Sugar juice color value online detection method and system and process parameter control method and system

Abstract

The present application relates to the technical field of sugar production, and particularly relates to a sugar juice color value online detection method and system and a process parameter control method and system. The method collects sugar juice color values and key process parameters in real time and constructs a digital feature vector; a dynamic coupling relationship model is trained using historical and real-time data, a baseline color value evolution trajectory is predicted, and a backup collaborative adjustment scheme library covering multiple process scenarios is generated in advance; real-time multi-step trend prediction is performed on the color value, when the prediction deviates from the optimal range, the scheme library is queried in real time according to the deviation characteristics, and a pre-adjustment scheme with the minimum comprehensive process cost is quickly calculated, the pre-adjustment scheme is analyzed and converted into feedforward collaborative adjustment instructions for the auxiliary material adding system and each process execution unit, active and accurate closed-loop control is realized, the lag problem of traditional offline detection and manual control is effectively solved, the predictability and collaboration of color value control are improved, and product quality stability and production efficiency are ensured.

Description

Technical Field

[0001] This invention relates to the field of sugar production technology, and in particular to a method and system for online detection of sugar juice color value and control of its process parameters. Background Technology

[0002] In the sugar industry, the color value of sugar juice is a key indicator for evaluating the quality of finished sugar, directly affecting the product's grade and market value. Traditional color value detection methods mainly rely on manual, timed sampling and offline laboratory analysis, which suffers from significant lag and cannot reflect color value changes during the production process in real time. This leads to untimely process adjustments and large fluctuations in product quality. Furthermore, there is a lack of closed-loop control methods that can directly and automatically correlate and adjust key process parameters (such as sulfur fumigation intensity, pH value, and heating temperature) based on color value detection results. This means that the production process still relies excessively on manual experience, resulting in insufficient control precision and stability, which restricts the uniformity of sugar quality and further improvement of production efficiency. Therefore, there is an urgent need for a control system that can achieve rapid online detection of sugar juice color value and optimize process parameters in real time based on this information. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides a method and system for online detection of sugar juice color value and control of its process parameters.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention discloses an online detection method for the color value of sugar syrup and a method for controlling its process parameters, comprising the following steps: The system collects sugar juice color value data from multiple preset sampling points during the sugar-making process in real time, and simultaneously acquires the corresponding auxiliary material addition parameters and key process parameters, which together form a digital feature vector reflecting the current process status. Based on historical and real-time data training models, and using the digital feature vectors as input, a dynamic coupling relationship model between the evolution of sugar juice color value and multiple process parameters is analyzed and established. Based on the dynamic coupling relationship model, a library of alternative collaborative adjustment schemes for covering multiple process scenarios is constructed. The dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value. When the trend prediction trigger signal indicates that the prediction trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time according to the deviation characteristics, and the pre-adjustment scheme that makes the color value return to the target range and minimizes the overall process cost is optimized and calculated online. According to the pre-adjustment scheme, feedforward collaborative adjustment and control are carried out on the auxiliary material addition system and key process execution units in the production process.

[0005] Preferably, the color value data of sugar juice at multiple preset sampling points during the sugar-making process are collected in real time, and the corresponding auxiliary material addition parameters and key process parameters are acquired simultaneously, together forming a digital feature vector reflecting the current process status, specifically: The real-time color value readings output by the online color value detection device at multiple preset sampling points are obtained to form an original color value data sequence; Meanwhile, the instantaneous values ​​of the instantaneous addition rate of auxiliary materials and key process parameters corresponding to the process positions of each sampling point are extracted in real time from the sugar production process control system to form the original process parameter data set; The original color value data sequence and the original process parameter data set are aligned and bound according to a unified time base to generate a synchronous dataset; The color value data in the synchronous dataset is filtered and denoised to obtain a stable color value sequence, and the validity of the process parameter data is verified to obtain a set of valid process parameters. The stable color value sequence and the set of effective process parameters are normalized and spliced ​​according to a preset dimension and order to form a digital feature vector representing the overall process state at the current moment.

[0006] Preferably, based on historical and real-time data, a model is trained, and using the digital feature vector as input, a dynamic coupling relationship model between the evolution of syrup color value and multiple process parameters is analyzed and established, specifically as follows: The digital feature vectors of continuous time series are extracted from historical databases and real-time data streams and arranged in process order to form a process state time series segment. Time lag correlation analysis is performed on the time series segments of the process state to calculate the correlation strength between color value data and each process parameter at different time lags, and a lag correlation strength map is formed. The correlation strength of each item in the hysteresis correlation strength spectrum is compared with a preset correlation threshold. Process parameters whose correlation strength exceeds the preset correlation threshold and their corresponding time lags are screened out, thereby defining the key hysteresis parameter set and their respective time lag intervals. Based on the set of key lag parameters and their time lag intervals, the corresponding parameter lag feature sub-vectors are extracted and reconstructed from the time series segments of the process state. The parameter lag feature vector is combined with the current time-time digital feature vector, and a multi-parameter interaction effect term representing the nonlinear interaction between parameters is introduced to jointly form an extended coupled feature set for model training. Using the extended coupling feature set as input and the color value change in a specific future time period as output target, the preset model structure is trained and verified to obtain a dynamic coupling weight matrix that can quantify the influence of multi-parameter coupling on color value. The dynamic coupling weight matrix is ​​combined with the model structure to encapsulate a dynamic coupling relationship model.

[0007] Preferably, a library of alternative collaborative adjustment schemes covering multiple process scenarios is constructed based on a dynamic coupling relationship model, specifically as follows: Using the set of historical and real-time digital feature vectors covering typical process ranges as input, the dynamic coupling relationship model is used to perform forward iterative calculations to predict the evolution trajectory of the reference color value under each process scenario. The current values ​​of adjustable process parameters contained in the digital feature vectors corresponding to each process scenario are analyzed, and combined with the predefined parameter safety operating limits, the feasible adjustment space of each adjustable process parameter is determined. Within the feasible adjustment space of each process scenario, a series of discrete simulation adjustment quantities are systematically generated and combined according to the preset sampling rules to form multiple comprehensive simulation intervention schemes under the corresponding process scenario. Substitute the simulated intervention schemes under all process scenarios into the dynamic coupling relationship model for recalculation to obtain the corresponding simulated color value response trajectory; For each simulated intervention scheme, the corresponding simulated color value response trajectory is compared with the baseline color value evolution trajectory under the corresponding process scenario to determine the control effectiveness evaluation value of the simulated intervention scheme in improving and maintaining the color value trajectory within the preset optimal range, as well as the process disturbance index caused by implementing the simulated intervention scheme. All simulated intervention schemes under all process scenarios, along with their corresponding simulated color value response trajectories, process disturbance indices, and control performance evaluation values, are associated, stored, and indexed to form the alternative collaborative adjustment scheme library.

