Intelligent responsive adaptive reaction system for microplastic degradation

Abstract

An adaptive reaction system for intelligent responsive microplastic degradation, relating to the field of microplastic pollution control technology, addresses the problem of rigid model cognition in existing machine learning models. It includes a sample input and pretreatment module, a plasma degradation module, a photocatalytic degradation module, an intelligent response control module, and a product output and monitoring module. The sample input and pretreatment module receives microplastic samples and outputs a homogeneous microplastic suspension. The plasma degradation module receives the microplastic suspension and outputs a reaction mixture containing intermediate degradation products. The photocatalytic degradation module receives the reaction mixture and outputs the final product. The intelligent response control module incorporates a metacognitive monitor and a controlled exploration unit. The product output and monitoring module receives the final product and outputs degradation product and efficiency data, feeding the data back to the intelligent response control module. This invention solves the problem of rigid model cognition in existing machine learning models.

Description

Technical Field

[0001] This invention relates to the field of microplastic pollution control technology, specifically to an intelligent responsive adaptive reaction system for microplastic degradation. Background Technology

[0002] Currently, microplastic degradation technologies mainly rely on physical, chemical, and biological methods, among which plasma degradation and photocatalytic degradation have attracted widespread attention due to their high efficiency and lack of secondary pollution. In recent years, the industry has focused on developing synergistic degradation systems, combining multiple technologies to improve degradation efficiency and energy utilization. For example, plasma technology can rapidly cleave microplastic macromolecules using high-energy active species, while photocatalysis utilizes photogenerated electron-hole pairs to further mineralize intermediate products, achieving deep degradation. With the development of intelligent control technology, machine learning models have been introduced into degradation systems to optimize reaction parameters in real time, thereby improving the system's adaptability and operational efficiency. However, existing systems still face problems such as unstable synergistic mechanisms and lag in parameter adjustment, limiting their performance and reliability in practical applications.

[0003] Existing technology systems generally include a sample input and pretreatment module, a plasma degradation module, a photocatalytic degradation module, an intelligent response control module, and a product output and monitoring module. The intelligent response control module adjusts the operating parameters of the plasma and photocatalytic modules in real time based on sensor data to optimize the degradation process. In practice, it has been found that the machine learning model in the intelligent response control module, due to its inherent reward mechanism, over-relies on and reinforces a single, successfully successful pattern—a fixed set of parameters—and gradually loses its ability to explore potentially better synergistic states. The root of this problem lies in the fact that, in long-term operation, the machine learning model tends to repeat historically successful strategies, neglecting the exploration of new parameter spaces, leading to model cognitive rigidity. In practical work, this problem can cause a series of negative impacts: First, the system cannot adapt to subtle changes in microplastic composition or environmental conditions. For example, when the source of microplastics differs or catalyst activity declines, the degradation efficiency slowly decreases and is difficult to recover through routine calibration. Second, energy utilization decreases because the system adheres to suboptimal parameter combinations and cannot dynamically adjust to achieve optimal synergy between plasma and photocatalysis. Finally, long-term performance degradation may lead to the accumulation of incomplete degradation products, increasing the risk of environmental pollution. These consequences not only weaken the system's reliability and economy but also limit its application in complex real-world scenarios. Therefore, a solution is urgently needed to overcome this technical problem, ensuring the system can continuously explore optimal operating states and maintain efficient and stable degradation performance.

[0004] In existing technologies, there is a lack of research on improving model adaptability through metacognitive monitoring and controlled exploration units, leading to the inability of the system to effectively avoid cognitive solidification problems during long-term operation. This invention addresses this technological gap by proposing an innovative intelligent response control mechanism. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent responsive adaptive reaction system for microplastic degradation, which addresses the problem that machine learning models in existing systems tend to repeat historically successful strategies during long-term operation, neglecting the exploration of new parameter spaces, leading to model cognitive rigidity.

