Automated Precision Control Method and System for the Production Process of Macrocyclic Acid Derivatives

Abstract

The present invention discloses an automatic precise control method and system for the production process of macrolide acid derivatives, which relates to the field of intelligent control technology. It collects the time-series data of multi-dimensional process parameters and the spectral feature data of raw materials of the reaction kettle in real time, performs dynamic adaptive filtering processing on the time-series data of multi-dimensional process parameters to generate a denoised process parameter sequence, performs baseline correction on the spectral feature data of raw materials to generate standard spectral data, conducts feature fusion, constructs a process parameter feature matrix, inputs it into an improved regression model with process knowledge constraints, outputs a key parameter sensitivity vector and a predicted product yield value, extracts a dominant parameter sub-matrix from the process parameter feature matrix, constructs a time-varying state space model using a process time series alignment algorithm, dynamically compares the time-varying state space model with the predicted product yield value, and generates an optimized control parameter set through a multi-objective optimization model, which can significantly improve the product yield of macrolide acid derivatives.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and specifically to an automatic precise control method and system for the production process of macrolide acid derivatives. Background Art

[0002] As an important chemical auxiliary, macrolide acid derivatives are widely used in fields such as fine chemicals, petrochemicals, papermaking, and light industry. Their production process usually involves multiple steps of reactions, strict process conditions, and complex reaction kinetics. During the production process, it is necessary to not only monitor in real time the conventional process parameters such as the temperature, pressure, and pH value in the reaction kettle, but also combine multi-dimensional data such as the spectral information, flow rate, and temperature of the raw materials to accurately control the reaction state. Due to the highly non-linear and time-varying nature of the reaction process, combined with factors such as fluctuations in the quality of raw materials, lag in equipment response, and environmental interference, traditional manual control and control methods based on static models are difficult to meet the requirements of product quality and production efficiency. Therefore, intelligent and automatic precise control has become an inevitable trend to achieve high-quality and stable production.

[0003] Existing process control methods are mostly based on linear or static parameter correction, which are difficult to cope with the complex situation of dynamic changes in process parameters over time, and lack in-depth fusion processing of multi-sensor and multi-dimensional data, making it difficult to utilize information such as process parameters and spectral characteristics simultaneously;

[0004] Therefore, the present invention proposes an automatic precise control method and system for the production process of macrolide acid derivatives. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. For this purpose, the present invention proposes an automatic precise control method and system for the production process of macrolide acid derivatives, which can significantly improve the product yield of macrolide acid derivatives.

[0006] To achieve the above object, an automatic precise control method for the production process of macrolide acid derivatives is proposed, including the following steps:

[0007] Step 1: Real-time collect the time-series data of multi-dimensional process parameters of the reaction kettle through a distributed sensing system, and synchronously obtain the spectral feature data of the raw materials;

[0008] Step 2: Perform dynamic adaptive filtering processing on the time-series data of multi-dimensional process parameters to generate a denoised process parameter sequence, and at the same time perform baseline correction on the spectral feature data of the raw materials to generate standard spectral data;

[0009] Step 3: Perform feature fusion on the denoised process parameter sequence and the standard spectral data to construct a process parameter feature matrix, input it into an improved regression model constrained by process knowledge, and output the key parameter sensitivity vector and the predicted value of the product yield;

[0010] Step 4: Based on the key parameter sensitivity vector, extract the dominant parameter sub-matrix from the process parameter feature matrix, and construct a time-varying state space model using the process time series alignment algorithm;

[0011] Step 5: Dynamically compare the time-varying state space model with the predicted product yield value. When the deviation exceeds the preset threshold, generate an optimized control parameter set through the multi-objective optimization model;

[0012] Step 6: Input the optimized control parameter set into the adaptive controller to generate execution instructions to control the process parameters.

[0013] The steps of real-time collecting the multi-dimensional process parameter time series data of the reactor through the distributed sensing system are as follows:

[0014] Step 101: Arrange a distributed sensor network for collecting various process parameters at the key process nodes of the reactor;

[0015] Step 102: Synchronously obtain the process parameter time series corresponding to each process parameter from the sensor network, and perform time domain alignment on the time of the process parameter time series using the time protocol to generate an original data set with a unified timestamp, which constitutes the multi-dimensional process parameter time series data.

[0016] The method for synchronously obtaining the raw material spectral feature data is as follows:

[0017] Use an on-line near-infrared spectrometer with a wavelength range of 900 - 1700 nm to scan the raw material flow in real time, obtain the original spectral matrix in each sampling period, and synchronously collect the raw material flow rate and the raw material temperature;

[0018] Perform spectral denoising on the original spectral matrix using the Savitzky-Golay filter to generate a standard spectral matrix;

[0019] The standard spectral matrix, the raw material flow rate, and the raw material temperature constitute the raw material spectral feature data.

[0020] The steps of performing dynamic adaptive filtering processing on the multi-dimensional process parameter time series data to generate a denoised process parameter sequence are as follows:

[0021] Step 211: Construct the original multi-dimensional process parameter time series data with a unified timestamp into the form of a data matrix X(t);

[0022] Step 212: Set the window length as Tw, and calculate the local mean and the local variance of each process parameter within the window to form the local noise statistical information. At the same time, define the local signal-to-noise ratio as the quantization index of the noise level;

[0023] Step 213: Design an adaptive filter model based on local noise statistical information;

[0024] Step 214: After the dynamic adaptive filter model is constructed, perform point-by-point filtering on the multi-dimensional parameters at each moment t;

[0025] Step 215: Collect the time-series data sequences of the denoised process parameters after dynamic adaptive filtering processing, denoted as the denoised process parameter sequences.

[0026] The simultaneous baseline correction of the raw material spectral feature data to generate standard spectral data includes the following steps:

[0027] Step 221: Denote the standard spectral matrix as Sraw, where each column represents the spectral response within the wavelength range at a sampling period;

[0028] Step 222: First, for the spectral curve within each sampling period, extract the local minimum values of the wavelength;

[0029] Step 223: Based on the extracted local minimum values, construct a baseline model for the entire spectral curve using a low-order polynomial fitting or adaptive smoothing method;

[0030] Step 224: Use the constructed baseline model to perform baseline correction on the standard spectral matrix.

