondary dimensionality reduction and analyze the comprehensive temperature change law, obtaining the inlet air temperature, material temperature, and outlet air temperature that have the highest correlation with material changes. Then, a drying endpoint judgment model is established based on historical key relevant parameters collected multiple times. Simultaneously, the inter-group mean difference of the historical key relevant parameters is processed to obtain the drying endpoint judgment threshold, i.e., the confidence line threshold. In subsequent monitoring, the dimensionality-reduced drying data of the material to be dried can be obtained through the drying endpoint judgment model, and the Hotling T-squared value of the material to be dried can be calculated. At this point, based on the Hotling T-squared value of the material to be dried and the drying endpoint judgment threshold, the drying endpoint of the material to be dried can be controlled in real time, achieving immediate release at the drying endpoint and ensuring the accuracy of the moisture prediction results.
[0013] In conjunction with the first aspect, in one embodiment, prior to collecting historical key relevant parameters of the fluidized bed across multiple batches during the drying process, the method further includes:
[0014] The material spectrum of the fluidized bed during the drying process is obtained, and the material spectrum is subjected to a dimension reduction process.
[0015] The material spectrum after a first dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process, and the combination result is subjected to a second dimensionality reduction process to determine the key relevant parameters of material changes from the process parameters; the process parameters include the expansion chamber pressure, air volume, air temperature, material temperature, and air outlet temperature.
[0016] Based on the above technical means, this application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during the drying process to perform secondary dimensionality reduction and analysis of the comprehensive temperature change law, thereby obtaining the inlet air temperature, material temperature, and outlet air temperature that have the highest correlation with material changes. At this point, the inlet air temperature, material temperature, and outlet air temperature with the highest correlation can represent material changes. Further analysis of the above three temperatures can predict the drying endpoint.
[0017] In conjunction with the first aspect, in one embodiment, the step of combining the material spectrum after a first dimensionality reduction process with the process parameters of the fluidized bed during the drying process, and then performing a second dimensionality reduction process on the combined result, to determine key relevant parameters of material changes from the process parameters, includes:
[0018] The material spectrum after a single dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process;
[0019] The combination results are subjected to a second dimensionality reduction process to convert the three-dimensional data corresponding to the combination results into two-dimensional data, and to obtain the comprehensive temperature change law of the fluidized bed during the drying process.
[0020] Based on the comprehensive temperature change pattern, key relevant parameters for material changes are determined from the process parameters.
[0021] In conjunction with the first aspect, in one implementation, before performing dimensionality reduction processing on the historical key-related parameters, the method further includes:
[0022] The historical key related parameters are standardized and preprocessed to convert them into a standard normal distribution.
[0023] Based on the aforementioned technical means, this application uses standardized preprocessing to adjust historical key parameters to a unified scale for subsequent data analysis or model training.
[0024] In conjunction with the first aspect, in one implementation, the step of performing dimensionality reduction processing on the historical key-related parameters and establishing a drying endpoint determination model includes:
[0025] The historical key parameters are converted into an initial parameter matrix by column.
[0026] The initial parameter matrix is zero-mean processed to obtain the covariance matrix corresponding to the initial parameter matrix;
[0027] Obtain the eigenvalues of the covariance matrix and the corresponding eigenvectors;
[0028] The feature vectors are arranged into a feature matrix according to the magnitude of their corresponding feature values to establish a drying endpoint determination model.
[0029] Based on the above technical means, this application establishes a drying endpoint judgment model to reduce the high-dimensional data of key related parameters to low-dimensionality, improves computational efficiency, and extracts the principal components of the data to achieve feature extraction.
[0030] In conjunction with the first aspect, in one implementation, the step of performing inter-group mean difference processing on the historical key-related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold includes:
[0031] The historical key relevant parameters after dimensionality reduction are subjected to a first inter-group mean difference calculation to obtain a first Hotling T-squared value;
[0032] Based on the first Hotling T-squared value and the sample size involved in the modeling of the drying endpoint judgment model, a confidence line threshold is obtained; the confidence line threshold indicates the drying endpoint judgment threshold.
[0033] Based on the aforementioned technical methods, after dimensionality reduction processing, this application calculates Hotelling's T. 2 Statistical measures are used to assess the arrival of the drying endpoint, Hotelling's T 2 Statistics can be used to determine whether there are significant differences in the data distribution at different stages of the drying process, thereby ensuring the accurate determination of the drying endpoint.
