Polycarboxylate superplasticizer fermentation monitoring system

Abstract

The invention relates to the technical field of polycarboxylate superplasticizer production monitoring, in particular to an old-age nursing service system based on an intelligent internet of things. Comprising a data acquisition unit, a self-adaptive decision control unit, an intelligent optimization and digital twinning unit and a security and redundancy unit, the data acquisition unit is used for acquiring fermentation parameters in real time through a wireless local area network and performing data aggregation and uploading through an edge gateway; and the self-adaptive decision control unit is based on a fuzzy control and model prediction control dynamic switching mechanism of fermentation process feature recognition, is combined with a self-adaptive prediction control model of online parameter updating, and is a collaborative architecture of edge end real-time control and cloud end optimization parameter adjustment. Through a fuzzy control and model predictive control dynamic switching mechanism of fermentation process feature recognition and an adaptive predictive control model of online parameter updating, a control strategy can be automatically adjusted according to the change of a fermentation stage, the control accuracy and flexibility are improved, and the fermentation efficiency and the product quality are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production monitoring of polycarboxylate water reducers, and specifically to a fermentation monitoring system for polycarboxylate water reducers. Background Art

[0002] As an important concrete admixture, polycarboxylate water reducers are widely used in the construction industry, and their quality and performance directly affect various performance indicators of concrete. In the production process of polycarboxylate water reducers, the fermentation link is one of the key steps. The stability of the fermentation process, the quality and yield of the products are affected by various factors, such as changes in parameters such as temperature, pH value, substrate concentration, dissolved oxygen, etc. In the prior art, most automated monitoring systems have a single monitored parameter, lack a comprehensive and in-depth analysis and understanding of the fermentation process, and it is difficult to adjust the control strategy in a timely manner according to the changes in the fermentation stage, resulting in low fermentation efficiency and unstable product quality. At the same time, the interface of remote visual monitoring is not friendly enough, and it is difficult for operators to quickly obtain key information.

[0003] Based on this, the present invention provides a fermentation monitoring system for polycarboxylate water reducers to solve the above-mentioned technical problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a fermentation monitoring system for polycarboxylate water reducers. Through the fuzzy control and model predictive control dynamic switching mechanism for fermentation process feature recognition of the present invention, and the adaptive predictive control model for online parameter update, the control strategy can be automatically adjusted according to the changes in the fermentation stage, improving the accuracy and flexibility of control, and effectively enhancing the fermentation efficiency and product quality.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] The present invention provides a fermentation monitoring system for polycarboxylate water reducers, including a data acquisition unit, an adaptive decision-making control unit, an intelligent optimization and digital twin unit, and a safety and redundancy unit, wherein:

[0007] The data acquisition unit: is used to collect fermentation parameters in real time through a wireless local area network, and perform data aggregation and upload through an edge gateway;

[0008] The adaptive decision-making control unit: based on the fuzzy control and model predictive control dynamic switching mechanism for fermentation process feature recognition, combined with the adaptive predictive control model for online parameter update, and through the collaborative architecture of real-time control at the edge and optimization of parameter tuning in the cloud;

[0009] The intelligent optimization and digital twin unit: fine-tunes the pre-trained universal fermentation model based on transfer learning to adapt to new scenarios, recommends optimal parameters based on the effects of digital twin simulation control strategies, and uses augmented reality monitoring to provide workers with abnormal point annotations and suggestions;

[0010] The security and redundancy unit adopts industrial-grade encryption protocols and local cache mechanisms to ensure data security of wireless communications and emergency control of network disconnection.

[0011] The data acquisition unit includes a sensor module, a wireless communication module, and an edge gateway module, wherein:

[0012] The sensor module is used to deploy various types of sensors in various places of the fermentation equipment to accurately collect various parameters in the fermentation process of polycarboxylate water reducer in real time;

[0013] The wireless communication module is used to package and encode the data collected by the multi-sensors and transmit them to the edge gateway module through the wireless local area network;

[0014] The edge gateway module is used to receive multi-sensor data, perform aggregation, initial screening, and noise reduction analysis, and then upload it to the cloud or subsequent units, and has the ability to temporarily store data when the network is abnormal.