[0031] The construction of the process parameter feature matrix by fusing the denoised process parameter sequences and the standard spectral data includes the following steps:

[0032] Step 311: Mark the collected denoised process parameter sequences as and mark the collected standard spectral data as ;

[0033] Step 312: Perform correlation analysis on each parameter in the denoised process parameter sequences, eliminate redundant highly correlated parameters, and for the spectral data, extract the principal components through a dimensionality reduction method;

[0034] Step 313: Concatenate the denoised process parameter sequences and the standard spectral data after dimensionality reduction along the time dimension to form a new fusion feature matrix;

[0035] Step 314: Perform standardization processing on the fusion feature matrix generated after fusion to obtain the process parameter feature matrix.

[0036] The input of the improved regression model with process knowledge constraints and the output of the key parameter sensitivity vector and the product yield prediction value include the following steps:

[0037] Step 321: Mark the process parameter feature vector of each row in the process parameter feature matrix as X(t); mark the target variable product yield as Y(t).

[0038] Step 322: Mark the improved regression model as , and introduce prior constraints for some key parameters according to the process experience of each process stage in the production process of macrolide acid derivatives.

[0039] Step 323: Pre-collect the process parameter feature vectors and the corresponding product yields at each moment in each process stage during several production processes of macrolide acid derivatives as a sample set.

[0040] Step 324: Use the process parameter feature vectors at each moment in the sample set as the input of the improved regression model, use the mean square error between the product yields corresponding to each moment in the sample set and the predicted values calculated by the improved regression model expression as the loss function, use the prior constraints of each key parameter as the inequality constraint conditions, and use the optimization algorithm with inequality constraints to solve the intercept term and sensitivity coefficients of the improved regression model.

[0041] Step 325: The sensitivity coefficients of each key parameter obtained by solving form a key parameter sensitivity vector; for a new moment in any production process, input the process parameter feature vector of this new moment into the improved regression model to obtain the predicted value of the product yield.

[0042] The steps of extracting the dominant parameter submatrix from the process parameter feature matrix based on the key parameter sensitivity vector include the following:

[0043] Step 411: Mark the key parameter sensitivity vector as , calculate the absolute value | | of each sensitivity coefficient; where d = 1, 2,..., D.

[0044] Step 412: Determine the dominant parameter set by the threshold method or the sorting method.

[0045] Step 413: Extract the parameter values corresponding to each dominant parameter in the dominant parameter set from the process parameter feature matrix to form the dominant parameter submatrix.

[0046] The method of constructing the time-varying state space model by using the process time series alignment algorithm is as follows:

[0047] Select the dynamic time warping algorithm as the process time series alignment algorithm.

[0048] Select a stable and representative dominant parameter time series template for the normal process state from the historical data.

[0049] Obtain the current real-time dominant parameter data sequence;

[0050] Use the dynamic time warping algorithm to compare two time series data, and then construct a cost matrix to calculate the matching cost between the dominant parameter time series template and the dominant parameter data sequence, so as to find the optimal alignment path π to achieve non-linear correction on the time axis;

[0051] Utilize the dominant parameter data aligned by the dynamic time warping algorithm to construct a time-varying state space model describing the dynamic evolution of the process.

[0052] The method of dynamically comparing the time-varying state space model with the predicted product yield value is as follows:

[0053] Input the real-time process parameter feature matrix into the improved regression model to obtain the predicted product yield value;

[0054] Input the real-time dominant parameter sub-matrix into the time-varying state space model to obtain the predicted expected yield value output by the time-varying state space model;

[0055] Compare the difference between the predicted product yield value and the predicted expected yield value, and judge whether the difference is greater than the preset difference threshold.

[0056] The method of generating the optimized control parameter set through the multi-objective optimization model is as follows:

[0057] Mark the difference threshold between the predicted product yield value and the predicted expected yield value as h;

[0058] Mark the predicted expected yield value as H;

[0059] Based on the predicted expected yield value and the improved regression model, construct a multi-objective optimization function Yu;

[0060] Based on the predicted expected yield value, the improved regression model and the difference threshold, construct a set of constraint conditions U;

[0061] Take minimizing the multi-objective optimization function as the optimization goal and the set of constraint conditions as the constraint conditions to construct a multi-objective optimization model;

[0062] By using an optimization algorithm to solve the multi-objective optimization model, obtain the solution values of the parameter variables to be solved for each key parameter, and the solution values of all parameter variables constitute the optimized control parameter set.

[0063] Propose an automatic precision control system for the production process of macrocyclic acid derivatives, including an original data collection module, a preprocessing module, a yield prediction module, a state space model training module, and a control parameter optimization module; among them, each module is connected electrically;

[0064] The raw data collection module collects the time-series data of multi-dimensional process parameters of the reactor in real time through a distributed sensing system, synchronously obtains the raw material spectral feature data, and sends the time-series data of multi-dimensional process parameters and the raw material spectral feature data to the preprocessing module;

[0065] The preprocessing module performs dynamic adaptive filtering on the time-series data of multi-dimensional process parameters to generate a denoised process parameter sequence. At the same time, it corrects the baseline of the raw material spectral feature data to generate standard spectral data, and sends the denoised process parameter sequence and the standard spectral data to the yield prediction module;

[0066] The yield prediction module fuses the features of the denoised process parameter sequence and the standard spectral data, constructs a process parameter feature matrix, inputs an improved regression model with process knowledge constraints, outputs a key parameter sensitivity vector and a product yield prediction value, and sends the key parameter sensitivity vector to the state space model training module, and sends the improved regression model and the product yield prediction value to the control parameter optimization module;

[0067] The state space model training module extracts a dominant parameter submatrix from the process parameter feature matrix based on the key parameter sensitivity vector, constructs a time-varying state space model using a process time series alignment algorithm, and sends the time-varying state space model to the control parameter optimization module;