[0034] In conjunction with the first aspect, in one embodiment, monitoring the drying dimensionality reduction data based on the drying endpoint determination threshold to control the drying endpoint of the material to be dried in real time includes:
[0035] The dry dimensionality reduction data is subjected to a second inter-group mean difference calculation to obtain the second Hotling T-squared value;
[0036] When the second-order Hotling T-squared value reaches or falls below the drying endpoint determination threshold, the drying process of the material to be tested is determined to have reached the drying endpoint.
[0037] Based on the above technical means, this application compares the magnitude of the second-order Hotling T-squared value with the drying endpoint judgment threshold to determine whether the drying process has reached the endpoint.
[0038] Secondly, this application provides a device for determining the endpoint of fluidized bed drying, the device comprising:
[0039] The historical key relevant parameter acquisition module is used to collect historical key relevant parameters of multiple batches during the fluidized bed drying process; the historical key relevant parameters include inlet air temperature, material temperature and outlet air temperature.
[0040] The drying endpoint determination model establishment module is used to perform dimensionality reduction processing on the historical key related parameters and establish a drying endpoint determination model;
[0041] The drying endpoint judgment threshold acquisition module is used to perform inter-group mean difference processing on the historical key related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold.
[0042] The drying dimension reduction data acquisition module is used to acquire the drying dimension reduction data of the material to be dried based on the drying endpoint judgment model.
[0043] The drying endpoint control module is used to monitor the drying dimensionality reduction data according to the drying endpoint judgment threshold, so as to control the drying endpoint of the material to be tested in real time.
[0044] Thirdly, this application provides a computer device, which includes a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described method for determining the endpoint of fluidized bed drying.
[0045] Fourthly, this application provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the above-described method for determining the endpoint of fluidized bed drying.
[0046] The technical solution provided in this application may include the following beneficial effects:
[0047] This application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during drying, performs secondary dimensionality reduction, and analyzes the comprehensive temperature change law to obtain the inlet air temperature, material temperature, and outlet air temperature, which have the highest correlation with material changes. Then, based on historical key relevant parameters collected multiple times, a drying endpoint judgment model is established. Simultaneously, the inter-group mean difference of the historical key relevant parameters is processed to obtain the drying endpoint judgment threshold, i.e., the confidence line threshold. In subsequent monitoring, the dimensionality-reduced drying data of the material to be dried can be obtained through the drying endpoint judgment model, and the Hotling T-squared value of the material to be dried can be calculated. At this point, based on the Hotling T-squared value of the material to be dried and the drying endpoint judgment threshold, the drying endpoint of the material to be dried can be controlled in real time, achieving immediate release at the drying endpoint and ensuring the accuracy of the moisture prediction results. Attached Figure Description
[0048] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0049] Figure 1 This is a schematic diagram of a fluidized bed drying endpoint determination system according to an exemplary embodiment.
[0050] Figure 2 This is a flowchart illustrating a method for determining the endpoint of fluidized bed drying according to an exemplary embodiment.
[0051] Figure 3 This is a flowchart illustrating a method for determining the endpoint of fluidized bed drying according to an exemplary embodiment.
[0052] Figure 4This is a schematic diagram of a two-dimensional load for principal component analysis (PCA) dimensionality reduction processing, according to an exemplary embodiment.
[0053] Figure 5 This is a schematic diagram of PCA two-dimensional scores for three temperature SCADA parameters during a multi-batch drying process, according to an exemplary embodiment.
[0054] Figure 6 This is a schematic diagram illustrating the first-order Hotelling T-squared value and 95% confidence line of historical key related parameters according to an exemplary embodiment.
[0055] Figure 7 This is a monitoring schematic diagram of a drying endpoint determination model according to an exemplary embodiment.
[0056] Figure 8 This is a structural block diagram of a fluidized bed drying endpoint determination device according to an exemplary embodiment.
[0057] Figure 9 A structural block diagram of a computer device illustrated in an exemplary embodiment of this application is shown. Detailed Implementation
[0058] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0059] Figure 1 This is a schematic diagram of a fluidized bed drying endpoint determination system according to an exemplary embodiment. The system includes a fluidized bed 110, a near-infrared spectrometer 120, and a server 130.
[0060] Furthermore, the near-infrared spectrometer 120 is installed on the fluidized bed 110. The near-infrared spectrometer 120 is used to collect the material spectrum on the fluidized bed 110 during the drying process, so as to screen out the variables most related to the material changes during the drying process, namely key related parameters, including inlet air temperature, material temperature and outlet air temperature.
[0061] Furthermore, the fluidized bed 110 is used to dry materials. During the drying process, the fluidized bed 110 places the bulk material on the perforated plate and conveys gas from below, causing the material particles to move on the gas distribution plate and be suspended in the airflow, generating a mixed bottom layer of material particles and gas. The material particles are in full contact with the gas in this mixed bottom layer, and heat and moisture transfer between the material and the gas are carried out.