[0015] The adaptive decision control unit includes a feature recognition module, a control strategy switching module, an adaptive predictive control model module, and an edge-cloud collaboration module, wherein:

[0016] The feature recognition module analyzes the fermentation data in real time based on the cluster analysis algorithm, identifies the fermentation stage and features to determine the fermentation status;

[0017] The control strategy switching module is used to adopt fuzzy control when the fermentation process is stable according to the feature recognition results, and switch to model predictive control when precise control prediction is required;

[0018] The adaptive predictive control model module is used to combine historical and real-time data to establish and update the model, predict future fermentation trends, adjust control parameters, and support online parameter updates to adapt to changes;

[0019] The edge-cloud collaboration module is used to coordinate real-time monitoring and rapid control at the edge, and deep analysis and parameter optimization at the cloud.

[0020] The cluster analysis algorithm analyzes the fermentation data in real time, identifies the fermentation stage and characteristics to determine the fermentation state. The specific formula of the cluster analysis algorithm is:

[0021] The weighted Euclidean distance is used to calculate the distance between the sample point X and the center:

[0022]

[0023] Wherein, w j (t) is the time-varying feature weight; x j ' is the normalization parameter;

[0024] Update the clustering center every 5 minutes:

[0025]

[0026] Wherein, s m is the set of samples of the m-th class;

[0027] When the within-class variance σ of 10 consecutive samples 2 <0, it is determined that the corresponding fermentation stage is entered.

[0028] The method for combining historical and real-time data to establish and update a model, predict the future trend of fermentation, adjust control parameters, and support online parameter update to adapt to changes, the specific operation content is:

[0029] A1. Establish a state space model based on the fermentation kinetics equation, where the relationship between the cell growth rate μ and the substrate concentration S is:

[0030]

[0031] The model parameters are updated online once an hour, u is the actual specific growth rate, representing the growth amount of microorganisms per unit time; μ max is the maximum specific growth rate, the growth rate limit value under ideal conditions; S is the substrate concentration, the instantaneous concentration of nutrients in the fermentation broth; K s is the half-saturation constant; K i is the inhibition coefficient;

[0032] A2. Take the minimization of the product molecular weight distribution variance in the next 1 hour as the objective function, and the expression of the objective function is:

[0033]

[0034] Wherein, PDI k is the product dispersion index in the k-th period, characterizing the uniformity of the molecular weight distribution; PDI target is the target dispersion index, the PDI value corresponding to the ideal product molecular weight distribution; u k is the control quantity vector in the k-th period; u k-1 is the control quantity in the previous period, used to limit the sudden change of the control action; λ is the adjustment coefficient;

[0035] A3. Use sequential quadratic programming to solve the optimal control quantity in the next 30 minutes;

[0036] A4. The predicted output is weighted and corrected based on real-time pH and dissolved oxygen data, and the weight coefficient α ∈ [0.2, 0.5].

[0037] The coordination of real-time monitoring and rapid control at the edge, and in-depth analysis and parameter optimization in the cloud, the specific operation content is as follows:

[0038] B1. Edge layer: Rapidly control the temperature and stirring speed with a cycle ≤ 1 second, and use first-order lag filtering for data preprocessing;

[0039] B2. Cloud layer: Receive edge data packets every 5 minutes, generate parameter optimization suggestions through digital twin simulation, and the downward voice volume ≤ 1KB;

[0040] B3. Disconnection disaster tolerance: When the network delay > 300ms, the edge device automatically enables the optimal parameter table cached locally and matches the control instructions according to the fermentation timestamp.

[0041] The intelligent optimization and digital twin unit includes a transfer learning module, a digital twin model construction module, a control strategy simulation and optimization module, and an augmented reality monitoring module, where:

[0042] The transfer learning module is used to pre-train a general fermentation model through public or historical data, and then use a small amount of new scenario data for transfer fine-tuning to adapt to the specific scenario of polycarboxylate superplasticizer fermentation;

[0043] The digital twin model construction module is used to construct a digital twin model based on the physical characteristics and process parameters of the actual fermentation equipment, and real-time simulate the changes in fermentation process parameters;

[0044] The control strategy simulation and optimization module is used to simulate different control strategies in the digital twin model, compare the effects, recommend the optimal control parameters and strategies for actual production, and improve the fermentation efficiency and product quality;

[0045] The augmented reality monitoring module is used to integrate virtual information and real vision using augmented reality technology. Workers can view the parameters through AR glasses, and AI analyzes and annotates anomalies and gives suggestions.