[0068] The control parameter optimization module dynamically compares the time-varying state space model with the product yield prediction value. When the deviation exceeds the preset threshold, it generates an optimized control parameter set through a multi-objective optimization model, and inputs the optimized control parameter set into an adaptive controller to generate an execution instruction to control the process parameters.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] The present invention collects process parameters such as temperature, pressure, pH, etc. and on-line near-infrared spectrum data in a reaction kettle through a distributed sensing system, and uses dynamic adaptive filtering and baseline correction to denoise the original data, improving the accuracy of the original data and reducing the pollution of noise. Then, the denoised process parameters and standardized spectrum data are dimension-reduced and weighted and fused to construct a high-dimensional feature matrix reflecting the process state. Based on the fused feature matrix and product yield label data, through a number of prior constraint conditions, an improved regression model is trained to output a key parameter sensitivity vector, realizing the accurate prediction of future yields. The dominant parameters are screened out from the fused feature matrix using the key parameter sensitivity vector, and the dynamic time warping algorithm is used for process time series alignment, and a time-varying state space model is constructed to capture the dynamic characteristics of the process, thereby obtaining an accurate expected yield. Finally, by dynamically comparing the predicted value of the time-varying state space model with the product yield collected in real time, the deviation is calculated and an optimal control parameter set is automatically generated using a multi-objective optimization model, realizing the intelligent monitoring and real-time optimal control of each key parameter in the process, thereby significantly improving the product yield and product quality, reducing energy consumption and safety risks, and enhancing the production automation level and stability. Description of the Drawings

[0071] Figure 1 It is a flowchart of the automatic precise control method for the production process of macrocyclic endoacid derivatives in Embodiment 1 of the present invention;

[0072] Figure 2 It is an example configuration diagram of the sensor network in Embodiment 1 of the present invention;

[0073] Figure 3 It is a module connection relationship diagram of the automatic precise control system for the production process of macrocyclic endoacid derivatives in Embodiment 2 of the present invention. Detailed Embodiments

[0074] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0075] In large-scale production, due to differences in raw material batches, equipment aging, and environmental factors, key parameters such as temperature, pressure, and pH value in the reactor often experience dynamic fluctuations, making it difficult to maintain stable product yield and quality. At the same time, traditional process control methods do not respond to these dynamic changes in a timely manner, and relying on manual intervention is not only inefficient and error-prone, but also prone to safety hazards. Therefore, how to use intelligent systems to capture process changes in real time, dynamically adjust control parameters, and ensure the accurate and stable operation of the entire production process is a key technical issue that needs to be solved urgently.

[0076] Example 1

[0077] like Figure 1 As shown, the method for automated and precise control of the production process of macrolide derivatives comprises the following steps:

[0078] Step 1: Use the distributed sensing system to collect the multi-dimensional process parameter time series data of the reactor in real time and simultaneously obtain the spectral characteristic data of the raw materials;

[0079] Step 2: Implement dynamic adaptive filtering on the multi-dimensional process parameter time series data to generate a denoised process parameter sequence, and perform baseline correction on the raw material spectral feature data to generate standard spectral data;

[0080] Step 3: Feature fusion of the denoised process parameter sequence and the standard spectral data, constructing a process parameter feature matrix, inputting an improved regression model constrained by process knowledge, and outputting key parameter sensitivity vectors and product yield prediction values;

[0081] Step 4: Based on the key parameter sensitivity vector, the dominant parameter sub-matrix is ​​extracted from the process parameter feature matrix, and the time-varying state space model is constructed using the process timing alignment algorithm;

[0082] Step 5: dynamically compare the time-varying state space model with the product yield prediction value. When the deviation exceeds the preset threshold, generate an optimized control parameter set through a multi-objective optimization model;

[0083] Step 6: Input the optimized control parameter set into the adaptive controller to generate execution instructions to control the process parameters.

[0084] In an embodiment of the present invention, the real-time acquisition of multi-dimensional process parameter time series data of the reactor by a distributed sensing system comprises the following steps:

[0085] Step 101: Arrange a distributed sensor network for collecting various process parameters at key process nodes of the reactor;

[0086] Specifically, the process parameters include but are not limited to the temperature, pressure and pH value in the reactor;

[0087] Specifically, ifFigure 2 As shown in the structural diagram of the reactor with a distributed sensor network, the arrangement of the distributed sensor network can be as follows:

[0088] The temperature sensing groups are evenly distributed along the axial direction of the reactor, and each sensing group contains 3 PT100 platinum resistors distributed at 120°; that is Figure 2 T1, T2, and T3 in

[0089] The three pressure sensing units are respectively installed at the top, middle, and bottom of the reactor, and piezoelectric sensors are used as the pressure sensing units; that is Figure 2 P1, P2, and P3 in

[0090] Two pH detection probes are symmetrically arranged above and below the impeller of the stirring paddle, and an automatic cleaning device is equipped; that is Figure 2 pH1 and pH2 in

[0091] All sensors are connected to the central control station through industrial Ethernet, the sampling frequency is uniformly set to 2Hz, and hardware-level synchronization is established with the stirring speed signal of the reactor;

[0092] Step 102: Synchronously obtain the process parameter time series corresponding to each process parameter from the sensor network through the Modbus-TCP protocol, and perform time domain alignment on the time of the process parameter time series using the IEEE 1588 precise time protocol to generate a raw data set with a unified timestamp to form multi-dimensional process parameter time series data; specifically including temperature time series, pressure time series, and pH time series;

[0093] Furthermore, the method for synchronously obtaining the raw material spectral characteristic data is as follows:

[0094] The raw material flow is scanned in real time by an on-line near-infrared spectrometer with a wavelength range of 900 - 1700nm, and the original spectral matrix is obtained in each sampling period, and the raw material flow rate and raw material temperature are synchronously collected;

[0095] The Savitzky-Golay filter is used to perform spectral denoising on the original spectral matrix to generate a standard spectral matrix. The window width of the Savitzky-Golay filter is 15 points, and the quadratic polynomial fitting method is applied for fitting;

[0096] The standard spectral matrix, raw material flow rate, and raw material temperature constitute the raw material spectral characteristic data.