[0008] Preferably, the dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value, specifically as follows: The latest digital feature vector is obtained as input to the dynamic coupling relationship model, and a dynamic coupling weight matrix matching the current process state is extracted from the dynamic coupling relationship model. The weighted summation of each component in the latest digital feature vector with the corresponding weight coefficient in the extracted dynamic coupling weight matrix is ​​performed to obtain the first color value prediction point for the next unit process cycle. Based on the first color value prediction point, and combined with the process state retention assumption derived from the digital feature vector, the prediction state vector for the next moment is updated and rolled forward. The predicted state vector is used as a new input and operated on with the dynamic coupling weight matrix again to obtain the second color value prediction point for subsequent unit process cycles. Repeat the above state rolling and model calculation steps, and iterate multiple times within the preset state inference window to obtain a future multi-step prediction trajectory composed of multiple ordered prediction points. The smoothness test and confidence assessment are performed on the multi-step predicted trajectory to generate the corresponding prediction confidence interval; The multi-step predicted trajectory and its prediction confidence interval are compared with the preset optimal color value range in real time, and a trend prediction trigger signal is output to determine whether to start control intervention.

[0009] Preferably, when the trend prediction trigger signal indicates that the prediction trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time based on the deviation characteristics, and the pre-adjustment scheme that minimizes the overall process cost and returns the color value to the target range is calculated online. Specifically: Based on the relationship between the multi-step prediction trajectory and the boundary position of the optimal color value range, the direction, magnitude, and urgency of the color value deviation are determined, and a deviation feature description is formed. Using the aforementioned deviation characteristics as the core query vector, the real-time matching degree between the core query vector and the initial process state vector corresponding to each simulated intervention scheme stored in the alternative collaborative adjustment scheme library is calculated. Based on the aforementioned urgency, a matching degree threshold is set to initially select a subset of scenario matching schemes. From the subset of scenario matching schemes, the simulated color value response trajectory, process disturbance index and control performance evaluation value of each scheme are extracted. The current color value deviation is combined with the magnitude and direction for weighted correction to generate the estimated control performance value and estimated comprehensive process cost value of each scheme for the current real-time deviation. A real-time multi-objective optimization function is constructed with the urgency of the current color value deviation as a weight factor. The real-time multi-objective optimization function integrates the estimated control efficiency value and the estimated comprehensive process cost value, and performs online re-evaluation and ranking of all schemes in the scenario matching scheme subset. Based on the sorting results and the preset decision threshold, the final optimal adjustment scheme is selected or integrated, and the optimal adjustment scheme is coded for feasibility. The output is a pre-adjustment scheme that includes specific auxiliary material and key process parameter adjustment instructions, timing logic and expected effects.

[0010] Preferably, according to the pre-adjustment scheme, feedforward collaborative adjustment and control are performed on the auxiliary material addition system and key process execution units in the production process, specifically as follows: The instruction sequence encoded in the pre-adjustment scheme is analyzed to extract the types of auxiliary materials to be adjusted, the target adjustment amounts of each key process parameter, and the corresponding process execution unit identifiers. Based on the equipment response characteristics and process timing constraints of each process execution unit, the target adjustment amount is decoupled and allocated in the time dimension, and arranged into an equipment instruction execution plan that includes specific execution time points and execution order; Based on the equipment instruction execution plan and combined with the real-time acquired process cycle synchronization signal, the adjustment amount in the equipment instruction execution plan is converted into physical control instructions that can be recognized by the corresponding execution unit controller. Before the planned execution time arrives, corresponding physical control commands are sent to the controllers of the relevant auxiliary material addition systems and process execution units in advance, and the equipment command reception confirmation signals returned by each controller are received. Within the preset collaborative execution window, the device command completion status returned by each execution unit is monitored. When all necessary units have reported completion of execution, a collaborative adjustment completion confirmation signal is generated, thus completing the feedforward collaborative adjustment control.

[0011] Specifically, the process cycle synchronization signal acquired in real time is obtained by: capturing key equipment status signals that characterize process stage switching or material flow in real time based on preset key physical event trigger points in the sugar production line, and aligning the key equipment status signals with the central control clock to generate a unified process cycle synchronization signal.

[0012] The second aspect of the present invention discloses an online detection and process parameter control system for sugar syrup color value. The system includes a memory and a processor. The memory stores a program for online detection and process parameter control of sugar syrup color value. When the program for online detection and process parameter control of sugar syrup color value is executed by the processor, the steps of the online detection and process parameter control method for sugar syrup color value described in any one of the present invention are implemented.

[0013] This invention addresses the technical deficiencies in the prior art and offers the following advantages: By constructing a dynamic coupling relationship model and a library of alternative collaborative adjustment schemes, and by performing real-time trend prediction and online optimization decisions, it achieves advanced perception and accurate prediction of changes in sugar juice color value during the sugar-making process, overcoming the lag of traditional offline detection. Through feedforward collaborative regulation and control, the optimized pre-adjustment scheme is timely and collaboratively transformed into precise linkage control commands for multiple key process execution units (such as auxiliary material addition, pH adjustment, and temperature control), thereby proactively intervening before the color value actually deviates. This transforms passive post-processing into proactive forward control, improving the real-time performance, predictability, and synergy of color value control, effectively suppressing product quality fluctuations. While ensuring the stable color of the finished sugar meets high-quality standards, it reduces auxiliary material waste and energy consumption caused by delayed or uncoordinated process adjustments, ultimately achieving synergistic optimization of the sugar-making process's quality and efficiency. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0015] Figure 1 A flowchart of a method for online detection of color value of sugar juice and control of its process parameters; Figure 2 This is a structural diagram of an online detection system for the color value of sugar juice and its process parameters. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] like Figure 1 As shown, the first aspect of this invention discloses an online detection method for the color value of sugar syrup and a method for controlling its process parameters, comprising the following steps: The system collects sugar juice color value data from multiple preset sampling points during the sugar-making process in real time, and simultaneously acquires the corresponding auxiliary material addition parameters and key process parameters, which together form a digital feature vector reflecting the current process status. Based on historical and real-time data training models, and using the digital feature vectors as input, a dynamic coupling relationship model between the evolution of sugar juice color value and multiple process parameters is analyzed and established. Based on the dynamic coupling relationship model, a library of alternative collaborative adjustment schemes for covering multiple process scenarios is constructed. The dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value. When the trend prediction trigger signal indicates that the prediction trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time according to the deviation characteristics, and the pre-adjustment scheme that makes the color value return to the target range and minimizes the overall process cost is optimized and calculated online. According to the pre-adjustment scheme, feedforward collaborative adjustment and control are carried out on the auxiliary material addition system and key process execution units in the production process.

[0019] Preferably, the color value data of sugar juice at multiple preset sampling points during the sugar-making process are collected in real time, and the corresponding auxiliary material addition parameters and key process parameters are acquired simultaneously, together forming a digital feature vector reflecting the current process status, specifically: The real-time color value readings output by the online color value detection device at multiple preset sampling points are obtained to form an original color value data sequence; Meanwhile, the instantaneous values ​​of the instantaneous addition rate of auxiliary materials and key process parameters corresponding to the process positions of each sampling point are extracted in real time from the sugar production process control system to form the original process parameter data set; The original color value data sequence and the original process parameter data set are aligned and bound according to a unified time base to generate a synchronous dataset; The color value data in the synchronous dataset is filtered and denoised to obtain a stable color value sequence, and the validity of the process parameter data is verified to obtain a set of valid process parameters. The stable color value sequence and the set of effective process parameters are normalized and spliced ​​according to a preset dimension and order to form a digital feature vector representing the overall process state at the current moment.