[0006] The technical solution adopted by this invention to solve its technical problem is: an adaptive reaction system for intelligent responsive microplastic degradation, comprising a sample input and pretreatment module, a plasma degradation module, a photocatalytic degradation module, an intelligent response control module, and a product output and monitoring module connected in sequence; the sample input and pretreatment module receives microplastic samples and outputs a uniform microplastic suspension to the plasma degradation module; the plasma degradation module receives the microplastic suspension and outputs a reaction mixture containing intermediate degradation products to the photocatalytic degradation module; the photocatalytic degradation module receives the reaction mixture and outputs a final product containing small molecule products to the product output and monitoring module; the intelligent response control module receives real-time reaction parameters from the plasma degradation module and the photocatalytic degradation module, performs intelligent analysis, and outputs control commands to the plasma degradation module and the photocatalytic degradation module; the intelligent response control module incorporates a metacognitive monitor and a controlled exploration unit; the product output and monitoring module receives the final product and outputs degradation product and efficiency data, and feeds the data back to the intelligent response control module.

[0007] Furthermore, the metacognitive monitor calculates the policy entropy value. The cognitive stability of the model is quantified; among which, The probability of the i-th control instruction type; when the entropy value Below the threshold At that point, the metacognitive monitor determined that the model had fallen into a state of cognitive rigidity.

[0008] Furthermore, the metacognitive monitor performs trend analysis on the degradation efficiency sequence in the performance data, calculating the rate of change over consecutive time intervals. ;in, Let be the rate of change of degradation efficiency at time t. and These are the degradation efficiency values ​​at adjacent sampling times. This is the sampling time interval. If When the absolute value of the value remains below 0.01% / second, the metacognitive monitor judgment model also falls into a state of cognitive solidification.

[0009] Furthermore, when the model falls into a state of cognitive rigidity, the metacognitive monitor generates an exploration trigger signal. Explore trigger signals After being input into the controlled exploration unit as output data, it is mapped into a set of exploratory instructions. ,in, The amplitude is constrained within the safe range based on the system stability boundary function; the parameters output by the controlled exploration unit temporarily cover the initial control commands of the core control model.

[0010] Furthermore, upon receiving the exploration trigger signal, the controlled exploration unit immediately activates the disturbance generation mechanism to perform restricted random disturbance operations on the preliminary control commands output by the core control model. The disturbance algorithm constructs exploratory parameter commands by introducing controlled changes based on the current parameter values, and its mathematical expression is: ;in, Let j be the j-th control parameter after the disturbance. These are the initial control parameters. The disturbance factor is used; if the disturbance result exceeds the safety threshold, the system automatically executes the boundary constraint correction function to limit the power parameter to below 500 watts and the light intensity parameter to below 100 milliwatts per square centimeter; the generated exploratory parameter command replaces the original command of the core control model and is sent to the plasma degradation module and photocatalytic degradation module within 5 to 10 minutes after triggering, so that the two can operate in coordination under the new control space.

[0011] Furthermore, the controlled exploration unit continuously collects real-time reaction data and monitors changes in degradation efficiency. The collected dataset, along with perturbation parameters, is input into the core control model for online retraining. The retraining process minimizes the performance error function. ;in, For retraining losses, Number of sampling times To measure the degradation efficiency, To improve the model's prediction efficiency; after retraining, the controlled exploration unit outputs updated control commands as new decision references and returns them to the core control model.