[0097] In the embodiment of the present invention, the steps for performing dynamic adaptive filtering processing on the multi-dimensional process parameter time series data to generate a denoised process parameter sequence include the following:

[0098] Step 211: Construct the original multi-dimensional process parameter time-series data with unified timestamps into the form of a data matrix X(t).

[0099] Specifically, the formula form of X(t) is: , i = 1, 2, 3...I; where I is the number of process parameters in the collected multi-dimensional process parameter time-series data; i is the process parameter number. represents the measured value of the i-th parameter at time t; t is any moment when process parameter data is collected through the distributed sensor network.

[0100] Meanwhile, according to the process flow, pre-define each process stage in the production process of macrocyclic lactone derivatives, and mark the process stage number as k; specifically, in the production process of macrocyclic lactone derivatives, the process stage can be divided into raw material pretreatment stage, initiation and cyclization stage, side chain modification stage, post-treatment and purification stage, and stability enhancement stage. Among them, the raw material pretreatment stage is used for raw material activation and impurity control, the initiation and cyclization stage is used for macrocyclic skeleton construction, the side chain modification stage is used for directional introduction of functional groups, and the stability enhancement stage is used for product separation and refinement.

[0101] Step 212: After matrix construction, to enable the filter to have dynamic adaptive capabilities, it is necessary to estimate the noise characteristics of each parameter within a sliding time window; for this purpose, set the window length as Tw, and calculate the local mean of each process parameter within the window. and the local variance to form local noise statistical information. At the same time, define the local signal-to-noise ratio as a quantization index of the noise level.

[0102] It should be noted that combined with the characteristics of the current process stage, this step provides a basis for the dynamic adjustment of adaptive parameters in the subsequent filter, enabling the response speed and filtering intensity of the filter to vary flexibly in different stages (such as start-up, stable, or end stages).

[0103] Step 213: Design an adaptive filter model based on the local noise statistical information.

[0104] Specifically, use the adaptive Kalman filter as the adaptive filter model. For the i-th parameter, adopt the following state update equation:

[0105] ; where represents the denoising estimate value obtained after filtering. is the adaptive Kalman gain, and its calculation formula is: , where is the error covariance at the previous moment, and It is dynamically adjusted according to the noise variance calculated within the current sliding window and the k-th process stage. For example, it can be set as , where is the preset adjustment coefficient corresponding to the k-th process stage;

[0106] Step 214: After the dynamic adaptive filter model is constructed, perform point-by-point filtering on the multi-dimensional parameters at each moment t;

[0107] Specifically, at each sampling moment t, according to the current process stage and local noise statistical characteristics, the error covariance is updated in real time, and the Kalman gain is recalculated to ensure that the filter can quickly respond to sudden noise or process fluctuations;

[0108] It can be understood that this dynamic update mechanism ensures that the filter can automatically increase the smoothing strength during periods of high noise or at the start of the process, while allowing more responses to capture real signal changes when the process is stable.

[0109] The whole process can be recorded as:

[0110]

[0111] where is the process noise covariance, which is also dynamically set according to the process stage;

[0112] Step 215: Collect the time-series data sequence of the denoised process parameters after dynamic adaptive filtering processing for each process parameter, denoted as the denoised process parameter sequence.

[0113] In the embodiment of the present invention, the simultaneous baseline correction of the raw material spectral characteristic data to generate standard spectral data includes the following steps:

[0114] Step 221: Denote the standard spectral matrix as Sraw, where each column represents the spectral response within the wavelength range in one sampling period;

[0115] Step 222: Considering that in the preprocessed raw spectral data, due to factors such as instrument response, environmental temperature change, and raw material flow rate fluctuation, there is often non-ideal baseline drift. Therefore, first, for the spectral curve in each sampling period, extract the local minimum value of the wavelength;

[0116] Specifically, the local minimum value detection refers to selecting a window interval with a preset sliding window size within the wavelength interval, calculating the local minimum value within the window interval, and using it as the preliminary estimation point of the baseline within the window interval to initially capture the low-value trend in the spectral curve;

[0117] Step 223: Based on the extracted local minima, construct a baseline model for the entire spectral curve using a low-order polynomial fitting or an adaptive smoothing method;

[0118] Specifically, taking the polynomial fitting method as an example, set the baseline model as:

[0119] where n is the fitting order, is the wavelength, and the coefficients are solved by the least squares method, using the extracted local minima as the label values of the spectral curve at the corresponding wavelengths. An adaptive algorithm such as the adaptive iterative weighted least squares method or the ALS method is used to compensate for the non-linear drift caused by the raw material flow rate and temperature while ensuring smooth fitting of the low signal-to-noise ratio region; this process ensures that the generated baseline can fully reflect the background trend of the actual spectrum;

[0120] Step 224: Use the constructed baseline model to perform baseline correction on the standard spectral matrix;

[0121] Specifically, the method for performing baseline correction on the standard spectral matrix is:

[0122] Perform correction through the following correction formula:

[0123] , where is the spectral value after baseline calibration, is the spectral value corresponding to the wavelength in the standard spectral matrix, and is the lowest spectral value fitted at this wavelength, which can be understood as the background spectral value. Subtracting the background spectral value realizes baseline correction.