[0020] It should be noted that the process involves acquiring real-time readings from the color value detection device to form a raw color value data sequence. Simultaneously, instantaneous values ​​of the corresponding auxiliary materials (such as sulfur dioxide) and other key process parameters (such as pH and temperature) are extracted from the process control system to form a raw process parameter set. The collected raw data is then time-aligned and bound using a unified time base. The synchronized color value data is then filtered and noise-reduced to eliminate random interference, resulting in a stable and reliable color value sequence. The process parameters are then validated to remove outliers, yielding a valid parameter set. The processed stable color value sequence and valid process parameter set are standardized according to a predefined dimensional order and then concatenated into a structured multidimensional data vector, i.e., a digital feature vector. This digital feature vector encapsulates the synchronization status of the system's key quality indicators and control variables at the current moment.

[0021] Preferably, based on historical and real-time data, a model is trained, and using the digital feature vector as input, a dynamic coupling relationship model between the evolution of syrup color value and multiple process parameters is analyzed and established, specifically as follows: The digital feature vectors of continuous time series are extracted from historical databases and real-time data streams and arranged in process order to form a process state time series segment. Time lag correlation analysis is performed on the time series segments of the process state to calculate the correlation strength between color value data and each process parameter at different time lags, and a lag correlation strength map is formed. The correlation strength of each item in the hysteresis correlation strength spectrum is compared with a preset correlation threshold. Process parameters whose correlation strength exceeds the preset correlation threshold and their corresponding time lags are screened out, thereby defining the key hysteresis parameter set and their respective time lag intervals. Based on the set of key lag parameters and their time lag intervals, the corresponding parameter lag feature sub-vectors are extracted and reconstructed from the time series segments of the process state. The parameter lag feature vector is combined with the current time-time digital feature vector, and a multi-parameter interaction effect term representing the nonlinear interaction between parameters is introduced to jointly form an extended coupled feature set for model training. Using the extended coupling feature set as input and the color value change in a specific future time period as output target, the preset model structure is trained and verified to obtain a dynamic coupling weight matrix that can quantify the influence of multi-parameter coupling on color value. The dynamic coupling weight matrix is ​​combined with the model structure to encapsulate a dynamic coupling relationship model.

[0022] It should be noted that the extended coupling feature set is paired with its corresponding measured color value changes over a specific future time period (e.g., the next 15 minutes) to form a supervised learning sample set, which is then divided into a training set and a validation set according to a preset ratio (e.g., 7:3). A preset data-driven model structure with learning and expressive capabilities is initialized, such as a fully connected feedforward neural network containing an input layer, at least one hidden layer, and an output layer, or a gradient boosting decision tree model containing multiple regression trees. Then, using the training set as input, the Adam optimizer is employed to iteratively update the model's internal parameters with the optimization objective of minimizing the error (e.g., mean squared error) between the model's predicted color value changes and the measured values. During or after training, the validation set is used to evaluate the model's prediction accuracy and generalization ability, and overfitting is prevented by adjusting hyperparameters (including the number of network layers, neurons, and the number and depth of trees), ensuring that the model performance meets the preset requirements. The learned quantitative relationship of the model, which maps input features (including current values, lag values, and interaction effect terms) to color value changes, is manifested as a series of parameter weights that can be extracted and fixed in the model. For linear models or specific layers of neural networks, these weights can be directly represented as matrix form. For tree models, they can be equivalently represented as a set of feature importance and split points. This invention refers to this set of parameters that can quantify the influence of multi-parameter coupling on color value as the dynamic coupling weight matrix, and combines it with the model's computational logic to encapsulate it into a dynamic coupling relationship model that can be deployed online.

[0023] It should be noted that continuous time-series digital feature vectors are extracted from historical databases and real-time data streams and arranged according to the process sequence to form process state time series segments. Time lag correlation analysis is performed on these segments, for example, using mutual information or time-lag correlation coefficients, to calculate the correlation strength between color values ​​and various process parameters at different lag times, forming a lag correlation strength map. As an example, the preset correlation threshold for selecting key parameters can be set to a mutual information value greater than 0.3 or an absolute correlation coefficient greater than 0.6. Based on this, parameters such as sulfur fumigation intensity (lag of 5-10 minutes) and pH value (lag of 0-5 minutes) can be selected as key lag parameters, and their respective time lag intervals are defined. According to the key parameters and their time lag intervals, the corresponding parameter lag feature sub-vectors are extracted and reconstructed from the original time series, for example, containing sulfur fumigation intensity values ​​at t-5 minutes, t-6 minutes…t-10 minutes.

[0024] Furthermore, when constructing the model input features, in addition to the current time-instance digital feature vector and the aforementioned lagged feature sub-vectors, a multi-parameter interaction effect term representing the nonlinear interaction between parameters is introduced through mathematical construction. This multi-parameter interaction effect term is defined as a product of selected key parameters, or between the parameters and their lagged values, or a normalized polynomial combination term, used to quantify the synergistic or antagonistic effects between parameters. For example, an interaction effect term could be the product of "current sulfur fumigation intensity" and "pH value 5 minutes ago," used to characterize the joint influence of these two parameters on color value changes under a specific time-series relationship. All basic features are combined with the constructed interaction effect terms to form an extended coupled feature set for model training. Finally, using the extended coupled feature set as input and the color value change in a specific future time period (e.g., the next 15 minutes) as the output target, training and validation are performed using a pre-defined model structure such as gradient boosting trees or neural networks. The training process will learn and output a dynamic coupling weight matrix that can quantify the influence of each input feature (including interaction effect terms) on the color value. The weight matrix is ​​then combined with the model structure to encapsulate a dynamic coupling relationship model that can be directly used for online prediction.

[0025] Preferably, a library of alternative collaborative adjustment schemes covering multiple process scenarios is constructed based on a dynamic coupling relationship model, specifically as follows: Using the set of historical and real-time digital feature vectors covering typical process ranges as input, the dynamic coupling relationship model is used to perform forward iterative calculations to predict the evolution trajectory of the reference color value under each process scenario. It should be noted that representative process state data from typical operating intervals of the sugar-making process are selected from long-term historical databases and real-time databases, and preprocessed into a set of digital feature vectors arranged in time series. Each independent vector represents a specific process scenario (for example, the i-th scenario vector represents the process state at a specific moment, such as sulfur fumigation intensity X, pH value Y, and temperature Z). Then, each digital feature vector in the set is used as the initial input and fed into a trained and validated dynamic coupling relationship model. The model starts with the aforementioned vectors and, based on its internal iterative logic (including the dynamic coupling weight matrix and model structure), it iterates through the model over a preset future time period (e.g., the next 3...). The total prediction time is 0 minutes. Forward iterative calculation is performed with a fixed time step (e.g., 1 minute). In each iteration, the predicted color value of the next time step is calculated based on the input vector at the current time. The predicted value, together with the preset or predicted values ​​of other process parameters (in the baseline prediction, the adjustable parameters are usually assumed to remain unchanged at the current value or change according to the historical average trend), is used to construct the input vector of the next time step. This is then fed back to the model for the next prediction. This process is repeated until the entire prediction period is completed, thus outputting a continuous curve composed of the predicted color values ​​of each time step connected in chronological order. This curve is the baseline color value evolution trajectory in the future period under the assumption that no active intervention is applied, corresponding to a specific process scenario.