[0046] Construct a digital twin model based on the physical characteristics and process parameters of the actual fermentation equipment, and real-time simulate the changes in fermentation process parameters. The specific operation content of model construction is as follows:

[0047] C1. Macro scale: Establish an overall kinetic model of the reactor based on the mass-energy conservation equation; Micro scale: Use an LSTM neural network to predict the distribution of microbial metabolic fluxes;

[0048] C2. Synchronize the temperature, pH, and dissolved oxygen data of the actual equipment to the twin every 2 minutes through the OPC UA protocol;

[0049] C3. When the deviation between the simulation value and the actual value continuously exceeds 5%, trigger the online parameter correction based on the Kalman filter.

[0050] In the digital twin model, different control strategies are simulated, and the comparison results recommend the optimal control parameters and strategies for actual production, improving the fermentation efficiency and product quality. The specific operation content is as follows:

[0051] D1. In the digital twin, parallelly simulate the product molecular weight distribution and conversion rate in the next 1 hour of three strategies: conventional PID control, model predictive control, and adaptive fuzzy control.

[0052] D2. With the constraints of PDI ≤ 1.3 and conversion rate ≥ 85%, use the NSGA-II algorithm to solve the Pareto optimal solution set.

[0053] D3. Push the top three optimization strategies and the comparison chart of expected benefits to the operator through the AR interface.

[0054] The safety and redundancy unit includes an encryption module, a local cache module, and an emergency control module, where:

[0055] The encryption module is used to encrypt at the data sending end and decrypt at the receiving end using the AES industrial-level encryption protocol.

[0056] The local cache module is used to temporarily store key data and control instructions with a certain storage capacity when the network is abnormal, and automatically upload them to the cloud after the network resumes, ensuring the coherence and integrity of the data.

[0057] The emergency control module is used to implement emergency control of the fermentation equipment according to local rules and preset strategies in the event of a network outage, maintain the basic operating parameters, and help the system quickly recover after troubleshooting.

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] 1. Through the fuzzy control and model predictive control dynamic switching mechanism for fermentation process feature recognition, and the adaptive predictive control model for online parameter update, the present invention can automatically adjust the control strategy according to the changes in the fermentation stage, improving the accuracy and flexibility of control, and effectively enhancing the fermentation efficiency and product quality.

[0060] 2. By using transfer learning to adapt to new scenarios, combining digital twin to simulate the control strategy effects and recommend optimal parameters, and providing intuitive and accurate operation guidance for workers through augmented reality monitoring, the present invention further optimizes the fermentation process management. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a system diagram of a polycarboxylate superplasticizer fermentation monitoring system of the present invention.

[0062] Description of the attached reference numerals:

[0063] 100, data acquisition unit; 101, sensor module; 102, wireless communication module; 103, edge gateway module; 200, adaptive decision control unit; 201, feature recognition module; 202, control strategy switching module; 203, adaptive predictive control model module; 204, edge-cloud collaboration module; 300, intelligent optimization and digital twin unit; 301, transfer learning module; 302, digital twin model construction module; 303, control strategy simulation and optimization module; 304, augmented reality monitoring module; 400, security and redundancy unit; 401, encryption module; 402, local cache module; 403, emergency control module. Detailed implementation manners

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] Embodiment:

[0066] As Figure 1 shown, this embodiment provides a polycarboxylate superplasticizer fermentation monitoring system, including a data acquisition unit 100, an adaptive decision control unit 200, an intelligent optimization and digital twin unit 300, and a security and redundancy unit 400, wherein: The data acquisition unit 100: is used to collect fermentation parameters in real time through a wireless local area network, and perform data aggregation and upload through an edge gateway; The adaptive decision control unit 200: is based on a fuzzy control and model predictive control dynamic switching mechanism for fermentation process feature recognition, combines an adaptive predictive control model with online parameter update, and adopts a collaborative architecture of real-time control at the edge and optimization of parameter tuning in the cloud; The intelligent optimization and digital twin unit 300: fine-tunes a pre-trained general fermentation model based on transfer learning to adapt to a new scenario, combines digital twin simulation to control the effect of the control strategy to recommend optimal parameters, and uses augmented reality monitoring to provide anomaly point marking and suggestions for workers; The security and redundancy unit 400: adopts an industrial-grade encryption protocol and a local cache mechanism to ensure the data security of wireless communication and emergency control in case of network disconnection.

[0067] In this embodiment, it should be noted that the data acquisition unit 100 collects fermentation parameters in real time through a wireless local area network and uploads them to the edge gateway. The adaptive decision-making control unit 200 dynamically switches between fuzzy control and model predictive control strategies based on feature recognition, and realizes parameter optimization through an edge-cloud collaborative architecture. The intelligent optimization and digital twin unit 300 recommends the optimal control scheme using transfer learning and digital twin technologies and presents it through AR interaction. At the same time, the security and redundancy unit 400 ensures the security of the full-link data and the emergency control of faults.

[0068] In addition, it should be noted that the wireless local area network adopts communication protocols such as Wi-Fi and ZigBee.

[0069] In the present invention, the data acquisition unit 100 includes a sensor module 101, a wireless communication module 102, and an edge gateway module 103, where: the sensor module 101 is used to deploy various types of sensors at various locations of the fermentation equipment to collect various parameters in the fermentation process of polycarboxylate superplasticizer in real time and accurately; the wireless communication module 102 is used to package and encode the data collected by multiple sensors through a wireless local area network and then transmit it to the edge gateway module 103; the edge gateway module 103 is used to receive multi-sensor data, perform aggregation, primary screening, and noise reduction analysis, and then upload it to the cloud or subsequent units, and has the ability to temporarily store data in case of network anomalies.

[0070] In this embodiment, it should be noted that the sensor module 101 collects various parameters of fermentation in real time, encrypts and transmits them through the wireless communication module 102 to the edge gateway module 103 for data fusion and preprocessing, and then uploads them to the cloud.

[0071] In addition, it should be noted that various types of sensors include temperature sensors, pH value sensors, pressure sensors, dissolved oxygen sensors, and substrate concentration sensors; various parameters include temperature, pH value, pressure, dissolved oxygen, and substrate concentration in the fermentation process; in case of network anomalies, built-in storage chips or storage devices such as small hard disks are used to temporarily store data, and the storage capacity can be set to different specifications such as 1GB - 10GB according to actual needs.

[0072] In the present invention, the adaptive decision-making control unit 200 includes a feature recognition module 201, a control strategy switching module 202, an adaptive predictive control model module 203, and an edge-cloud collaborative module 204, where: the feature recognition module 201 analyzes fermentation data in real time based on the clustering analysis algorithm, identifies the fermentation stage and features to judge the fermentation state; the specific formula of the clustering analysis algorithm is:

[0073] The weighted Euclidean distance is used to calculate the distance between the sample point X and the center:

[0074]

[0075] where, w j (t) is the time-varying feature weight; x j ' is the normalization parameter;

[0076] Update the clustering center every 5 minutes:

[0077]

[0078] where, s m is the set of samples of the m-th class;

[0079] When the within-class variance σ of 10 consecutive samples 2 < 0, it is determined that the corresponding fermentation stage is entered.