[0124] In the embodiments of the present invention, the process of fusing the denoising process parameter sequence with the standard spectral data to construct a process parameter feature matrix includes the following steps:

[0125] Step 311: Mark the collected denoising process parameter sequence as , and mark the collected standard spectral data as ;

[0126] Specifically, is the time-series data of multi-dimensional process parameters, in the form of:

[0127]

[0128] where each element corresponds to a process parameter, representing the denoising sequence of the corresponding process parameter at time t;

[0129] And the standard spectral data set is the standard spectral matrix within each sampling period p, and its form is:

[0130]

[0131] where N represents the number of wavelength points, Q is the number of sampling periods, and each element in this standard spectral matrix represents the standard spectral data value of the corresponding wavelength point within the corresponding sampling period;

[0132] Further preferably, since the time intervals between the denoising process parameter sequence and the standard spectral data may be different, it is also possible to first perform time alignment on the denoising process parameter sequence and the standard spectral data. Therefore, an interpolation algorithm such as linear interpolation or spline interpolation is used to perform time alignment on one of the data sets so that it is consistent with the sampling time points of the process parameters. After completion of the alignment, the obtained denoising process parameter sequence and standard spectral data will be aligned at the same time stamp;

[0133] As an example, in an actual reaction process, a set of denoising process parameter sequences is represented as:

[0134] ; each row corresponds to a time point (t = 1 to 5), which are temperature, pressure, and pH value in sequence;

[0135] The collected standard spectral data is represented as:

[0136] , which represents that the spectral data takes 4 points in the wavelength direction (wavelengths 1000, 1200, 1400, 1600 nm) and is collected at 5 sampling moments, forming a 5×4 matrix;

[0137] Step 312: Since both the denoising process parameter sequence and the standard spectral data may contain redundant or highly correlated features, in order to improve the fusion effect and reduce the computational burden, feature selection and dimensionality reduction processing are also required; for this purpose, correlation analysis is performed on each parameter in the denoising process parameter sequence to eliminate redundant highly correlated parameters, and for the spectral data, the principal components are extracted through a dimensionality reduction method;

[0138] Specifically, the correlation analysis uses the Pearson correlation coefficient or the mutual information method to quantify the correlation between each process parameter, and selects the main parameters that are highly correlated with the yield of the target product;

[0139] And the dimensionality reduction methods include, but are not limited to, dimensionality reduction methods such as principal component analysis (PCA), weighted average, or independent component analysis (ICA);

[0140] For the example in step 311, the process of dimensionality reduction is as follows:

[0141] In this example, since process parameters such as temperature, pressure, and pH are all key parameters, they are directly retained without dimensionality reduction.

[0142] For the standard spectral data, the original 4D spectral data usually has strong correlations. Dimensionality reduction is performed using simple weighted averaging, and the weight vector for the dimension is set to [0.25, 0.25, 0.25, 0.25].

[0143] For each sampling moment, the data at 4 wavelength points are projected into a scalar. For example, for the first moment, its spectral feature is 0.25×(0.80 + 0.75 + 0.70 + 0.65) = 0.25×2.90 = 0.725.

[0144] The dimensionality reduction results for each moment are calculated in sequence: 0.725, 0.745, 0.735, 0.755, 0.765.

[0145] Thus, a feature vector of dimension 5×1 is obtained.

[0146] Step 313: Concatenate the denoised process parameter sequence after dimensionality reduction and the standard spectral data along the time dimension to form a new fused feature matrix.

[0147] Specifically, in the feature process, each wavelength point of the standard spectral data is regarded as an independent feature, and each parameter of the denoised process parameter sequence is also taken as an independent feature. A weighted fusion method is used to configure the weights of the process parameters and spectral data, and each weight is set in advance.

[0148] For the example in step 312, the process of feature fusion is as follows:

[0149] Set the weight of each process parameter to 1.

[0150] At t = 1 moment, the process parameters are [80, 2.0, 7.0]; the spectral feature is 0.725, and the fused feature vector after fusion is [80, 2.0, 7.0, 0.725]. Similarly, for t = 2, 3, 4, 5 moments, the corresponding fused feature vectors are [82, 2.1, 7.1, 0.745], [83, 2.0, 7.0, 0.735], [84, 2.2, 7.2, 0.755], [85, 2.1, 7.3, 0.765] respectively.

[0151] The fused feature matrix formed by concatenating all fused feature vectors in chronological order is:

[0152] ;

[0153] Step 314: Standardize the fused feature matrix to obtain a process parameter feature matrix;

[0154] For the example of Step 313, the process parameter feature matrix obtained by standardization is approximately:

[0155] .

[0156] In the embodiments of the present invention, the improved regression model with input process knowledge constraints outputs a key parameter sensitivity vector and a product yield prediction value, including the following steps:

[0157] Step 321: Mark the process parameter feature vector of each row in the process parameter feature matrix as X(t); mark the target variable product yield as Y(t);

[0158] Wherein, where D represents the feature dimension obtained after fusion;

[0159] Step 322: Mark the improved regression model as , and introduce prior constraints for some key parameters according to the process experience of each process stage in the production process of macrolide acid derivatives;

[0160] The specific expression of the improved regression model is set as:

[0161] ; where is the intercept term, is the sensitivity coefficient of the d-th key parameter;

[0162] Specifically, the prior constraint can require a positive impact on some key parameters, that is, the corresponding >0, or set the influence range for some key parameters, that is, the corresponding is within a preset range;

[0163] Step 323: Pre-collect the process parameter feature vectors and the corresponding product yields at each moment in each process stage during several production processes of macrolide acid derivatives as a sample set;

[0164] Step 324: Use the process parameter feature vectors at each moment in the sample set as the input of the improved regression model, use the mean square error between the product yields at each moment in the sample set and the predicted values calculated by the improved regression model expression as the loss function, and use the prior constraints of each key parameter as inequality constraint conditions, and use the optimization algorithm with inequality constraints to solve the intercept term and sensitivity coefficient of the improved regression model;

[0165] Step 325: The sensitivity coefficients of the obtained key parameters form a key parameter sensitivity vector; for any new moment in a production process, input the process parameter feature vector at this new moment into the improved regression model to obtain the predicted product yield value.