[0026] The current values ​​of adjustable process parameters contained in the digital feature vectors corresponding to each process scenario are analyzed, and combined with the predefined parameter safety operating limits, the feasible adjustment space of each adjustable process parameter is determined. Within the feasible adjustment space of each process scenario, a series of discrete simulation adjustment quantities are systematically generated and combined according to the preset sampling rules to form multiple comprehensive simulation intervention schemes under the corresponding process scenario. Substitute the simulated intervention schemes under all process scenarios into the dynamic coupling relationship model for recalculation to obtain the corresponding simulated color value response trajectory; It should be noted that, based on the multiple comprehensive simulation intervention schemes (i.e., a series of specific combinations of adjustable process parameter adjustments, such as increasing sulfur fumigation intensity by 5% and decreasing pH value by 0.2) already generated for each process scenario, for each specific simulation intervention scheme, the original digital feature vector of its corresponding process scenario is modified according to the scheme. That is, the values ​​of the corresponding adjustable parameters in the vector are updated with the adjustment amounts specified in the scheme, while other non-adjustable parameters in the vector (such as the current color value, feed flow rate, etc.) or environmental parameters remain unchanged. This generates a simulation intervention feature vector representing the initial state after intervention. The simulation intervention feature vector is used as a new starting input, substituted into the dynamic coupling relationship model, and the same forward iteration as the predicted baseline trajectory is adopted. The computational mechanism is recalculated: the model uses the vector as the initial state and iteratively predicts the color value at each future time step (e.g., 30 minutes in the future, with a step size of 1 minute) based on the time delay and nonlinear coupling relationship between parameters characterized by its internal dynamic coupling weight matrix. At the same time, during the iteration process, the future values ​​of the adjustable process parameters involved in the model as input will be set according to the adjustment logic defined by the intervention scheme (e.g., remain constant after adjustment, or change according to a simple trend defined by the scheme), rather than using the assumptions of the baseline prediction. Through a series of iterative calculations, a continuous prediction curve is output, starting from the intervention start time, depicting how the color value will change in response to the specific control measure in the future prediction period.

[0027] For each simulated intervention scheme, the corresponding simulated color value response trajectory is compared with the baseline color value evolution trajectory under the corresponding process scenario to determine the control effectiveness evaluation value of the simulated intervention scheme in improving and maintaining the color value trajectory within the preset optimal range, as well as the process disturbance index caused by implementing the simulated intervention scheme. All simulated intervention schemes under all process scenarios, along with their corresponding simulated color value response trajectories, process disturbance indices, and control performance evaluation values, are associated, stored, and indexed to form the alternative collaborative adjustment scheme library.

[0028] It should be noted that a set of historical and real-time digital feature vectors covering typical process ranges (e.g., operating conditions under different raw material batches and production loads) is selected, where each vector represents a process scenario. These vectors are input one by one into a dynamic coupling relationship model. Through forward iterative calculation (recursively predicting the color value for each minute within the next 60 minutes), the baseline color value evolution trajectory for each scenario is obtained without any intervention. Then, the current values ​​of adjustable process parameters (such as sulfur fumigation intensity setpoint and pH adjustment valve opening) in each scenario vector are analyzed. Based on production safety and equipment limitations, combined with predefined parameter safety operating limits (e.g., sulfur fumigation intensity can be adjusted within ±10%, and pH value adjustment upper limit is ±0.3), the feasible adjustment space for each adjustable parameter is quickly determined. Next, within the feasible adjustment space of each scenario, all possible combinations of parameter adjustment amounts are generated according to the preset sampling rules (i.e., each parameter is sampled discretely within its feasible range at a fixed step size, such as a sulfur fumigation intensity step size of 0.5%), thereby forming a simulated intervention scheme under the corresponding scenario. Then, each simulated intervention scheme (i.e., a set of adjusted parameter values) is substituted into the dynamic coupling relationship model to obtain the predicted simulated color value response trajectory under the corresponding intervention.

[0029] The key to this step lies in the quantitative evaluation of each scheme: When calculating the control effectiveness evaluation value, the simulated color value response trajectory is first compared with the corresponding baseline color value evolution trajectory within the same future time period. Specifically, the color value difference between the two trajectories is calculated, and the area of ​​the simulated trajectory deviating from the preset optimal color value range (e.g., the target color value range is 100-150 ICUMSA units) is integrated and compared with the area of ​​the baseline trajectory deviating from this range. The control effectiveness evaluation value can be defined as the difference between the baseline deviation area and the simulated deviation area; the larger this value, the more significant the improvement effect. Simultaneously, the process disturbance index is calculated: based on the adjustment magnitude (absolute change or relative rate of change relative to the current value) and the frequency or severity of adjustment (e.g., parameter change rate) of each parameter in the simulated intervention scheme, the disturbances of each parameter are weighted and summed (the weights can be set according to the importance of the parameter or the adjustment cost). The process disturbance index is this weighted sum; the smaller the value, the smoother the process adjustment and the lower the cost.

[0030] Finally, all simulated intervention schemes under all scenarios, their corresponding simulated color value response trajectories, calculated process disturbance indices, and control performance evaluation values ​​are associated and stored and indexed in a database (e.g., using process scenario identifiers and parameter adjustment combinations as index keys), thereby constructing a library of alternative collaborative adjustment schemes that can be directly queried and called online for quick access.

[0031] Preferably, the dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value, specifically as follows: The latest digital feature vector is obtained as input to the dynamic coupling relationship model, and a dynamic coupling weight matrix matching the current process state is extracted from the dynamic coupling relationship model. The weighted summation of each component in the latest digital feature vector with the corresponding weight coefficient in the extracted dynamic coupling weight matrix is ​​performed to obtain the first color value prediction point for the next unit process cycle. Based on the first color value prediction point, and combined with the process state retention assumption derived from the digital feature vector, the prediction state vector for the next moment is updated and rolled forward. The predicted state vector is used as a new input and operated on with the dynamic coupling weight matrix again to obtain the second color value prediction point for subsequent unit process cycles. Repeat the above state rolling and model calculation steps, and iterate multiple times within the preset state inference window to obtain a future multi-step prediction trajectory composed of multiple ordered prediction points. The smoothness test and confidence assessment are performed on the multi-step predicted trajectory to generate the corresponding prediction confidence interval; The multi-step predicted trajectory and its prediction confidence interval are compared with the preset optimal color value range in real time, and a trend prediction trigger signal is output to determine whether to start control intervention.