[0080] The control strategy switching module 202 is used to adopt fuzzy control when the fermentation process is stable according to the feature recognition result, and switch to model predictive control when precise control is required for prediction; the adaptive predictive control model module 203 is used to establish and update the model by combining historical and real-time data, predict the future trend of fermentation, adjust the control parameters, and support online parameter update to adapt to changes; the specific operation content is as follows:

[0081] A1. Establish a state space model based on the fermentation kinetics equation, where the relationship between the cell growth rate μ and the substrate concentration S is:

[0082]

[0083] The model parameters are updated online once an hour. u is the actual specific growth rate, representing the growth amount of microorganisms per unit time; μ max is the maximum specific growth rate, the growth rate limit value under ideal conditions; S is the substrate concentration, the instantaneous concentration of nutrients in the fermentation broth; K s is the half-saturation constant; K i is the inhibition coefficient;

[0084] A2. Take the minimization of the variance of the product molecular weight distribution in the next 1 hour as the objective function, and the expression of the objective function is:

[0085]

[0086] where, PDI k is the product dispersion index in the k-th period, characterizing the uniformity of the molecular weight distribution; PDI target is the target dispersion index, the PDI value corresponding to the ideal product molecular weight distribution; the PDI value is usually 1.0 - 1.5; u k is the control quantity vector in the k-th period; the control quantity vector includes process parameters such as temperature and feeding rate; u k-1is the control amount of the previous period, which is used to limit the sudden change of control action; λ is the adjustment coefficient, λ∈0.1,0.3;

[0087] A3. Use sequential quadratic programming to solve the optimal control quantity for the next 30 minutes;

[0088] A4. Use real-time pH and dissolved oxygen data to make weighted corrections to the predicted output, with a weight coefficient α∈[0.2,0.5]. The edge-cloud collaboration module 204 is used to coordinate real-time monitoring and rapid control at the edge, deep analysis and parameter optimization at the cloud. The specific operations are as follows: B1. Edge layer: Rapidly control temperature and stirring speed with an execution cycle of ≤1 second, and first-order lag filtering is used for data preprocessing; B2. Cloud layer: Receive edge data packets every 5 minutes, generate parameter optimization suggestions through digital twin simulation, and send volume ≤1KB; B3. Disaster recovery when the network is disconnected: When the network delay is greater than 300ms, the edge automatically enables the optimal parameter table cached locally and matches the control instructions according to the fermentation timestamp.

[0089] In this embodiment, it should be noted that the feature recognition module 201 analyzes the multi-dimensional parameter characteristics of the fermentation process in real time, and outputs the results to the control strategy switching module 202 for dynamic selection between fuzzy control and model predictive control. The selected strategy is executed by online optimization of parameters by the adaptive predictive control model module 203, while the edge-cloud collaboration module 204 realizes distributed collaborative computing and dynamic parameter adjustment of control instructions.

[0090] In addition, it should be noted that the specific operation steps of the control strategy switching module 202 are as follows: ① Real-time receiving the fermentation stage determination result and stability index output by the feature recognition module 201, such as the intra-class variance σ 2 ②When σ 2 < threshold and the parameter fluctuation is gentle, the fuzzy control strategy is activated; ③ When a mutation or prediction error of key parameters such as dissolved oxygen and pH is detected> 5%, it automatically switches to model predictive control; ④ The weighted fusion algorithm is used to smooth the switching process to avoid jumps in the control quantity, and the transition time is ≤ 10 seconds; ⑤ If the optimization fails for three consecutive times, an alarm is triggered and the control is switched to constant speed stirring + constant temperature control. PDI calculation formula: Among them, M W is the weight average molecular weight, M n is the number average molecular weight. s ∈[0.1,0.5]g / L; K i ∈[0.01,0.05]L / g.