[0062] Furthermore, the fluidized bed 110 and the near-infrared spectrometer 120 communicate with the server 130. The server 130 performs a first dimensionality reduction on the material spectrum collected by the near-infrared spectrometer 120, and then combines the material spectrum after the first dimensionality reduction with the process parameters of the fluidized bed during the drying process to perform a second dimensionality reduction and analyze the comprehensive temperature change law to obtain the inlet air temperature, material temperature and outlet air temperature that are most correlated with the material change. Then, based on the three temperature parameters of inlet air temperature, material temperature and outlet air temperature, a drying endpoint judgment model is established, and the drying endpoint judgment threshold, i.e., the confidence line threshold, is calculated.
[0063] During subsequent monitoring, the server 130 can obtain the reduced-dimensionality data of the material to be dried through the drying endpoint judgment model, calculate the Hotling T-square value of the material to be dried, and when the Hotling T-square value of the material to be dried reaches or falls below the drying endpoint judgment threshold, the server 130 feeds back to the SCADA system (data acquisition and monitoring control system) to close the steam valve 110 on the fluidized bed, thereby controlling the drying endpoint in real time, realizing the immediate release of the drying endpoint, and ensuring the accuracy of the moisture prediction results.
[0064] Figure 2 This is a flowchart illustrating a method for determining the endpoint of fluidized bed drying according to an exemplary embodiment. The method is executed by a computer device, which may be, for example... Figure 1 Server 130 is shown in the image. (As shown...) Figure 2 As shown, the method may include the following steps:
[0065] Step S201: Collect historical key relevant parameters of multiple batches during the fluidized bed drying process; these historical key relevant parameters include inlet air temperature, material temperature and outlet air temperature.
[0066] In one possible implementation, before collecting historical key relevant parameters in multiple batches, this embodiment first analyzes the process parameters of the fluidized bed during the drying process, i.e., the SCADA parameters. The SCADA parameters to be monitored in the fluidized bed drying process include the expansion chamber pressure, inlet air volume, inlet air temperature, material temperature, and outlet air temperature. However, not all parameters are highly correlated with material changes. In order to screen out the variables most related to material changes during the drying process, this embodiment obtains the three key relevant parameters with the highest correlation to material changes from the SCADA parameters through the acquisition of material spectra and dimensionality reduction of SCADA parameters, namely, inlet air temperature, material temperature, and outlet air temperature. Therefore, the material changes of the fluidized bed during the drying process can be analyzed based on the key relevant parameters, and historical key relevant parameters of the fluidized bed in multiple batches during the drying process can be collected.
[0067] Step S202: Dimensionality reduction of the relevant historical parameters and establishment of a drying endpoint judgment model.
[0068] In one possible implementation, after collecting multiple batches of historical key related parameters, the historical key related parameters are first subjected to dimensionality reduction processing to extract the feature values of the historical key related parameters, and then a drying endpoint judgment model is constructed. Through this drying endpoint judgment model, the key related parameters corresponding to the material to be dried can be directly subjected to dimensionality reduction processing in the subsequent drying process.
[0069] Step S203: Perform inter-group mean difference processing on the historical key related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold.
[0070] In one possible implementation, this embodiment compares the differences in means between different groups through inter-group mean difference processing, such as Hotelling's T. 2 (Hotling T-squared distribution) By processing the difference in mean between groups, the Hotling T-squared value of the historical key related parameter is first obtained, and then the drying endpoint judgment threshold is calculated based on the Hotling T-squared value.
[0071] Step S204: Based on the drying endpoint judgment model, obtain the drying dimension reduction data of the material to be dried.
[0072] In one possible implementation, after obtaining the drying endpoint judgment threshold, the dimensionality reduction data of the drying material to be tested can be monitored through the drying endpoint judgment threshold. At this time, the key relevant parameters corresponding to the drying material to be tested are first input into the completed drying endpoint judgment model for dimensionality reduction processing, and the dimensionality reduction data of the drying material to be tested can be obtained.
[0073] Step S205: Monitor the drying dimensionality reduction data based on the drying endpoint determination threshold to control the drying endpoint of the material to be tested in real time.
[0074] In one possible implementation, after obtaining the dimensionality-reduced drying data of the material to be tested, a secondary inter-group mean difference calculation is performed on the dimensionality-reduced drying data to obtain the Hotling T-square value corresponding to the material to be tested. The Hotling T-square value corresponding to the material to be tested is compared with the drying endpoint judgment threshold to control the drying endpoint of the material to be tested in real time. Generally, when the Hotling T-square value corresponding to the material to be tested reaches or falls below the drying endpoint judgment threshold, it is determined that the drying process of the material to be tested has reached the drying endpoint.