[0032] It should be noted that the digital feature vector is used as the immediate input to the dynamic coupling relationship model. Simultaneously, based on the operating condition characteristics represented by the vector, the dynamic coupling weight matrix that best matches the specific process state is extracted or called from the model library to obtain the most applicable parameter coupling relationship. Then, the first prediction step is performed: the components of the digital feature vector (i.e., each parameter value) are weighted and summed with the corresponding weight coefficients in the dynamic coupling weight matrix, directly outputting the color value prediction point for the next unit process cycle (e.g., 1 minute later), i.e., the first color value prediction point. Subsequently, an iterative multi-step prediction stage begins: based on the first color value prediction point, combined with the process state maintenance assumption derived from the current digital feature vector (e.g., assuming that adjustable process parameters will remain unchanged in the short term, while only the color value changes according to the model's rules), a new prediction state vector containing the predicted color value and other parameter prediction values ​​(based on the assumption) is constructed. This vector is used as the input for the next moment and is again calculated with the same dynamic coupling weight matrix to obtain the second color value prediction point for the second unit process cycle (e.g., 2 minutes later). Repeat this iterative process, performing multiple calculations within a preset state projection window (e.g., predicting the next 30 minutes, iterating 30 times), and finally generating a future multi-step prediction trajectory composed of multiple ordered prediction points (e.g., 30 points).

[0033] In addition, to assess the reliability of the prediction results, the trajectory needs to be post-processed: a smoothness test is performed (e.g., calculating the difference between predicted points; if consecutive difference values ​​exceed a preset threshold, such as 10 ICUMSA units / minute, a smoothness warning is triggered) and a confidence assessment is performed (e.g., based on the standard deviation of historical prediction errors, the 95% confidence interval for each point on the predicted trajectory is calculated; the half-width of the interval can be used as a confidence index). The obtained multi-step predicted trajectory and its confidence interval (e.g., trajectory value is [predicted value], confidence interval is [predicted value ± 5 units]) are compared in real time with the preset optimal color value range (e.g., target range is 100-150 ICUMSA units): if the predicted trajectory will continue to exceed the optimal range during a critical period in the future (e.g., the next 10-15 minutes), or if the lower limit of its confidence interval is higher than the upper limit of the range (or the upper limit is lower than the lower limit of the range), a trend prediction trigger signal is output, indicating that there is a risk of deviation in the current process status, and subsequent control intervention procedures need to be initiated immediately, thereby realizing the transformation from passive monitoring to proactive early warning.

[0034] Preferably, when the trend prediction trigger signal indicates that the prediction trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time based on the deviation characteristics, and the pre-adjustment scheme that minimizes the overall process cost and returns the color value to the target range is calculated online. Specifically: Based on the relationship between the multi-step prediction trajectory and the boundary position of the optimal color value range, the direction, magnitude, and urgency of the color value deviation are determined, and a deviation feature description is formed. Using the aforementioned deviation characteristics as the core query vector, the real-time matching degree between the core query vector and the initial process state vector corresponding to each simulated intervention scheme stored in the alternative collaborative adjustment scheme library is calculated. Based on the aforementioned urgency, a matching degree threshold is set to initially select a subset of scenario matching schemes. From the subset of scenario matching schemes, the simulated color value response trajectory, process disturbance index and control performance evaluation value of each scheme are extracted. The current color value deviation is combined with the magnitude and direction for weighted correction to generate the estimated control performance value and estimated comprehensive process cost value of each scheme for the current real-time deviation. A real-time multi-objective optimization function is constructed with the urgency of the current color value deviation as a weight factor. The real-time multi-objective optimization function integrates the estimated control efficiency value and the estimated comprehensive process cost value, and performs online re-evaluation and ranking of all schemes in the scenario matching scheme subset. Based on the sorting results and the preset decision threshold, the final optimal adjustment scheme is selected or integrated, and the optimal adjustment scheme is coded for feasibility. The output is a pre-adjustment scheme that includes specific auxiliary material and key process parameter adjustment instructions, timing logic and expected effects.

[0035] It should be noted that, based on the boundary position relationship between the multi-step predicted trajectory and the preset optimal color value range (e.g., 100-150 ICUMSA units), the color value deviation characteristics are determined, including the deviation direction (above the upper limit or below the lower limit), the magnitude (e.g., the maximum predicted deviation is 20 units above the upper limit), and the urgency (e.g., quantified by the time point when the predicted trajectory first continuously exceeds the range; if it exceeds the range within 5 minutes, the urgency is "high"; if it exceeds the range after 15 minutes, the urgency is "low").

[0036] Using the deviation feature description (which can be quantified as a feature vector containing elements such as direction, amplitude, and urgency level) as the core query vector, the real-time matching degree between it and the initial process state vectors corresponding to all schemes in the alternative collaborative adjustment scheme library is calculated (e.g., calculating Euclidean distance or cosine similarity). A matching degree threshold is set based on urgency (e.g., when urgency is "high," a similarity threshold greater than 0.85 is set to quickly filter the most relevant schemes; when urgency is "low," the threshold can be set greater than 0.7 to obtain a wider range of schemes), thus initially filtering out a subset of scenario-matching schemes. The pre-stored simulated color value response trajectory, process disturbance index, and control performance evaluation value for each scheme are extracted from the subset. The pre-stored values ​​are weighted and corrected based on the amplitude and direction of the current real-time deviation: for example, if the current deviation amplitude is 1.5 times the pre-stored baseline deviation, the pre-stored control performance evaluation value is scaled proportionally; if the deviation direction is opposite, the performance value is inverted. This generates the estimated control effectiveness value and the estimated comprehensive process cost value for each scheme for the current specific deviation (the latter can be directly reflected by the corrected process disturbance index).

[0037] Then, a real-time multi-objective optimization function is constructed, with the following form: Total Evaluation Value = α * Estimated Control Effectiveness Value - β * Estimated Comprehensive Process Cost Value, where the weighting factors α and β are dynamically adjusted according to the current urgency of the deviation (e.g., when the urgency is "high", α = 0.8, β = 0.2, focusing on rapid correction; when the urgency is "low", α = 0.5, β = 0.5, focusing on process stability). Using this function, the corresponding total evaluation value is calculated synchronously for each scheme in the scenario matching scheme subset, and the schemes are quickly sorted from high to low based on the total evaluation value, thereby completing online re-evaluation and sorting.

[0038] Finally, based on the ranking results and preset decision thresholds (e.g., the total evaluation value must be greater than threshold T=0.6, and the adjustment range of a single process parameter must not exceed 80% of its safe operating limit), the final optimal adjustment scheme is generated by selecting or merging (e.g., weighted averaging of the top three schemes). The scheme is then coded for feasibility, outputting a pre-adjustment scheme containing specific instructions, timing logic, and expected effects, and sent to the process control system for execution, thus completing the final link from intelligent prediction to closed-loop optimization control.

[0039] Preferably, according to the pre-adjustment scheme, feedforward collaborative adjustment and control are performed on the auxiliary material addition system and key process execution units in the production process, specifically as follows: The instruction sequence encoded in the pre-adjustment scheme is analyzed to extract the types of auxiliary materials to be adjusted, the target adjustment amounts of each key process parameter, and the corresponding process execution unit identifiers. Based on the equipment response characteristics and process timing constraints of each process execution unit, the target adjustment amount is decoupled and allocated in the time dimension, and arranged into an equipment instruction execution plan that includes specific execution time points and execution order; Based on the equipment instruction execution plan and combined with the real-time acquired process cycle synchronization signal, the adjustment amount in the equipment instruction execution plan is converted into physical control instructions that can be recognized by the corresponding execution unit controller. Before the planned execution time arrives, corresponding physical control commands are sent to the controllers of the relevant auxiliary material addition systems and process execution units in advance, and the equipment command reception confirmation signals returned by each controller are received. Within the preset collaborative execution window, the device command completion status returned by each execution unit is monitored. When all necessary units have reported completion of execution, a collaborative adjustment completion confirmation signal is generated, thus completing the feedforward collaborative adjustment control.