[0091] In the present invention, the intelligent optimization and digital twin unit 300 includes a transfer learning module 301, a digital twin model construction module 302, a control strategy simulation and optimization module 303, and an augmented reality monitoring module 304, where: The transfer learning module 301 is used to pre-train a general fermentation model through public or historical data, and then use a small amount of new scenario data for transfer fine-tuning to adapt to the specific scenario of polycarboxylate superplasticizer fermentation; The digital twin model construction module 302 is used to construct a digital twin model based on the physical characteristics and process parameters of the actual fermentation equipment, and simulate the changes in fermentation process parameters in real time; The specific operation content of model construction is: C1. Macro scale: Establish an overall kinetic model of the reactor based on the mass-energy conservation equation; Micro scale: Use an LSTM neural network to predict the distribution of microbial metabolic fluxes; C2. Synchronize the temperature, pH, and dissolved oxygen data of the actual equipment to the twin body every 2 minutes through the OPC UA protocol; C3. When the deviation between the simulation value and the actual value continuously > 5%, trigger online parameter correction based on Kalman filtering. The control strategy simulation and optimization module 303 is used to simulate different control strategies in the digital twin model, compare the effects, recommend the optimal control parameters and strategies for actual production, and improve the fermentation efficiency and product quality; The specific operation content is: D1. Parallelly simulate the product molecular weight distribution and conversion rate in the next 1 hour of three strategies, namely conventional PID control, model predictive control, and adaptive fuzzy control, in the digital twin; D2. Use the NSGA-II algorithm to solve the Pareto optimal solution set with the constraints of PDI ≤ 1.3 and conversion rate ≥ 85%; D3. Push the top three optimization strategies and the comparison chart of expected benefits to the operator through the AR interface. The augmented reality monitoring module 304 is used to integrate virtual information with the real vision using augmented reality technology. Workers can view the parameters through AR glasses, and AI analyzes and marks abnormalities and gives suggestions.

[0092] In this embodiment, it should be noted that the transfer learning module 301 quickly adapts to the dedicated fermentation model. The digital twin model construction module 302 establishes a multi-scale virtual-real mapping model and calibrates it in real time. The control strategy simulation and optimization module 303 parallelly simulates multiple strategies in the digital twin environment and recommends the Pareto optimal solution. Finally, the augmented reality monitoring module 304 realizes visual decision-making interaction.

[0093] In addition, it should be noted that the process of the NSGA-II algorithm for solving the Pareto optimal solution set is as follows: I. Define the bi-objective optimization functions f1(x) = |PDI(x) - 1.3| and f2(x) = -Conversion(x), where x is the control variable vector; PDI(x) is the product dispersion index under strategy x; Conversion(x) is the substrate conversion rate under strategy x. II. Use dynamic non-dominated sorting to divide the solution set into levels. The basis for dividing the non-dominated front is as follows: If individual i is not inferior to individual j in all objectives and is superior to individual j in at least one objective, then individual i dominates individual j. The individuals that are not dominated by any other individuals form the first non-dominated front F1. After removing F1 from the population, the individuals among the remaining individuals that are not dominated by the other remaining individuals form the second non-dominated front F2, and so on. Calculate the crowding distance to ensure the distribution of the solution set. where M is the number of objective functions, f m (i) is the value of individual i on the m-th objective function, and i + 1, i - 1 are the adjacent solutions of individual i after sorting by objective m, is the maximum / minimum value of objective m in the current population. III. Generate a new generation of population through the elitist retention strategy, and finally output the 10 sets of solutions with the largest crowding distance on the Pareto front.

[0094] In the present invention, the safety and redundancy unit 400 includes an encryption module 401, a local cache module 402, and an emergency control module 403, where: The encryption module 401 is used to encrypt at the data sending end and decrypt at the receiving end using the AES industrial-grade encryption protocol. The local cache module 402 is used to temporarily store key data and control instructions with a certain storage capacity when the network is abnormal, and automatically upload them to the cloud after the network is restored to ensure the coherence and integrity of the data. The emergency control module 403 is used to perform emergency control on the fermentation equipment according to local rules and preset strategies in the event of a network outage, maintain the basic operating parameters, and assist the system to quickly recover after troubleshooting.

[0095] In this embodiment, it should be noted that the encryption module 401 realizes end-to-end encryption of data transmission, the local cache module 402 temporarily stores key data and maintains the continuity of instructions when the network is abnormal, and the emergency control module 403 performs emergency regulation of the equipment based on preset rules.