[0012] The beneficial effects of this invention are: (1) By introducing an innovative mechanism that combines a metacognitive monitor with a controlled exploration unit, the problem of cognitive solidification caused by excessive reliance on historical successful patterns in long-term operation of machine learning models is solved, enabling the system to continuously explore a better parameter space while maintaining stability, thereby achieving continuous optimization of degradation efficiency. (2) By using the metacognitive monitoring mechanism to monitor the diversity of control instructions and the trend of system performance in parallel, the model solidification trend can be identified in the early stage, and proactive intervention can be made before the system performance declines significantly, effectively preventing the occurrence of collaborative deharmonicity and ensuring that the degradation system always maintains the optimal working state. (3) By incorporating an exploratory training data dynamic update mechanism into the improved machine learning model training method, the model can continuously learn and adapt to system changes, overcoming the model aging problem caused by traditional fixed dataset training, and significantly improving the adaptability and robustness of the system under different working conditions. (4) By constructing a complete technical chain from parameter acquisition, intelligent decision-making, execution feedback to product monitoring, intelligent collaborative regulation of plasma and photocatalytic degradation processes is realized, solving the problem of model cognitive solidification while ensuring that the system degradation efficiency is increased by more than 20% and energy consumption is reduced by more than 15%. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall architecture of the reaction system of the present invention;

[0014] Figure 2 This is a schematic diagram of the internal structure of the intelligent response control module of the present invention;

[0015] Figure 3 This is a schematic diagram of the process flow for the sample input and pretreatment module.

[0016] Figure 4 This is a flowchart of the plasma-photocatalytic synergistic degradation reaction. Detailed Implementation

[0017] The reaction system of this invention first inputs and pre-treats the sample; then it performs plasma degradation and photocatalytic degradation; simultaneously, the intelligent response control module receives real-time reaction parameters from the plasma degradation module and the photocatalytic degradation module, and intelligently analyzes and outputs control commands to the plasma degradation module and the photocatalytic degradation module to achieve dynamic optimization. The intelligent response control module introduces a metacognitive monitor and a controlled exploration unit to prevent the model from losing its exploration ability due to over-reliance on historical successful patterns; finally, the product output and monitoring module receives the final product and outputs degradation product and efficiency data, and feeds the data back to the intelligent response control module to form a closed loop. This invention ensures that the reaction system can continuously explore better operating states and maintain efficient and stable degradation performance.

[0018] This invention discloses an adaptive reaction system for intelligent responsive microplastic degradation, comprising a sample input and pretreatment module, a plasma degradation module, a photocatalytic degradation module, an intelligent response control module, and a product output and monitoring module connected in sequence. The sample input and pretreatment module receives microplastic samples and outputs a uniform microplastic suspension to the plasma degradation module. The plasma degradation module receives the microplastic suspension and outputs a reaction mixture containing intermediate degradation products to the photocatalytic degradation module. The photocatalytic degradation module receives the reaction mixture and outputs a final product containing small molecule products to the product output and monitoring module. The intelligent response control module receives real-time reaction parameters from the plasma degradation module and the photocatalytic degradation module, intelligently analyzes them, and outputs control commands to the plasma degradation module and the photocatalytic degradation module to achieve dynamic optimization. The intelligent response control module incorporates a metacognitive monitor and a controlled exploration unit to prevent loss of exploration capabilities due to over-reliance on historical success patterns. The product output and monitoring module receives the final product and outputs degradation product and efficiency data, feeding the data back to the intelligent response control module.

[0019] I. Sample Input and Preprocessing Module.

[0020] like Figure 3 As shown, the sample input and pretreatment module undertakes the initial standardization task of microplastic samples in the system. Its processing directly affects the efficiency and stability of subsequent plasma and photocatalytic synergistic degradation. The sample input and pretreatment module includes a crushing unit, a suspension preparation unit, and a delivery pump.

[0021] Microplastic samples collected from the external environment first enter the crushing unit, which is equipped with a high-speed rotating cutter disc and a fixed sieve plate. Through mechanical shearing and friction, the original plastic fragments are gradually ground into micron-sized particles with a diameter between 1 micrometer and 100 micrometers. Particle size control is adjusted based on the dynamic feedback of the particle size distribution function. The particle size uniformity index is monitored by a real-time particle size analyzer and must meet the condition that the standard deviation does not exceed a set threshold.