[0075] In summary, this application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during drying, performs secondary dimensionality reduction, and analyzes the comprehensive temperature change law to obtain the inlet air temperature, material temperature, and outlet air temperature, which have the highest correlation with material changes. Then, based on historical key relevant parameters collected multiple times, a drying endpoint judgment model is established. Simultaneously, the inter-group mean difference of the historical key relevant parameters is processed to obtain the drying endpoint judgment threshold, i.e., the confidence line threshold. In subsequent monitoring, the dimensionality-reduced drying data of the material to be dried can be obtained through the drying endpoint judgment model, and the Hotling T-squared value of the material to be dried can be calculated. At this point, based on the Hotling T-squared value of the material to be dried and the drying endpoint judgment threshold, the drying endpoint of the material to be dried can be controlled in real time, achieving immediate release at the drying endpoint and ensuring the accuracy of the moisture prediction results.
[0076] Figure 3 This is a flowchart illustrating a method for determining the endpoint of fluidized bed drying according to an exemplary embodiment. The method is executed by a computer device, which may be, for example... Figure 1 Server 130 is shown in the image. (As shown...) Figure 3 As shown, the method may include the following steps:
[0077] Step S301: Obtain the material spectrum of the fluidized bed during the drying process, and perform a dimensionality reduction process on the material spectrum. This dimensionality reduction process is a principal component analysis (PCA) dimensionality reduction process.
[0078] Step S302: Combine the material spectrum after the first dimensionality reduction process with the process parameters of the fluidized bed during the drying process, and perform a second dimensionality reduction process on the combined result to determine the key relevant parameters of material changes from the process parameters. These process parameters include the expansion chamber pressure, inlet air volume, inlet air temperature, material temperature, and outlet air temperature. The key relevant parameters include the inlet air temperature, material temperature, and outlet air temperature. This second dimensionality reduction process is a second principal component analysis (PCA) dimensionality reduction process.
[0079] In one possible implementation, step S302 includes:
[0080] The spectrum of the material after a single dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process;
[0081] The combination result is subjected to a second dimensionality reduction process to convert the three-dimensional data corresponding to the combination result into two-dimensional data, and to obtain the comprehensive temperature change law of the fluidized bed during the drying process.
[0082] Based on the overall temperature change pattern, the key relevant parameters for material changes are determined from the process parameters.
[0083] Furthermore, the SCADA parameters (i.e., the aforementioned process parameters) to be monitored during the fluidized bed drying process include the expansion chamber pressure, inlet air volume, inlet air temperature, material temperature, and outlet air temperature. To identify the variables most relevant to material changes during the drying process, this embodiment installs a near-infrared spectrometer on the fluidized bed to collect the material spectrum during the drying process. The material spectrum after initial principal component analysis (PCA) dimensionality reduction is then combined with the process parameters and subjected to another PCA dimensionality reduction. (See [link to PCA documentation] for details.) Figure 4 The diagram shows a two-dimensional loading of principal component analysis (PCA) dimensionality reduction. This two-dimensional loading diagram mainly reflects the importance of variables and the relationships between them. The horizontal axis represents the PCA first principal component score, and the vertical axis represents the PCA second principal component score. Figure 4 The results show that the material changes are most strongly correlated with the inlet air temperature, the material temperature, and the outlet air temperature, and have very low correlation with the inlet air volume and the expansion chamber pressure. Therefore, only these three temperature parameters need to be analyzed in the future.
[0084] At this point, see again Figure 5 The diagram shows the PCA two-dimensional scores of three temperature SCADA parameters during multiple batches of drying processes. If the inlet air temperature, material temperature, and outlet air temperature SCADA parameters were plotted as three-dimensional data, it would be inconvenient to observe data patterns. Therefore, this embodiment uses principal component analysis (PCA) dimensionality reduction algorithm to project the original three-dimensional data into two-dimensional data. Figure 5 The x-axis represents the score of the first principal component of the PCA algorithm, and the y-axis represents the score of the second principal component. The total contribution of the two principal components is 96%, indicating that the two principal components after projection can represent the original three-dimensional data patterns and reflect the comprehensive temperature change patterns during the drying process. Figure 5 The different shapes of the curves represent different batches of materials (such as 230802-1, 230802-2, 230802-3, 230802-4, 230803-3, 230804-3, and 230803-1); from Figure 5 The data shows distinct trajectories for the three temperature SCADA parameters for each batch of material. However, due to differences in environment and initial material conditions, the data trajectories for different batches show significant variations. Based on the above analysis, the three temperature SCADA parameters can be used to represent material changes. Therefore, the trajectory differences are mainly due to variations in the initial moisture content of the material and different equipment conditions, leading to different heating rates for the three temperature SCADA parameters, and consequently, different times to reach the material release standard. Although there are significant differences in the starting point and process trajectory, Figure 5 The endpoints of the three SCADA parameters for each batch all projected into the same range, indicating that when the material reaches the drying endpoint, the material temperature, inlet air temperature, and outlet air temperature all reach a dynamic equilibrium.