[0096] In the description of this specification, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0097] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A fermentation monitoring system for polycarboxylate water reducer, characterized in that, It includes a data acquisition unit (100), an adaptive decision-making control unit (200), an intelligent optimization and digital twin unit (300), and a safety and redundancy unit (400), where: The data acquisition unit (100): is used to collect fermentation parameters in real time through a wireless local area network, and perform data aggregation and upload through an edge gateway; The adaptive decision-making control unit (200): is based on a dynamic switching mechanism of fuzzy control and model predictive control for fermentation process feature recognition, combines an adaptive predictive control model with online parameter update, and adopts a collaborative architecture of real-time control at the edge and optimization of tuning parameters in the cloud; The intelligent optimization and digital twin unit (300): fine-tunes a pre-trained general fermentation model based on transfer learning to adapt to new scenarios, combines digital twin simulation control strategy effects to recommend optimal parameters, and uses augmented reality monitoring to provide anomaly point annotation and suggestions for workers; The safety and redundancy unit (400): adopts an industrial-grade encryption protocol and a local caching mechanism to ensure data security of wireless communication and emergency control in case of network disconnection.

2. The fermentation monitoring system of a polycarboxylate water reducer according to claim 1, characterized in that, The data acquisition unit (100) includes a sensor module (101), a wireless communication module (102), and an edge gateway module (103), where: The sensor module (101) is used to deploy various types of sensors at various parts of the fermentation equipment to accurately collect various parameters in the fermentation process of polycarboxylate superplasticizer in real time; The wireless communication module (102) is used to package and encode the data collected by multiple sensors through a wireless local area network and transmit it to the edge gateway module (103); The edge gateway module (103) is used to receive multi-sensor data, perform aggregation, preliminary screening, and noise reduction analysis, and then upload it to the cloud or subsequent units, and has the ability to temporarily store data in case of network anomalies.

3. A polycarboxylate water reducer fermentation monitoring system according to claim 1, characterized in that, The adaptive decision-making control unit (200) includes a feature recognition module (201), a control strategy switching module (202), an adaptive predictive control model module (203), and an edge-cloud collaboration module (204), where: The feature recognition module (201) performs real-time analysis of fermentation data based on a clustering analysis algorithm to identify fermentation stages and features to judge the fermentation state; The control strategy switching module (202) is used to adopt fuzzy control when the fermentation process is stable according to the feature recognition result, and switch to model predictive control when precise control prediction is required; The adaptive predictive control model module (203) is used to establish and update a model by combining historical and real-time data, predict the future trend of fermentation, adjust control parameters, and support online parameter update to adapt to changes; The edge-cloud collaboration module (204) is used to coordinate real-time monitoring and rapid control at the edge, and in-depth analysis and parameter optimization work in the cloud.

4. The polycarboxylate water reducer fermentation monitoring system according to claim 3, characterized in that, The clustering analysis algorithm performs real-time analysis of fermentation data to identify fermentation stages and features to judge the fermentation state. The specific formula of the clustering analysis algorithm is: The weighted Euclidean distance is used to calculate the distance between the sample point X and the center: where \(w\) j (t) is the time-varying feature weight; \(x\) j ' is the normalization parameter; Update the clustering center every 5 minutes: where s m is the set of samples of the m-th class; When the within-class variance σ of 10 consecutive samples 2 is less than 0, it is determined that the corresponding fermentation stage is entered.

5. The polycarboxylate water reducer fermentation monitoring system according to claim 4, characterized in that, The method for establishing and updating a model by combining historical and real-time data, predicting the future trend of fermentation, adjusting control parameters, and supporting online parameter update to adapt to changes is as follows: A1. Establish a state space model based on the fermentation kinetics equation. The relationship between the cell growth rate μ and the substrate concentration S is: The model parameters are updated online every hour. u is the actual specific growth rate, representing the growth amount of microorganisms per unit time; μ max is the maximum specific growth rate, the limiting value of the growth rate under ideal conditions; S is the substrate concentration, the instantaneous concentration of nutrients in the fermentation broth; K s is the half-saturation constant; K i is the inhibition coefficient; A2. Take the minimization of the variance of the product molecular weight distribution in the next 1 hour as the objective function. The expression of the objective function is: where PDI k is the product dispersion index in the k-th period, representing the uniformity of the molecular weight distribution; PDI target is the target dispersion index, the PDI value corresponding to the ideal product molecular weight distribution; u k is the control quantity vector in the k-th period; u k-1 is the control quantity in the previous period, used to limit the mutation of the control action; λ is the adjustment coefficient; A3. Use sequential quadratic programming to solve the optimal control quantity for the next 30 minutes; A4. Perform weighted correction on the predicted output through real-time pH and dissolved oxygen data, and the weight coefficient α ∈ [0.2, 0.5].