[0040] Specifically, the process cycle synchronization signal acquired in real time is obtained by: capturing key equipment status signals that characterize process stage switching or material flow in real time based on preset key physical event trigger points in the sugar production line, and aligning the key equipment status signals with the central control clock to generate a unified process cycle synchronization signal.

[0041] It should be noted that the generated pre-adjustment scheme is analyzed to extract the coded instruction information, clarifying the type of auxiliary material to be adjusted (e.g., sulfur dioxide solution), the target adjustment amount of each key process parameter (e.g., pH value, heating temperature), and the corresponding process execution unit (e.g., sulfur fumigation tower solenoid valve, pH adjustment pump, heater power controller). Based on the equipment response characteristics of each execution unit (e.g., valve opening and closing delay, pump flow ramp-up time) and process timing constraints (e.g., pH adjustment needs to be performed within a specific time period after sulfur fumigation begins), the target adjustment amount is decoupled and finely allocated in the time dimension, and arranged into an equipment instruction execution plan that includes specific execution time points (e.g., sending the sulfur fumigation intensity instruction at time T0, and sending the pH adjustment instruction 30 seconds later at T0) and a strict execution sequence. Based on the equipment instruction execution plan, combined with the real-time acquired process cycle synchronization signal (e.g., synchronization with the main production line PLC clock), each adjustment amount in the plan is converted into physical control instructions (e.g., 4-20mA analog signals or specific register write values) that can be directly recognized and executed by each execution unit controller (e.g., PLC, DCS). Then, with an appropriate lead time (e.g., one control cycle earlier) before the planned execution time, corresponding physical control commands are sent to the controllers of the relevant auxiliary material adding systems and process execution units in advance, and command reception confirmation signals are received from each controller to ensure reliable delivery of commands. Within a preset collaborative execution window (e.g., within 120 seconds from the first command), the completion status of equipment commands returned by each execution unit is monitored in real time (e.g., valves reaching the specified opening degree, temperature reaching the set value). When all necessary execution units have reported completion, a collaborative adjustment completion confirmation signal is generated and output, marking the completion of this feedforward collaborative adjustment control closed loop. This ensures that multiple process parameters are adjusted accurately and synchronously according to the optimized scheme, achieving efficient and stable correction of color value deviations.

[0042] In this invention, the control method may further include the following steps: The given target color value range of a specific finished sugar product is jointly encoded with the key process constraints of the production process to generate a multi-dimensional target-constraint joint space. Using the dynamic coupling relationship model as a forward simulator, a set of candidate process parameter vectors covering the feasible region is generated in the solution space consisting of all adjustable process parameters by employing sparse grid sampling technology. Each candidate process parameter vector is input into the dynamic coupling relationship model, and forward simulation calculation is performed to obtain the corresponding predicted color value trajectory. The predicted color value trajectory is then compared and integrated with the boundary of the target color value range point by point to calculate the color value deviation index corresponding to each candidate vector. Meanwhile, the candidate process parameter vectors are correlated and weighted with the unit output energy consumption and auxiliary material consumption records in the historical database to generate a comprehensive process economic index corresponding to each candidate vector. The calculated color value deviation index and the comprehensive process economic index are normalized and used as two optimization objectives. Multi-objective Pareto front solutions are then performed on all candidate process parameter vector sets to select the non-dominated solution set located on the Pareto front. Based on the time-series smoothness requirements of the production process, the process parameter vectors in the non-dominated solution set are sorted and interpolated according to the process segment order to generate a continuous and smooth optimal process parameter baseline setting curve.

[0043] It should be noted that in sugar production, traditional fixed process parameter settings are difficult to adapt to raw material fluctuations and market demand for multi-grade sugar products, often facing the dilemma of pursuing high quality leading to soaring costs or controlling costs at the expense of quality. To solve this problem, this embodiment, based on the established dynamic coupling relationship model, further performs the following steps: For the specific target color value range of finished sugar required by a specific production order (such as the more stringent range required for high-end refined sugar) and the key process constraints of the actual production line (such as maximum sulfur fumigation capacity and upper limit of steam pressure), both are jointly encoded. Specifically, the upper and lower limits of the color value range and the boundary values ​​of each constraint are used as dimensions to construct a multi-dimensional target-constraint joint space, which defines all theoretically acceptable process operation areas. Using a dynamic coupling model as a forward simulator, within the solution space spanned by all adjustable process parameters (such as sulfur fumigation intensity, pH value, and temperature of each chamber of the multi-effect evaporator), a sparse grid sampling technique is employed to generate a set of candidate process parameter vectors that can broadly cover the feasible domain while taking into account computational efficiency. For example, each parameter is taken at sparse grid points within its allowable range, and full combination or Latin hypercube sampling is performed.

[0044] Then, simulation and evaluation are performed: Each candidate parameter vector is input into a forward simulator to simulate the duration of a future production batch, obtaining the predicted color value trajectory. This trajectory is then compared point-by-point with the boundary of the target color value range and integrated to calculate the area or weighted deviation of the color value from the target range throughout the entire prediction period, which serves as the color value deviation index for this scheme (the smaller the value, the better). Simultaneously, the candidate parameter vector is correlated with historical database records of unit output energy consumption and auxiliary material (such as sulfur and lime) consumption under the same or similar operating conditions. A weighted summation (weights can be set based on real-time energy and material costs) is used to calculate the comprehensive process economic index corresponding to this scheme (the smaller the value, the lower the cost).

[0045] Next, multi-objective optimization decision-making is performed: the color value deviation index and the comprehensive process economic index of all candidate schemes are normalized to eliminate the influence of dimensions. Using these two indices as optimization objectives that need to be minimized simultaneously, Pareto front solutions are obtained using multi-objective optimization algorithms such as NSGA-II on the entire set of candidate schemes. The non-dominated solution set that is better in at least one objective and is not surpassed by other schemes at the same time is selected. These solutions represent a series of optimal trade-offs between quality and cost. Based on the requirements of the production process for the smoothness of parameter changes over time (to avoid drastic parameter jumps that could lead to production instability), the process parameter vectors in the non-dominated solution set are sorted according to the process segment sequence (such as clarification, evaporation, and sugar boiling), and fitted using algorithms such as spline interpolation. This generates a continuous and smooth optimal process parameter benchmark setting curve, which can be directly sent to the control systems of each process segment as a benchmark for setting value tracking. This achieves the connection from offline global optimization to online stable execution, thus solving the problem of balancing fixed set values ​​and comprehensive benefits. It provides standardized guidance for the entire production cycle that balances quality standards, cost optimization, and operational feasibility, fundamentally improving the overall economic efficiency and control level of sugar production.