[0085] Step S303: Collect historical key relevant parameters for multiple batches during the fluidized bed drying process; these historical key relevant parameters include inlet air temperature, material temperature, and outlet air temperature. The moisture content of these multiple batches of historical key relevant parameters is qualified and relatively uniform.
[0086] Furthermore, since establishing a drying endpoint determination model requires offline research and collection of historical data, and the drying endpoints in historical data are determined by workers, considering the differences in operation time among different workers and the different endpoint determination points among different workers, it is necessary to combine the moisture content at the end of drying and select qualified batches with relatively uniform moisture content as standard modeling batches. A complete production process is considered as one batch. In this embodiment, at least 20 batches with qualified moisture content are collected, along with key historical parameters (inlet air temperature, material temperature, and outlet air temperature). Temperature sensors are installed on the fluidized bed drying equipment, and data is collected by the SCADA system. The SCADA system can collect temperature sensor data in real time, and the data changes over time, with a collection interval of several seconds per point. The key historical parameters near the drying endpoint of 20 batches (exemplarily, 5 to 10 time points are collected for each batch) are used to establish the subsequent drying endpoint determination model.
[0087] Step S304: Perform standardization preprocessing on the historical key related parameters to convert them into a standard normal distribution.
[0088] Furthermore, before performing dimensionality reduction on the historical key related parameters, this embodiment also needs to perform standardization preprocessing on the collected historical key related parameters (inlet air temperature, outlet air temperature, and material temperature). The calculation method for standardization preprocessing is as follows:
[0089]
[0090] Where Z represents the standardized preprocessed data, X represents the original historical key relevant parameters, μ represents the mean, and σ represents the standard deviation.
[0091] Step S305: After standardization preprocessing, dimensionality reduction is performed on the relevant historical parameters, and a drying endpoint determination model is established. This dimensionality reduction is a three-stage process, specifically a three-stage principal component analysis (PCA) dimensionality reduction.
[0092] In one possible implementation, step S305 includes:
[0093] Convert the relevant parameters of this historical key point into an initial parameter matrix by column;
[0094] The initial parameter matrix is zero-mean normalized to obtain the covariance matrix corresponding to the initial parameter matrix;
[0095] Obtain the eigenvalues of the covariance matrix and the corresponding eigenvectors;
[0096] The feature vectors are arranged into a feature matrix according to the magnitude of their corresponding feature values to establish a drying endpoint determination model.
[0097] Furthermore, in this embodiment, the historical key-related parameters after standardization preprocessing are reduced in dimensionality using the PCA algorithm, and a PCA model (i.e., the aforementioned drying endpoint determination model) is established, including:
[0098] 1) Arrange the historical key parameters into an n x m matrix Z (i.e., the initial parameter matrix mentioned above);
[0099] 2) Zero-mean value is applied to each row of matrix Z, i.e., the mean of that row is subtracted;
[0100] 3) Calculate the covariance matrix;
[0101] 4) Find the eigenvalues and corresponding eigenvectors of the covariance matrix;
[0102] 5) Arrange the feature vectors into a matrix (i.e., the feature matrix above) from top to bottom according to the size of the corresponding feature values, take the first k rows to form matrix P, and then construct the above drying endpoint judgment model.
[0103] 6) Input the new key relevant parameters of the material to be dried into the drying endpoint judgment model. Y = PXnew is the drying dimension-reduced data after being reduced to k dimensions.
[0104] Step S306: Perform a first inter-group mean difference calculation on the historical key relevant parameter after dimensionality reduction to obtain a first Hotelling T-squared value. The dimensionality reduction here is a three-stage dimensionality reduction, i.e., a three-stage principal component analysis (PCA) dimensionality reduction.
[0105] Furthermore, in this embodiment, Hotelling's T is calculated using the score data after dimensionality reduction processing by principal component analysis (PCA). 2 The 95% confidence line is used as the threshold for determining the drying endpoint; first, Hotelling's T is calculated. 2 The statistic, namely the first-order Hotelling T-squared value mentioned above, is obtained using the following formula:
[0106] T 2 =X b PΛ -1 P T X;
[0107] Λ=diag(λ1,…λ k};
[0108] Step S307: Obtain the confidence line threshold based on the first Hotling T-squared value and the sample size involved in the modeling of the drying endpoint judgment model; the confidence line threshold indicates the drying endpoint judgment threshold.