[0166] In the embodiment of the present invention, the step of extracting the dominant parameter submatrix from the process parameter feature matrix based on the key parameter sensitivity vector includes the following steps:

[0167] Step 411: Mark the key parameter sensitivity vector as , calculate the absolute value | | of each sensitivity coefficient; where d = 1, 2,..., D;

[0168] It can be understood that | | characterizes the importance of the d-th key parameter and is a direct index for measuring the influence intensity of the d-th key parameter on the target variable;

[0169] Step 412: Determine the dominant parameter set through the threshold method or the sorting method;

[0170] Specifically, the threshold method refers to presetting an importance threshold, and forming a dominant parameter set by the key parameters whose absolute values are greater than the importance threshold;

[0171] The sorting method refers to sorting the absolute values | | of all sensitivity coefficients according to the numerical size, and selecting D' key parameters from largest to smallest to form a dominant parameter set, where D' is the preset number of dominant parameters;

[0172] Step 413: Extract the parameter values corresponding to each dominant parameter in the dominant parameter set from the process parameter feature matrix to form a dominant parameter submatrix;

[0173] It can be understood that through the extraction of the above-mentioned dominant parameter submatrix, a sensitivity vector is obtained from model training, and the absolute influence index of each parameter is calculated, determining the dominant parameter index that needs to be concerned, greatly reducing the dimension of the process parameter feature matrix.

[0174] Further, the method of constructing a time-varying state space model by using the process time series alignment algorithm is as follows:

[0175] Select the dynamic time warping algorithm as the process time series alignment algorithm;

[0176] Select a dominant parameter time series template that is stable and represents the normal process state from historical data;

[0177] Obtain the current real-time dominant parameter data sequence;

[0178] Use the dynamic time warping algorithm to compare two time series data, and then construct a cost matrix to calculate the matching cost between the dominant parameter time series template and the dominant parameter data sequence, so as to find the optimal alignment path π to achieve non-linear correction on the time axis;

[0179] It can be understood that through the above non-linear correction, each moment j in the real-time sequence can be mapped to a certain moment i in the reference sequence template, so as to eliminate the time distortion that appears in the process;

[0180] Use the dominant parameter data aligned by the dynamic time warping algorithm to construct a time-varying state space model that describes the dynamic evolution of the process;

[0181] Specifically, the form of the constructed time-varying state space model that describes the dynamic evolution of the process is:

[0182]

[0183] Among them, x(t) represents the state vector, and its value is the value of the aligned dominant parameter sub-matrix at time t. u(t) is a controllable input such as adjustment parameters and compensation variables. y(t) is the observed output, corresponding to the key process indicators measured in real time, such as the predicted value of product yield. A(t), B(t), and C(t) are time-varying system matrices, which reflect the dynamic characteristics of the process in different time periods; w(t) and v(t) are process noise and measurement noise respectively;

[0184] Finally, through the system identification method of parameter update using the recursive least squares method or the Kalman filter for historical aligned data, the above system matrices can be updated online in real time, so that the state space model can dynamically reflect the current process state and evolution trend;

[0185] It should be noted that the constructed time-varying state space model is used to capture the dynamic change law of the process. By comparing the deviation between the real-time predicted product yield data and the product yield in the template, process anomalies or process drifts can be detected in time, so that production correction instructions can be generated in time to adjust the controllable parameters in the production process in time.

[0186] In the embodiment of the present invention, the method of dynamically comparing the time-varying state space model with the predicted value of product yield is:

[0187] Input the real-time process parameter feature matrix into the improved regression model to obtain the predicted value of the product yield;

[0188] Input the real-time dominant parameter sub-matrix into the time-varying state space model to obtain the expected yield prediction value output by the time-varying state space model;

[0189] Compare the difference between the predicted value of the by-product yield and the predicted value of the expected yield, and determine whether the difference is greater than a preset difference threshold;

[0190] It can be understood that the improved regression model makes predictions based on the parameter values of the current process parameters, while the time-varying state space model makes predictions based on the empirical template of the yield in the historical production process. Generally speaking, the yield value predicted by the improved regression model is based on data fitting for the entire current production cycle and is the predicted future yield, while the time-varying state space model makes predictions based on the historical template, so the predicted value is an expected value, that is, the yield that is expected to be achieved when producing according to the current production parameters. If the predicted values of the two differ greatly, it indicates that there may be some abnormalities in the current production parameters and adjustments are needed.

[0191] In the embodiment of the present invention, the method for generating the optimized control parameter set by the multi-objective optimization model is as follows:

[0192] Mark the difference threshold between the predicted value of the by-product yield and the predicted value of the expected yield as h;

[0193] Mark the predicted value of the expected yield as H;

[0194] Construct a multi-objective optimization function Yu;

[0195] where the expression of Yu is , is the parameter variable to be solved for the d-th key parameter, that is, it represents the adjusted parameter value, is the parameter value of the current d-th key parameter; it can be understood that, measures the difference between the predicted value of the by-product yield and the predicted value of the expected yield. When is 0, it means that the current process parameters best meet the expectations, measures the adjustment range of each key parameter; where, g0 and gd are both preset proportionality coefficients;

[0196] Construct a set of constraint conditions U;

[0197] Specifically, the set of constraint conditions U includes at least < h, indicating that after the key parameters are adjusted, the difference between the predicted yield and the predicted value of the expected yield is less than the difference threshold, so as to meet the expectations;

[0198] Further, the set of constraint conditions U may further include the allowable range and the change range of each key parameter; the allowable range is the range in which each key parameter should be during the production process. For example, the temperature should be maintained within a certain range, and the change range refers to that the difference between the parameter value of each key parameter after adjustment and the current parameter value should be within a certain range, so as to reduce the amplitude of parameter change and avoid drastic changes in the environment inside the reactor, resulting in the destruction of raw materials;

[0199] Taking the minimization of the multi-objective optimization function as the optimization goal and the set of constraint conditions as the constraints, a multi-objective optimization model is constructed;

[0200] By using an optimization algorithm to solve the multi-objective optimization model, the solution values of the parameter variables to be solved for each key parameter are obtained, and the solution values of all parameter variables constitute the optimized control parameter set;

[0201] And inputting the optimized control parameter set into the adaptive controller to generate an execution instruction means taking the solution values of the parameter variables to be solved for each key parameter as control instructions, and adjusting the parameter values of each key parameter to the corresponding solution values by the production program background, so as to improve the matching degree between the yield of macrolide acid derivatives and the expected yield.