6. The fermentation monitoring system of a polycarboxylate water reducer according to claim 5, characterized in that, The method for coordinating real-time monitoring and rapid control at the edge, and in-depth analysis and parameter optimization in the cloud is as follows: B1. Edge layer: Execute rapid control of temperature and stirring speed with a cycle ≤ 1 second, and use first-order lag filtering for data preprocessing; B2. Cloud layer: Receive edge data packets every 5 minutes, generate parameter optimization suggestions through digital twin simulation, and the download volume ≤ 1KB; B3. Disconnection disaster tolerance: When the network delay > 300ms, the edge automatically enables the optimal parameter table cached locally and matches the control instructions according to the fermentation timestamp.

7. A polycarboxylate superplasticizer fermentation monitoring system according to claim 1, characterized in that, The intelligent optimization and digital twin unit (300) includes a transfer learning module (301), a digital twin model construction module (302), a control strategy simulation and optimization module (303), and an augmented reality monitoring module (304), where: The transfer learning module (301) is used to pre-train a general fermentation model through public or historical data, and then use a small amount of new scenario data for transfer fine-tuning to adapt to the specific scenario of polycarboxylate superplasticizer fermentation; The digital twin model construction module (302) is used to construct a digital twin model based on the physical characteristics and process parameters of the actual fermentation equipment, and simulate the changes in fermentation process parameters in real time; The control strategy simulation and optimization module (303) is used to simulate different control strategies in the digital twin model, compare the effects, recommend the optimal control parameters and strategies for actual production, and improve the fermentation efficiency and product quality; The augmented reality monitoring module (304) is used to integrate virtual information with the real vision by using augmented reality technology. Workers can view the parameters through AR glasses, and AI analyzes and marks abnormalities and gives suggestions.

8. A polycarboxylate superplasticizer fermentation monitoring system according to claim 7, characterized in that, Construct a digital twin model based on the physical characteristics and process parameters of the actual fermentation equipment, and simulate the changes in fermentation process parameters in real time. The specific operation content of model construction is as follows: C1. Macro scale: Establish an overall kinetic model of the reactor based on the mass-energy conservation equation; Micro scale: Use an LSTM neural network to predict the distribution of the cell metabolic flux; C2. Synchronize the temperature, pH, and dissolved oxygen data of the actual equipment to the twin body every 2 minutes through the OPC UA protocol; C3. When the deviation between the simulation value and the actual value continuously > 5%, trigger online parameter correction based on Kalman filtering.

9. A polycarboxylate water reducer fermentation monitoring system according to claim 7, characterized in that, Simulate different control strategies in the digital twin model, compare the effects, recommend the optimal control parameters and strategies for actual production, and improve the fermentation efficiency and product quality. The specific operation content is: D1. Parallelly simulate the product molecular weight distribution and conversion rate in the next 1 hour of three strategies, namely conventional PID control, model predictive control, and adaptive fuzzy control, in the digital twin; D2. With the constraints of PDI ≤ 1.3 and conversion rate ≥ 85%, use the NSGA-II algorithm to solve the Pareto optimal solution set; D3. Push the comparison chart of the top three optimization strategies and expected benefits to the operator through the AR interface.

10. A polycarboxylate water reducer fermentation monitoring system according to claim 1, characterized in that, The safety and redundancy unit (400) includes an encryption module (401), a local cache module (402), and an emergency control module (403), where: The encryption module (401) is used to encrypt at the data sending end and decrypt at the receiving end using the AES industrial-grade encryption protocol; The local cache module (402) is used to temporarily store key data and control instructions with a certain storage capacity during network anomalies, and automatically upload them to the cloud after the network is restored to ensure the coherence and integrity of the data; The emergency control module (403) is used to implement emergency control of the fermentation equipment according to local rules and preset strategies in the event of a network outage, maintain basic operating parameters, and assist the system to quickly recover after troubleshooting.

Images