[0022] After grinding, the particles are conveyed through a pipeline into the suspension preparation unit. This unit is equipped with a stirring system and a liquid distribution system, forming a uniformly dispersed suspension system through physical mixing of the liquid medium and the particles. The energy input for the mixing process is determined by both the stirring rate and the fluid viscosity. The suspension concentration is controlled between 0.1 g / L and 10 g / L to ensure the stability of the particle distribution and the uniformity of subsequent plasma interaction. The particle distribution state during stirring can be characterized by the average dispersion function, defined as... .in, Indicates dispersion. The number of sampling points. Let be the local particle concentration at the i-th sampling point. This represents the overall average concentration of the system. (This is achieved through monitoring...) By controlling the variation trend and maintaining it within a stable range, the uniformity of the suspension on a macroscopic scale can be ensured. The mixed suspension is continuously output via a delivery pump at a flow rate of 0.5 mL / min to 5 mL / min, with flow control achieved by a closed-loop pump speed regulation system to maintain fluid stability input to the plasma degradation module. This entire process constitutes a continuous pathway from solid sample to fluid input, ensuring standardization of particle size and concentration, resulting in a more uniform energy distribution and more efficient action of active species in the subsequent plasma reaction, thereby improving the overall reaction efficiency and reproducibility of the degradation system.

[0023] II. Plasma Degradation Module.

[0024] The plasma degradation module undertakes the initial pyrolysis and oxidation of the microplastic suspension. The plasma degradation module includes a gas supply unit, a temperature control unit, and a low-temperature plasma reaction chamber.

[0025] The low-temperature plasma reaction chamber is made of corrosion-resistant materials and contains parallel electrode pairs and a gas distributor to ensure the uniformity of the electric field and gas flow. The input microplastic suspension is uniformly sprayed into the reaction zone through the inlet, where it interacts with the ionized gas in a strong electric field environment. The electrode pairs are connected to a plasma generator with an output power between 50 and 500 watts. Frequency adjustment within the range of 10 kHz to 50 kHz achieves a balance between discharge stability and energy density.

[0026] The gas supply unit continuously delivers argon, oxygen, or air to the gas distributor at a flow rate controlled between 1 and 10 liters per minute. The type and concentration of reactive species are adjusted by regulating the gas ratio. During the reaction, high-energy electrons collide with gas molecules to generate reactive particles, including OH, NO2, and HO2. These particles break down the molecular chains on the surface of the microplastics, leading to the degradation of the polymer carbon chains into aldehyde and ketone intermediates. The reaction rate is determined by both the energy input density and the reactant concentration. .in, This represents the energy input density per unit volume, expressed in joules per liter. This refers to plasma discharge power, measured in watts. The flow rate of the reactant gas is expressed in liters per second. Energy input density characterizes the degree of efficient energy transfer to the reaction system. When the microplastic chain segments are stabilized within the optimal range, the fracture rate and oxidation rate of the microplastic chain segments reach equilibrium, thereby achieving efficient energy utilization at the molecular level.

[0027] The temperature control unit maintains the reaction temperature between 20°C and 60°C to suppress side reactions caused by overheating and stabilize free radical lifetimes. After plasma treatment, the macromolecular polymers in the suspension are converted into low molecular weight intermediates, forming a reaction mixture containing aldehydes and ketones. This mixture flows continuously through the outlet to the photocatalytic degradation module, realizing dynamic connection between reaction stages and forming a continuous degradation chain from polymer chain breakage to deep oxidation of intermediates.

[0028] III. Photocatalytic Degradation Module.

[0029] The photocatalytic degradation module performs deep mineralization and detoxification of the reaction mixture from the plasma degradation module. Its operation is based on a synergistic mechanism of photo-energy excitation and surface reaction within the photocatalytic reactor. The photocatalytic degradation module includes a photocatalytic reactor and a light source unit.