[0109] Furthermore, in calculating Hotelling's T 2 The statistic, namely the Hotelling's T-squared value mentioned above, is used in this embodiment to find Hotelling's T-squared value. 2 The 95% confidence line of the statistic is used as the threshold for determining the drying endpoint, which is obtained using the following formula:
[0110]
[0111] Where k represents the dimension of the dimensionality reduction process using Principal Component Analysis (PCA), n represents the sample size involved in the modeling, which is the product of the number of collection batches of the historical key relevant parameter and the number of time points extracted in each batch, 1-α represents the confidence level, which is set to 0.05 here, F α (k,nk) represents an F-distribution with k degrees of freedom and nk degrees of freedom; please refer to [link to relevant documentation]. Figure 6 The diagram shows the first-order Hotelling T-squared value and 95% confidence line of the historical key relevant parameters. The horizontal axis corresponds to the time point near the drying end of each batch, the vertical axis is the Hotelling T-squared value, the horizontal line represents the 95% confidence line, and the curve represents the first-order Hotelling T-squared value of the historical key relevant parameters for each batch.
[0112] Step S308: Based on the drying endpoint judgment model, obtain the drying dimension reduction data of the material to be dried.
[0113] In one possible implementation, the key relevant parameters of the dried material to be tested are input into the drying endpoint determination model. The drying endpoint determination model can then perform principal component analysis (PCA) dimensionality reduction on the key relevant parameters of the dried material to be tested and output the dimensionality reduction data of the dried material to be tested.
[0114] Step S309: Perform a second inter-group mean difference calculation on the dried dimensionality reduction data to obtain the second Hotling T-squared value.
[0115] For further details, please see Figure 7 The diagram shown illustrates the monitoring of the drying endpoint determination model. The horizontal axis represents each time point, and the vertical axis represents the Hotelling T-squared value. The curve represents the Hotelling T-squared value at each time point, and the horizontal line represents the 95% confidence line. This embodiment uses the same formula as described above to perform a secondary inter-group mean difference calculation on the drying dimensionality reduction data to obtain the secondary Hotelling T-squared value, which will not be elaborated upon here.
[0116] Step S310: When the second-order Hotling T-squared value reaches or falls below the drying endpoint judgment threshold, it is determined that the drying process of the material to be tested has reached the drying endpoint.
[0117] Furthermore, after obtaining the second-order Hotling T-squared value, comparing it with the drying endpoint judgment threshold allows for the determination of whether the drying process of the material under test has reached the drying endpoint. A second-order Hotling T-squared value greater than the drying endpoint judgment threshold is considered an outlier. When the second-order Hotling T-squared value falls below the drying endpoint judgment threshold, it indicates that the data is similar to the modeled drying endpoint data. This data is then fed back to the SCADA system, which closes the steam valve of the fluidized bed and controls the drying endpoint in real time.
[0118] In summary, this application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during drying, performs secondary dimensionality reduction, and analyzes the comprehensive temperature change law to obtain the inlet air temperature, material temperature, and outlet air temperature, which have the highest correlation with material changes. Then, based on historical key relevant parameters collected multiple times, a drying endpoint judgment model is established. Simultaneously, the inter-group mean difference of the historical key relevant parameters is processed to obtain the drying endpoint judgment threshold, i.e., the confidence line threshold. In subsequent monitoring, the dimensionality-reduced drying data of the material to be dried can be obtained through the drying endpoint judgment model, and the Hotling T-squared value of the material to be dried can be calculated. At this point, based on the Hotling T-squared value of the material to be dried and the drying endpoint judgment threshold, the drying endpoint of the material to be dried can be controlled in real time, achieving immediate release at the drying endpoint and ensuring the accuracy of the moisture prediction results.
[0119] Figure 8 This is a structural block diagram illustrating a fluidized bed drying endpoint determination device according to an exemplary embodiment. The device includes:
[0120] The historical key relevant parameter acquisition module 801 is used to collect historical key relevant parameters of multiple batches during the drying process of the fluidized bed; these historical key relevant parameters include inlet air temperature, material temperature and outlet air temperature.
[0121] The drying endpoint judgment model establishment module 802 is used to perform dimensionality reduction processing on the relevant historical key parameters and establish a drying endpoint judgment model;
[0122] The drying endpoint judgment threshold acquisition module 803 is used to perform inter-group mean difference processing on the historical key related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold.