[0202] Embodiment 2

[0203] As Figure 3 shown, the automatic precision control system for the production process of macrolide acid derivatives includes an original data collection module, a preprocessing module, a yield prediction module, a state space model training module, and a control parameter optimization module; among them, each module is connected electrically.

[0204] The original data collection module collects the time-series data of multi-dimensional process parameters of the reactor in real time through a distributed sensing system, synchronously obtains the raw material spectral feature data, and sends the time-series data of multi-dimensional process parameters and the raw material spectral feature data to the preprocessing module;

[0205] The preprocessing module performs dynamic adaptive filtering on the time-series data of multi-dimensional process parameters to generate a denoised process parameter sequence, and at the same time performs baseline correction on the raw material spectral feature data to generate standard spectral data, and sends the denoised process parameter sequence and the standard spectral data to the yield prediction module;

[0206] The yield prediction module fuses the features of the denoised process parameter sequence and the standard spectral data, constructs a process parameter feature matrix, inputs an improved regression model constrained by process knowledge, outputs a key parameter sensitivity vector and a predicted value of the product yield, and sends the key parameter sensitivity vector to the state space model training module, and sends the improved regression model and the predicted value of the product yield to the control parameter optimization module;

[0207] The state space model training module extracts the dominant parameter sub-matrix from the process parameter feature matrix based on the key parameter sensitivity vector, constructs a time-varying state space model using the process time series alignment algorithm, and sends the time-varying state space model to the control parameter optimization module;

[0208] The control parameter optimization module dynamically compares the time-varying state space model with the predicted product yield value. When the deviation exceeds the preset threshold, it generates an optimized control parameter set through a multi-objective optimization model, inputs the optimized control parameter set into an adaptive controller to generate an execution instruction to control the process parameters.

[0209] In addition, parts of the above technical solutions provided in the embodiments of the present application that are consistent with the corresponding technical solutions in the prior art in terms of implementation principles are not described in detail to avoid excessive elaboration.

[0210] As described above in the specific embodiments, the purpose, technical solutions, and beneficial effects of the present invention are further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

[0211] The above preset parameters or preset thresholds are all set by those skilled in the art according to the actual situation or obtained through a large number of data simulations.

[0212] The above embodiments are only used to illustrate the technical methods of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical methods of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for automated and precise control of the production process of macrolide derivatives, characterized in that: The following steps are involved: Step 1: Use the distributed sensing system to collect the multi-dimensional process parameter time series data of the reactor in real time and simultaneously obtain the spectral characteristic data of the raw materials; Multidimensional process parameters include temperature, pressure, and pH time series; Step 2: Implement dynamic adaptive filtering on the multi-dimensional process parameter time series data to generate a denoised process parameter sequence, and perform baseline correction on the raw material spectral feature data to generate standard spectral data; Step 3: Feature fusion of the denoised process parameter sequence and the standard spectral data, constructing the process parameter feature matrix through dimensionality reduction and weighted fusion, inputting the improved regression model constrained by process knowledge, and outputting the key parameter sensitivity vector and product yield prediction value; The improved regression model constrained by process knowledge includes: training the improved regression model based on the fusion feature matrix and product yield label data through prior constraints; Step 4: Based on the key parameter sensitivity vector, the dominant parameter sub-matrix is ​​extracted from the process parameter feature matrix, and the time-varying state space model is constructed by using the process timing alignment algorithm; the construction of the time-varying state space model includes: determining the dominant parameter set by a threshold method or a sorting method, extracting the parameter value corresponding to each dominant parameter in the dominant parameter set from the process parameter feature matrix, and forming a dominant parameter sub-matrix; constructing the time-varying state space model; Step 5: dynamically compare the product yield prediction value output by the time-varying state space model with the expected yield prediction value, and when the deviation exceeds a preset threshold, generate an optimization control parameter set through a multi-objective optimization model; including: constructing a multi-objective optimization function Yu based on the expected yield prediction value and the improved regression model; the multi-objective optimization function Yu is a weighted sum of the difference between the product yield prediction value and the expected yield prediction value and the adjustment range of various key parameters; Based on the expected yield prediction value, the improved regression model and the difference threshold, a constraint condition set U is constructed; Taking minimization of multi-objective optimization function as optimization objective and constraint condition set as constraint condition, a multi-objective optimization model is constructed. By using an optimization algorithm to solve the multi-objective optimization model, the solution values ​​of the parameter variables to be solved for each key parameter are obtained, and the solution values ​​of all parameter variables constitute an optimization control parameter set; Step 6: Input the optimized control parameter set into the adaptive controller to generate execution instructions to control the process parameters.

2. The method for automated and precise control of the production process of macrolide derivatives according to claim 1, characterized in that: The real-time acquisition of multi-dimensional process parameter time series data of the reactor by the distributed sensing system comprises the following steps: Step 101: Arrange a distributed sensor network for collecting various process parameters at key process nodes of the reactor; Step 102: synchronously obtain the process parameter time series corresponding to each process parameter from the sensor network, and use the time protocol to time-domain align the time of the process parameter time series to generate an original data set with a unified timestamp to form multi-dimensional process parameter time series data.

3. The method for automated and precise control of the production process of macrolide derivatives according to claim 2, characterized in that: The method of synchronously acquiring the raw material spectral characteristic data is: The raw material flow is scanned in real time by an online near-infrared spectrometer with a wavelength range of 900-1700 nm, and the raw spectrum matrix is ​​obtained in each sampling period, and the raw material flow rate and raw material temperature are collected synchronously; The Savitzky-Golay filter is used to perform spectral denoising on the original spectral matrix to generate a standard spectral matrix; The standard spectrum matrix, raw material flow rate and raw material temperature constitute the raw material spectrum characteristic data.