[0030] The input reaction mixture, upon entering the photocatalytic reactor, is first stirred to create a stable turbulent flow, ensuring a uniform distribution of suspended photocatalytic particles in the liquid flow field and thus increasing the reaction interface area. The photocatalytic reactor is made of a highly transparent material to ensure sufficient irradiance flux from the incident light. The photocatalyst is uniformly loaded on the inner wall of the reactor or suspended in the liquid phase. Under light irradiation, the active sites on the catalyst surface are excited to generate electron-hole pairs. Photogenerated electrons migrate to the conduction band, while photogenerated holes remain in the valence band, creating a strong redox environment. The electrons and holes react with oxygen and water molecules adsorbed on the catalyst surface, respectively, to generate superoxide anion radicals and hydroxyl radicals. These highly reactive particles undergo a chain oxidation reaction with intermediate degradation product molecules, causing carbon chains to break and gradually transforming into low-molecular-weight organic acids. The pH of the reaction system is maintained between 3 and 9 by an automatic titration adjustment unit to maintain a stable charge state on the catalyst surface, thereby ensuring the separation of photogenerated carriers and the continuity of the reaction.

[0031] The light source unit provides stable output based on the set wavelength and intensity. Ultraviolet light, within the wavelength range of 200 to 400 nanometers, effectively excites the bandgap energy of titanium dioxide or zinc oxide. Under visible light irradiation, the composite catalyst maintains high electron-hole pair migration efficiency through energy level modulation. During the reaction, light intensity and stirring rate jointly determine the spatial uniformity of free radical distribution in the liquid phase, directly affecting the continuity of the degradation pathway and the final degree of mineralization. As the reaction progresses, aldehydes and ketone intermediates in the system are gradually oxidized to carbon dioxide, water, and a small amount of short-chain organic acids. The product concentration tends to stabilize over time, indicating that the reaction has reached the mineralization endpoint. The final product after photocatalytic reaction is continuously transported to the product output and monitoring module. The entire microplastic degradation process maintains a dynamic balance in terms of energy input, reaction kinetics, and material transformation, realizing a continuous transformation chain from initial pyrolysis to complete mineralization.

[0032] IV. Intelligent Response Control Module.

[0033] The intelligent response control module, serving as the core of the system's dynamic optimization, maintains efficient coordination between plasma and photocatalytic reaction processes through real-time data fusion and a cognitive-level control mechanism. The intelligent response control module comprises a sensor array, a core control module, a metacognitive monitor, and a controlled exploration unit.

[0034] like Figure 1 As shown, the intelligent response control module first receives real-time reaction parameter inputs from the plasma degradation module and the photocatalytic degradation module, including the reaction temperature. Plasma power Light intensity and the concentration of intermediate degradation products .like Figure 2 As shown, these parameters are synchronously acquired through a multi-point distributed sensor array, and the signals are time-aligned and noise-filtered to form feature vectors. The input is processed by the core control model. The core control model constructs a state-action mapping function based on machine learning algorithms. Output control command vector These correspond to parameters for power adjustment, light intensity adjustment, gas flow rate adjustment, and solution pH adjustment, respectively. The performance loss function is minimized. Achieve control over energy density Gas-liquid ratio Deviation from intermediate product concentration Multi-objective constrained optimal control, where * denotes the corresponding optimal value of the objective. These are the weighting coefficients. The core control model continuously updates the policy distribution during iterative training. Strategy distribution Indicates the state Next selection control command The probability of.