[0046] In this invention, the control method may further include the following steps: Based on the dynamic coupling relationship model and historical process data, a process state transition diagram representing the steady state of the system under different parameter combinations is constructed, using multiple adjustable process parameters as dimensions. Guided by the process state transition diagram, spatial discretization sampling is performed in the multi-dimensional parameter space composed of the key auxiliary material addition rate and key process parameters to generate a multi-dimensional parameter grid covering the operable domain. The parameter combination corresponding to each multidimensional grid vertex is used as the initial condition and substituted into the dynamic coupling relationship model for forward simulation and deduction to obtain the corresponding steady-state color value and production prediction index, thereby fitting and generating a multidimensional process response hypersurface with process parameters as input and comprehensive quality and production index as output. On this multi-dimensional process response hypersurface, gradient calculation and contour analysis are performed to identify the parameter optimization gradient field pointing to the region of optimal quality and yield. Using the optimized gradient field as navigation, and combined with the preset process constraint boundary, the process parameter adjustment path that enables the comprehensive index to reach Pareto optimality is solved on the hypersurface. The path is then discretized into a series of specific auxiliary material and process parameter coordinated adjustment instruction sequences, which are stored as the global optimal adjustment scheme.

[0047] It should be noted that a process state transition diagram is constructed using multiple adjustable process parameters (such as sulfur fumigation intensity, phosphoric acid addition, and primary juice pH) as dimensions. This diagram is a directed graph network where nodes represent the quasi-steady state of the system defined by specific parameter combinations, and edges represent the state transition relationships achievable through parameter adjustment and their dynamic costs (such as transition time and energy consumption). During construction, typical steady-state nodes can be identified from historical data using clustering algorithms, and the relationships between edges can be established by simulating the state transition process using a model. Guided by the process state transition diagram, the parameter space range requiring fine-grained exploration is determined. Spatial discretization sampling is performed within the multivariate parameter space composed of key auxiliary material addition rates and key process parameters. For example, each parameter is divided into equally spaced or non-uniformly partitioned sections within its operable range, generating a multidimensional parameter grid covering the entire operable domain (e.g., an N-dimensional grid structure, where N is the number of parameters).

[0048] Then, global simulation and response surface fitting are performed: the parameter combination corresponding to each multidimensional grid vertex is used as the initial condition and substituted into the dynamic coupling relationship model for long-term forward simulation until the system output (color value, intensity, and instantaneous output calculated from flow rate) reaches stability, thereby obtaining the steady-state color value and output prediction index corresponding to each parameter combination. Using the simulation results of all grid vertices (input parameter combination and output index), a multidimensional process response hypersurface is generated through methods such as Kriging interpolation or neural network fitting, with process parameters as input and comprehensive quality (such as color value) and output index as output. Analysis and navigation are then performed on this hypersurface: gradient calculation (such as calculating the partial derivative of the comprehensive index with respect to each parameter at any point on the hypersurface) and contour analysis (drawing contour lines of quality and output indexes) are performed to identify the parameter optimization gradient field pointing to the optimal region of comprehensive quality and output.

[0049] Finally, using the parameter optimization gradient field as navigation basis, and combined with preset process constraint boundaries (such as equipment safety limits and material balance constraints), a path search algorithm (such as a combination of gradient descent and constraint processing) is employed on the hypersurface to solve for the process parameter adjustment path that enables the comprehensive index to reach Pareto optimality. This path is a feasible and efficient trajectory from the current state to the Pareto front. This path is discretized in the time and parameter dimensions, transforming it into a series of specific, time-series-related instructions for the coordinated adjustment of auxiliary materials and process parameters, ultimately stored as a globally optimal adjustment scheme that can be directly invoked. This embodiment overcomes the shortcomings of traditional methods that easily get trapped in local optima and struggle to balance multiple objectives in complex high-dimensional process spaces. By constructing and analyzing the global process response hypersurface, an intuitive grasp and quantitative analysis of the global situation of the process system is achieved. It can scientifically and automatically find the globally optimal or near-optimal operating regions under both quality and yield indicators, and generate smooth and feasible adjustment instruction sequences. This improves the overall production efficiency and resource utilization of the entire sugar production system while ensuring the stability of the core quality index (color value) of the sugar juice.

[0050] like Figure 2 As shown, the second aspect of the present invention discloses an online detection and process parameter control system for sugar syrup color value. The system includes a memory and a processor. The memory stores a program for online detection and process parameter control of sugar syrup color value. When the program for online detection and process parameter control of sugar syrup color value is executed by the processor, the steps of the online detection and process parameter control method for sugar syrup color value described in any one of the claims are implemented.

[0051] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0052] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0054] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0056] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for online detection of color value of sugar syrup and control of its process parameters, characterized in that, Includes the following steps: The system collects sugar juice color value data from multiple preset sampling points during the sugar-making process in real time, and simultaneously acquires the corresponding auxiliary material addition parameters and key process parameters, which together form a digital feature vector reflecting the current process status. Based on historical and real-time data training models, and using the digital feature vectors as input, a dynamic coupling relationship model between the evolution of sugar juice color value and multiple process parameters is analyzed and established. Based on the dynamic coupling relationship model, a library of alternative collaborative adjustment schemes for covering multiple process scenarios is constructed. The dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value. When the trend prediction trigger signal indicates that the prediction trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time according to the deviation characteristics, and the pre-adjustment scheme that makes the color value return to the target range and minimizes the overall process cost is optimized and calculated online. According to the pre-adjustment scheme, feedforward collaborative adjustment and control are carried out on the auxiliary material addition system and key process execution units in the production process.

2. The method for online detection of sugar juice color value and control of its process parameters according to claim 1, characterized in that, The system collects sugar juice color value data from multiple preset sampling points during the sugar-making process in real time, and simultaneously acquires corresponding auxiliary material addition parameters and key process parameters. Together, these parameters form a digital feature vector reflecting the current process status, specifically: The real-time color value readings output by the online color value detection device at multiple preset sampling points are obtained to form an original color value data sequence; Meanwhile, the instantaneous values ​​of the instantaneous addition rate of auxiliary materials and key process parameters corresponding to the process positions of each sampling point are extracted from the sugar production process control system in real time to form the original process parameter data set; The original color value data sequence and the original process parameter data set are aligned and bound according to a unified time base to generate a synchronous dataset; The color value data in the synchronous dataset is filtered and denoised to obtain a stable color value sequence, and the validity of the process parameter data is verified to obtain a set of valid process parameters. The stable color value sequence and the set of effective process parameters are normalized and spliced ​​according to a preset dimension and order to form a digital feature vector representing the overall process state at the current moment.

3. The method for online detection of sugar juice color value and control of its process parameters according to claim 1, characterized in that, Based on historical and real-time data training, and using the aforementioned digital feature vector as input, a dynamic coupling relationship model between the evolution of sugar syrup color value and multiple process parameters is analyzed and established, specifically as follows: The digital feature vectors of continuous time series are extracted from historical databases and real-time data streams and arranged in process order to form a process state time series segment. Time lag correlation analysis is performed on the time series segments of the process state to calculate the correlation strength between color value data and each process parameter at different time lags, and a lag correlation strength map is formed. The correlation strength of each item in the hysteresis correlation strength spectrum is compared with a preset correlation threshold. Process parameters whose correlation strength exceeds the preset correlation threshold and their corresponding time lags are screened out, thereby defining the key hysteresis parameter set and their respective time lag intervals. Based on the set of key lag parameters and their time lag intervals, the corresponding parameter lag feature sub-vectors are extracted and reconstructed from the time series segments of the process state. The parameter lag feature vector is combined with the current time-time digital feature vector, and a multi-parameter interaction effect term representing the nonlinear interaction between parameters is introduced to jointly form an extended coupled feature set for model training. Using the extended coupling feature set as input and the color value change in a specific future time period as output target, the preset model structure is trained and verified to obtain a dynamic coupling weight matrix that can quantify the influence of multi-parameter coupling on color value. The dynamic coupling weight matrix is ​​combined with the model structure to encapsulate a dynamic coupling relationship model.