[0123] The drying dimension reduction data acquisition module 804 is used to acquire the drying dimension reduction data of the material to be dried based on the drying endpoint judgment model.
[0124] The drying endpoint control module 805 is used to monitor the drying dimensionality reduction data according to the drying endpoint judgment threshold, so as to control the drying endpoint of the material to be tested in real time.
[0125] In one possible implementation, the device is also used for:
[0126] Obtain the material spectrum of the fluidized bed during the drying process, and perform a dimension reduction process on the material spectrum;
[0127] The material spectrum after a first dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process, and the combined result is subjected to a second dimensionality reduction process to determine the key relevant parameters of material change from the process parameters. The process parameters include the expansion chamber pressure, air volume, air temperature, material temperature, and air outlet temperature.
[0128] In one possible implementation, the device is also used for:
[0129] The spectrum of the material after a single dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process;
[0130] The combination result is subjected to a second dimensionality reduction process to convert the three-dimensional data corresponding to the combination result into two-dimensional data, and to obtain the comprehensive temperature change law of the fluidized bed during the drying process.
[0131] Based on the overall temperature change pattern, the key relevant parameters for material changes are determined from the process parameters.
[0132] In one possible implementation, the device is also used for:
[0133] The relevant parameters of this historical key are standardized and preprocessed to transform them into a standard normal distribution.
[0134] In one possible implementation, the drying endpoint determination model establishment module 802 is further configured to:
[0135] Convert the relevant parameters of this historical key point into an initial parameter matrix by column;
[0136] The initial parameter matrix is zero-mean normalized to obtain the covariance matrix corresponding to the initial parameter matrix;
[0137] Obtain the eigenvalues of the covariance matrix and the corresponding eigenvectors;
[0138] The feature vectors are arranged into a feature matrix according to the magnitude of their corresponding feature values to establish a drying endpoint determination model.
[0139] In one possible implementation, the drying endpoint determination threshold acquisition module 803 is further configured to:
[0140] A first-order inter-group mean difference calculation is performed on the historical key relevant parameters after dimensionality reduction to obtain a first-order Hotling T-squared value;
[0141] Based on the first Hotling T-squared value and the sample size involved in the modeling of the drying endpoint judgment model, a confidence line threshold is obtained; this confidence line threshold indicates the drying endpoint judgment threshold.
[0142] In one possible implementation, the drying endpoint control module 805 is further configured to:
[0143] A second-order inter-group mean difference calculation was performed on the dried dimensionality reduction data to obtain the second-order Hotling T-squared value;
[0144] When the second-order Hotling T-squared value reaches or falls below the drying endpoint judgment threshold, the drying process of the material to be tested is judged to have reached the drying endpoint.
[0145] In summary, this application combines the material spectrum after dimensionality reduction with the process parameters of the fluidized bed during drying, performs secondary dimensionality reduction, and analyzes the comprehensive temperature change law to obtain the inlet air temperature, material temperature, and outlet air temperature, which have the highest correlation with material changes. Then, based on historical key relevant parameters collected multiple times, a drying endpoint judgment model is established. Simultaneously, the inter-group mean difference of the historical key relevant parameters is processed to obtain the drying endpoint judgment threshold, i.e., the confidence line threshold. In subsequent monitoring, the dimensionality-reduced drying data of the material to be dried can be obtained through the drying endpoint judgment model, and the Hotling T-squared value of the material to be dried can be calculated. At this point, based on the Hotling T-squared value of the material to be dried and the drying endpoint judgment threshold, the drying endpoint of the material to be dried can be controlled in real time, achieving immediate release at the drying endpoint and ensuring the accuracy of the moisture prediction results.
[0146] Please see Figure 9 This is a schematic diagram of a computer device provided according to an exemplary embodiment of the present application. The computer device includes a memory and a processor. The memory is used to store a computer program. When the computer program is executed by the processor, it implements the above-described method for determining the endpoint of fluidized bed drying.
[0147] The processor can be a central processing unit (CPU). It can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations thereof.
[0148] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of this application. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the methods in the above-described embodiments.
[0149] The memory may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
[0150] In one exemplary embodiment, a computer-readable storage medium is also provided for storing at least one computer program, which is loaded and executed by a processor to implement all or part of the steps in the above-described method. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, or optical data storage device, etc.
[0151] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.
[0152] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.