[0035] The metacognitive monitor maintains a connection with the core control model, and its function is to analyze the diversity and exploration degree of the model's output policies in real time by calculating the policy entropy value. The cognitive stability of the model is quantified. Among other things, This represents the probability of the i-th type of control instruction. When the entropy value... Below the threshold When the absolute value of the rate of change in degradation efficiency remains below 0.01% / second, the metacognitive monitor determines that the model has fallen into a state of cognitive solidification and generates an exploration trigger signal. The exploration trigger signal, after being input into the controlled exploration unit, is mapped into a set of exploratory commands. ,in, The amplitude is constrained within a safe range by the system stability boundary function to ensure that no reaction instability occurs during the exploration process. The parameters output by the controlled exploration unit temporarily overwrite the initial control commands of the core control model, allowing the system to obtain cognitive stimulation of parameter perturbations while maintaining safety, and to achieve active resampling of the reaction space. After feedback from the exploration phase, the intelligent response control module adjusts the control weights according to the trend of degradation efficiency changes and the convergence speed of the loss function, so that the system maintains a dynamic balance between utilizing existing knowledge and exploration. The final output integrated control command is transmitted to the plasma degradation module and the photocatalytic degradation module via the signal bus, driving their operating power, illumination conditions and hydrodynamic parameters to be adjusted synchronously, thereby maintaining the optimal state of energy coupling, reaction rate and product stability, and preventing the problem of strategy simplification caused by the control model's over-reliance on historical experience through a metacognitive monitoring mechanism.

[0036] The following section describes the detailed working mechanism of each unit in the intelligent response control module.

[0037] (1) Sensor array.

[0038] like Figure 2 As shown, the sensor array, as the core sensing subunit of the intelligent response control module, achieves real-time data input to the reaction system through multi-source synchronous acquisition and signal standardization conversion. The array is distributed at key locations in the plasma reaction chamber and photocatalytic reactor to ensure the representativeness and temporal consistency of the acquired data. The sensor array includes thermocouple sensors, a power meter, a photometer, and a spectrometer.

[0039] Thermocouple sensors are embedded in the reaction liquid region and at the gas-liquid interface to acquire temperature signals. Its output voltage is proportional to the temperature difference, and the real-time temperature value is obtained after linear calibration.

[0040] The power meter is connected to the output of the plasma generator to measure instantaneous power. Instantaneous power It reflects the dynamic effect of energy input on the density of reactive particles.

[0041] A photometer was placed at the light-transmitting window on the outer wall of the reactor to measure the light intensity. It is used to monitor photocatalytic excitation efficiency.

[0042] The spectrometer receives real-time spectral signals from the photocatalytic reaction solution, and calculates the concentration of intermediate degradation products using absorption peak identification and integration algorithms. .

[0043] Each unit of the sensor array at the sampling frequency Within a range of periods Input parameters are acquired synchronously, and the raw signals are converted into voltage signal matrices by an analog-to-digital converter module. Then, a standardization process is performed to give different physical quantities a consistent scale. The standardization process is defined as follows: .in, Let represent the measured value of the original sensor signal of type i at time t. This is the moving average of the signal. Its standard deviation, This is the standardized signal value. This processing makes the outputs of various sensors comparable in the numerical domain and eliminates biases caused by different measurement units. The standardized data is transmitted in real time to the core control model and metacognitive monitor via a high-bandwidth data bus, forming a multi-dimensional feature input vector. Upon receiving the vector, the core control model calculates the optimal control command, while the metacognitive monitor determines the system's dynamic stability by continuously monitoring the rate of change of the input distribution. The output of the entire sensor array exists in the form of a standardized parameter dataset, ensuring that subsequent control and optimization processes maintain temporal consistency and signal reliability at the information level, thereby providing a high-precision sensing foundation and continuous data support for the entire intelligent response control system.

[0044] (2) Core control model.