4. The method for automated and precise control of the production process of macrolide derivatives according to claim 3, characterized in that: The method of implementing dynamic adaptive filtering on the multi-dimensional process parameter time series data to generate a denoising process parameter sequence comprises the following steps: Step 211: constructing the original multi-dimensional process parameter time series data with a unified time stamp into the form of a data matrix X(t); Step 212: Set the window length to Tw and calculate the local mean of each process parameter within the window. With local variance Constitute the local noise statistics and define the local signal-to-noise ratio As a quantitative indicator of noise level; Step 213: designing an adaptive filter model based on local noise statistical information; Step 214: after the dynamic adaptive filter model is constructed, point-by-point filtering is performed on the multi-dimensional parameters at each time t; Step 215: Collect the time series data sequence after the denoising of each process parameter after dynamic adaptive filtering processing, and express it as a denoising process parameter sequence.

5. The method for automated and precise control of the production process of macrolide derivatives according to claim 4, characterized in that: The method of simultaneously performing baseline correction on the raw material spectral characteristic data to generate standard spectral data comprises the following steps: Step 221: record the standard spectrum matrix as Sraw, where each column represents the spectrum response within a wavelength range within a sampling period; Step 222: first, extract the local minimum value of the wavelength for the spectrum curve within each sampling period; Step 223: Based on the extracted local minimum, a baseline model of the entire spectral curve is constructed using a low-order polynomial fitting or an adaptive smoothing method; Step 224: Using the constructed baseline model, perform baseline correction on the standard spectrum matrix.

6. The method for automated and precise control of the production process of macrolide derivatives according to claim 5, characterized in that: The process of fusing the denoising process parameter sequence with the standard spectral data to construct a process parameter feature matrix includes the following steps: Step 311: Mark the collected denoising process parameter sequence as , the collected standard spectral data are marked as ; Step 312: performing correlation analysis on each parameter in the denoising process parameter sequence, removing redundant highly correlated parameters, and extracting the main components of the spectral data by using a dimensionality reduction method; Step 313: splicing the denoising process parameter sequence and the standard spectrum data after dimension reduction along the time dimension to form a new fusion feature matrix; Step 314: Standardize the fusion feature matrix generated after fusion to obtain a process parameter feature matrix.

7. The method for automated and precise control of the production process of macrolide derivatives according to claim 6, characterized in that: The improved regression model with input process knowledge constraints and outputting key parameter sensitivity vectors and product yield prediction values ​​comprises the following steps: Step 321: label the process parameter feature vector of each row in the process parameter feature matrix as X(t); label the target variable product yield as Y(t); Step 322: Label the improved regression model as , and introduce a priori constraints for some key parameters based on the process experience of each process stage in the production of macrolide derivatives; Step 323: pre-collecting several process parameter feature vectors and corresponding product yields at various moments in various process stages during the production of macrolide derivatives as a sample set; Step 324: using the process parameter feature vectors at each moment in the sample set as the input of the improved regression model, using the mean square error between the product yield corresponding to each moment in the sample set and the predicted value calculated by the improved regression model expression as the loss function, using the prior constraints of various key parameters as inequality constraints, and using the inequality constraint optimization algorithm to solve the intercept term and sensitivity coefficient of the improved regression model; Step 325: The sensitivity coefficients of the solved key parameters are used to form a key parameter sensitivity vector; for any new moment in the production process, the process parameter characteristic vector of the new moment is input into the improved regression model to obtain a predicted value of the product yield.

8. The method for automated and precise control of the production process of macrolide derivatives according to claim 7, characterized in that: The method of constructing the time-varying state space model using the process timing alignment algorithm is as follows: The dynamic time warping algorithm is selected as the process timing alignment algorithm; Select a dominant parameter timing template that is stable and represents a normal process state from historical data; Get the current real-time dominant parameter data sequence; The dynamic time warping algorithm is used to compare the two time series data, and then the cost matrix is ​​constructed to calculate the matching cost between the dominant parameter time series template and the dominant parameter data sequence to find the optimal alignment path π to achieve nonlinear correction on the time axis; The time-varying state-space model describing the dynamic evolution of the process is constructed using the dominant parameter data aligned by the dynamic time warping algorithm.

9. The method for automated and precise control of the production process of macrolide derivatives according to claim 8, characterized in that: The method of dynamically comparing the time-varying state space model with the product yield prediction value is: Inputting a real-time process parameter characteristic matrix into the improved regression model to obtain a predicted product yield prediction value; Inputting a real-time dominant parameter sub-matrix into the time-varying state-space model to obtain an expected yield prediction value output by the time-varying state-space model; Compare the difference between the product yield prediction value and the expected yield prediction value to determine whether the difference is greater than a preset difference threshold.

10. An automated and precise control system for the production process of macrolide derivatives, which is used to implement the automated and precise control method for the production process of macrolide derivatives according to any one of claims 1 to 9, characterized in that: It includes a raw data collection module, a preprocessing module, a yield prediction module, a state space model training module, and a control parameter optimization module; wherein each module is electrically connected; The raw data collection module collects the multi-dimensional process parameter time series data of the reactor in real time through the distributed sensing system, synchronously obtains the raw material spectrum characteristic data, and sends the multi-dimensional process parameter time series data and the raw material spectrum characteristic data to the preprocessing module; The preprocessing module implements dynamic adaptive filtering processing on the multi-dimensional process parameter time series data to generate a denoising process parameter sequence, and at the same time performs baseline correction on the raw material spectral feature data to generate standard spectral data, and sends the denoising process parameter sequence and the standard spectral data to the yield prediction module; The yield prediction module performs feature fusion of the denoised process parameter sequence and the standard spectral data, constructs a process parameter feature matrix, inputs an improved regression model constrained by process knowledge, outputs key parameter sensitivity vectors and product yield prediction values, and sends the key parameter sensitivity vectors to the state space model training module, and sends the improved regression model and product yield prediction values ​​to the control parameter optimization module; The state space model training module extracts the dominant parameter sub-matrix from the process parameter feature matrix based on the key parameter sensitivity vector, constructs the time-varying state space model using the process timing alignment algorithm, and sends the time-varying state space model to the control parameter optimization module; The control parameter optimization module dynamically compares the time-varying state space model with the product yield prediction value. When the deviation exceeds the preset threshold, the multi-objective optimization model generates an optimized control parameter set, which is then input into the adaptive controller to generate execution instructions to control the process parameters.

Images