Claims
1. A method for determining the endpoint of fluidized bed drying, characterized in that, The method includes: Historical key relevant parameters of multiple batches during the fluidized bed drying process were collected; the historical key relevant parameters include inlet air temperature, material temperature, and outlet air temperature. The historical key parameters are subjected to dimensionality reduction processing, and a drying endpoint judgment model is established; The historical key parameters after dimensionality reduction are processed to obtain the threshold for judging the drying endpoint by inter-group mean difference processing. Based on the drying endpoint determination model, the drying dimension reduction data of the material to be dried is obtained; The drying dimensionality reduction data is monitored based on the drying endpoint determination threshold to control the drying endpoint of the material to be dried in real time. Prior to collecting historical key parameters from multiple batches during the fluidized bed drying process, the method further includes: The material spectrum of the fluidized bed during the drying process is obtained, and the material spectrum is subjected to a dimension reduction process. The material spectrum after a first dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process, and the combination result is subjected to a second dimensionality reduction process to determine the key relevant parameters of material changes from the process parameters; the process parameters include the expansion chamber pressure, air volume, air inlet temperature, material temperature, and air outlet temperature.
2. The method according to claim 1, characterized in that, The process involves combining the material spectrum after a first dimensionality reduction with the process parameters of the fluidized bed during drying, and then performing a second dimensionality reduction on the combined result to determine key relevant parameters of material changes from the process parameters, including: The material spectrum after a single dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process; The combination results are subjected to a second dimensionality reduction process to convert the three-dimensional data corresponding to the combination results into two-dimensional data, and to obtain the comprehensive temperature change law of the fluidized bed during the drying process. Based on the comprehensive temperature change pattern, key relevant parameters for material changes are determined from the process parameters.
3. The method according to claim 1, characterized in that, Before performing dimensionality reduction on the historical key-related parameters, the method further includes: The historical key related parameters are standardized and preprocessed to convert them into a standard normal distribution.
4. The method according to claim 1, characterized in that, The step of performing dimensionality reduction on the historical key parameters and establishing a drying endpoint determination model includes: The historical key parameters are converted into an initial parameter matrix by column. The initial parameter matrix is zero-mean processed to obtain the covariance matrix corresponding to the initial parameter matrix; Obtain the eigenvalues of the covariance matrix and the eigenvectors corresponding to the eigenvalues; The feature vectors are arranged into a feature matrix according to the magnitude of their corresponding feature values to establish a drying endpoint determination model.
5. The method according to claim 1, characterized in that, The step of performing inter-group mean difference processing on the historical key related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold includes: The historical key relevant parameters after dimensionality reduction are subjected to a first inter-group mean difference calculation to obtain a first Hotling T-squared value; Based on the first Hotling T-squared value and the sample size involved in the modeling of the drying endpoint judgment model, a confidence line threshold is obtained; the confidence line threshold indicates the drying endpoint judgment threshold.
6. The method according to any one of claims 1 to 5, characterized in that, The step of monitoring the drying dimensionality reduction data based on the drying endpoint determination threshold to control the drying endpoint of the material to be dried in real time includes: The dry dimensionality reduction data is subjected to a second inter-group mean difference calculation to obtain the second Hotling T-squared value; When the second-order Hotling T-squared value reaches or falls below the drying endpoint determination threshold, the drying process of the material to be tested is determined to have reached the drying endpoint.
7. A device for determining the endpoint of fluidized bed drying, characterized in that, The device includes: The historical key relevant parameter acquisition module is used to collect historical key relevant parameters of multiple batches during the fluidized bed drying process; the historical key relevant parameters include inlet air temperature, material temperature and outlet air temperature. The drying endpoint determination model establishment module is used to perform dimensionality reduction processing on the historical key related parameters and establish a drying endpoint determination model; The drying endpoint judgment threshold acquisition module is used to perform inter-group mean difference processing on the historical key related parameters after dimensionality reduction to obtain the drying endpoint judgment threshold. The drying dimension reduction data acquisition module is used to acquire the drying dimension reduction data of the material to be dried based on the drying endpoint judgment model. The drying endpoint control module is used to monitor the drying dimensionality reduction data according to the drying endpoint judgment threshold, so as to control the drying endpoint of the material to be dried in real time. This device is also used for: The material spectrum of the fluidized bed during the drying process is obtained, and the material spectrum is subjected to a dimension reduction process. The material spectrum after a first dimensionality reduction process is combined with the process parameters of the fluidized bed during the drying process, and the combination result is subjected to a second dimensionality reduction process to determine the key relevant parameters of material changes from the process parameters; the process parameters include the expansion chamber pressure, air volume, air temperature, material temperature, and air outlet temperature.
8. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement a method for determining the endpoint of fluidized bed drying as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement a method for determining the endpoint of fluidized bed drying as described in any one of claims 1 to 6.