4. The method for online detection of sugar juice color value and control of its process parameters according to claim 1, characterized in that, A library of alternative collaborative adjustment schemes covering multiple process scenarios is constructed based on a dynamic coupling relationship model, specifically as follows: Using the set of historical and real-time digital feature vectors covering typical process ranges as input, the dynamic coupling relationship model is used to perform forward iterative calculations to predict the evolution trajectory of the reference color value under each process scenario. The current values ​​of adjustable process parameters contained in the digital feature vectors corresponding to each process scenario are analyzed, and combined with the predefined parameter safety operating limits, the feasible adjustment space of each adjustable process parameter is determined. Within the feasible adjustment space of each process scenario, a series of discrete simulation adjustment quantities are systematically generated and combined according to the preset sampling rules to form multiple comprehensive simulation intervention schemes under the corresponding process scenario. Substitute the simulated intervention schemes under all process scenarios into the dynamic coupling relationship model for recalculation to obtain the corresponding simulated color value response trajectory; For each simulated intervention scheme, the corresponding simulated color value response trajectory is compared with the baseline color value evolution trajectory under the corresponding process scenario to determine the control effectiveness evaluation value of the simulated intervention scheme in improving and maintaining the color value trajectory within the preset optimal range, as well as the process disturbance index caused by implementing the simulated intervention scheme. All simulated intervention schemes under all process scenarios, along with their corresponding simulated color value response trajectories, process disturbance indices, and control performance evaluation values, are associated, stored, and indexed to form the alternative collaborative adjustment scheme library.

5. The method for online detection of sugar juice color value and control of its process parameters according to claim 1, characterized in that, The dynamic coupling relationship model is used to predict the trend of real-time sugar juice color value, specifically as follows: The latest digital feature vector is obtained as input to the dynamic coupling relationship model, and a dynamic coupling weight matrix matching the current process state is extracted from the dynamic coupling relationship model. The weighted summation of each component in the latest digital feature vector with the corresponding weight coefficient in the extracted dynamic coupling weight matrix is ​​performed to obtain the first color value prediction point for the next unit process cycle. Based on the first color value prediction point, and combined with the process state retention assumption derived from the digital feature vector, the prediction state vector for the next moment is updated and rolled forward. The predicted state vector is used as a new input and operated on with the dynamic coupling weight matrix again to obtain the second color value prediction point for subsequent unit process cycles. Repeat the above state rolling and model calculation steps, and iterate multiple times within the preset state inference window to obtain a future multi-step prediction trajectory composed of multiple ordered prediction points. The smoothness test and confidence assessment are performed on the multi-step predicted trajectory to generate the corresponding prediction confidence interval; The multi-step predicted trajectory and its prediction confidence interval are compared with the preset optimal color value range in real time, and a trend prediction trigger signal is output to determine whether to start control intervention.

6. The method for online detection of sugar juice color value and control of its process parameters according to claim 5, characterized in that, When the trend prediction trigger signal indicates that the predicted trajectory deviates from the preset optimal color value range, the alternative collaborative adjustment scheme library is queried in real time based on the deviation characteristics, and the pre-adjustment scheme that minimizes the overall process cost and returns the color value to the target range is calculated online. Specifically: Based on the relationship between the multi-step prediction trajectory and the boundary position of the optimal color value range, the direction, magnitude, and urgency of the color value deviation are determined, and a deviation feature description is formed. Using the aforementioned deviation characteristics as the core query vector, the real-time matching degree between the core query vector and the initial process state vector corresponding to each simulated intervention scheme stored in the alternative collaborative adjustment scheme library is calculated. Based on the aforementioned urgency, a matching degree threshold is set to initially select a subset of scenario matching schemes. From the subset of scenario matching schemes, the simulated color value response trajectory, process disturbance index and control performance evaluation value of each scheme are extracted. The current color value deviation is combined with the magnitude and direction for weighted correction to generate the estimated control performance value and estimated comprehensive process cost value of each scheme for the current real-time deviation. A real-time multi-objective optimization function is constructed with the urgency of the current color value deviation as a weight factor. The real-time multi-objective optimization function integrates the estimated control efficiency value and the estimated comprehensive process cost value, and performs online re-evaluation and ranking of all schemes in the scenario matching scheme subset. Based on the sorting results and the preset decision threshold, the final optimal adjustment scheme is selected or integrated, and the optimal adjustment scheme is coded for feasibility. The output is a pre-adjustment scheme that includes specific auxiliary material and key process parameter adjustment instructions, timing logic and expected effects.

7. The method for online detection of sugar syrup color value and control of its process parameters according to claim 1, characterized in that, According to the aforementioned pre-adjustment scheme, feedforward collaborative adjustment and control are implemented for the auxiliary material addition system and key process execution units in the production process, specifically as follows: The instruction sequence encoded in the pre-adjustment scheme is analyzed to extract the types of auxiliary materials to be adjusted, the target adjustment amounts of each key process parameter, and the corresponding process execution unit identifiers. Based on the equipment response characteristics and process timing constraints of each process execution unit, the target adjustment amount is decoupled and allocated in the time dimension, and arranged into an equipment instruction execution plan that includes specific execution time points and execution order; Based on the equipment instruction execution plan and combined with the real-time acquired process cycle synchronization signal, the adjustment amount in the equipment instruction execution plan is converted into physical control instructions that can be recognized by the corresponding execution unit controller. Before the planned execution time arrives, corresponding physical control commands are sent to the controllers of the relevant auxiliary material addition systems and process execution units in advance, and the equipment command reception confirmation signals returned by each controller are received. Within the preset collaborative execution window, the device command completion status returned by each execution unit is monitored. When all necessary units have reported completion of execution, a collaborative adjustment completion confirmation signal is generated, thus completing the feedforward collaborative adjustment control.

8. The method for online detection of sugar juice color value and control of its process parameters according to claim 7, characterized in that, The process cycle synchronization signal combined with real-time acquisition is specifically as follows: based on the preset key physical event trigger points in the sugar production line, the key equipment status signals that characterize the process stage switching or material flow are captured in real time, and the key equipment status signals are aligned with the central control clock to generate a unified process cycle synchronization signal.

9. An online detection and process parameter control system for sugar syrup color value, characterized in that, The system includes a memory and a processor. The memory stores a program for online detection of sugar syrup color value and control of its process parameters. When the program for online detection of sugar syrup color value and control of its process parameters is executed by the processor, the steps of the method for online detection of sugar syrup color value and control of its process parameters as described in any one of claims 1 to 8 are implemented.

Images