[0045] The core control model plays a crucial role in real-time decision-making and dynamic adjustment within the intelligent response control module. Its operation consists of four stages: input reception, feature extraction, weight update, and control command generation. For example... Figure 2 As shown, the model first receives a normalized parameter dataset from the sensor array, which includes temperature, plasma power, light intensity, and intermediate product concentration. These parameters are mapped to the node vectors of the input layer. The input layer normalizes the original signal before passing it to the hidden layer. The neurons in the hidden layer use a linear rectified activation function to achieve a non-linear mapping, expressed as follows: .in, The output of the j-th hidden layer neuron. These are the weight parameters between the input layer and the hidden layer. For input parameter components, This is the bias term. This process achieves a weighted combination of input features through nonlinear activation, enabling the model to maintain sensitivity to multidimensional coupled features under different reaction conditions. The output layer performs a weighted summation of all hidden layer results and generates a predicted degradation efficiency value. Predicted degradation efficiency This corresponds to a comprehensive degradation capacity index based on the synergistic reaction of photocatalysis and plasma. During the training phase, the model uses historical datasets for weight updates, employing mean squared error as the loss function, defined as... .in, For the total loss, The actual degradation efficiency of the i-th sample. For the corresponding predicted value, This represents the total number of training samples. The model minimizes the loss function using the gradient descent algorithm, for each parameter... The update follows Here, η is the learning rate, and the model parameters gradually converge through multiple rounds of iterative training. The core control model continuously updates the policy distribution during the iterative training process. Strategy distribution Indicates the state Next selection control command The probability of this is determined. After training, the model is deployed in a real-time control environment, continuously receiving the latest data stream from the sensor array and generating preliminary control commands, including power adjustment signals and light intensity correction values. The output results are analyzed by the metacognitive monitor and the controlled exploration unit, and then fed back to form the final control decision, thereby enabling the degradation system to adaptively adjust and stably operate in a dynamic environment.

[0046] (3) Metacognitive monitor.

[0047] The metacognitive monitor plays a role in self-reflection and cognitive correction within the intelligent response control module. Its operation relies on the joint analysis of historical and real-time performance feedback from the core control model to identify states of convergent decision-making and efficiency stagnation. For example... Figure 2 As shown, firstly, the metacognitive monitor receives historical data of preliminary control commands from the core control model and real-time performance parameters from the sensor array, forming a multi-dimensional time series input. The system then performs statistical modeling on the control command sequence, normalizing the frequency of occurrence of various control commands to form a set of probability distributions. The diversity level of the current control strategy is measured by calculating the strategy entropy value, which is defined as follows: .in, The policy entropy value. Let H be the relative frequency of the i-th type of control instruction in the historical record. The logarithmic weighted sum of the instruction distribution reflects the breadth of the system's exploration in the control space; a higher entropy value indicates greater decision diversity. When the system tends to use a fixed type of instruction for a long time, the entropy value decreases significantly. The metacognitive monitor continuously monitors the time-varying curve of the policy entropy value. If H is below the threshold of 0.5 and remains below it for more than ten minutes, the system is judged to be in a state of cognitive solidification. Simultaneously, the metacognitive monitor performs trend analysis on the degradation efficiency sequence in the performance data, calculating the rate of change over consecutive time intervals. .in, Let be the rate of change of degradation efficiency at time t. and These represent the degradation efficiency values ​​at adjacent sampling times. This is the sampling time interval. If If the absolute value of H remains consistently below 0.01% / second, it indicates that the system is in a state of cognitive rigidity. The metacognitive monitor detects H below the threshold of 0.5 and... When the absolute value of the plasma remains below 0.01% / second, an exploration trigger signal is generated when any condition is met. This signal is transmitted as output data to the controlled exploration unit to guide the model in strategy perturbation and parameter re-optimization. Through this mechanism, the system possesses self-examination and strategy regeneration capabilities during long-term operation, ensuring the continuous adaptability and stable efficiency of the plasma-photocatalytic co-degradation process under multivariable environments.

[0048] (4) Controlled exploration unit.

[0049] The controlled exploration unit in the intelligent response control module is responsible for breaking cognitive rigidity and uncovering potential optimal parameters. Its operation is based on the dynamic fusion of dual inputs from the core control model and the metacognitive monitor. For example... Figure 2 As shown, upon receiving the exploration trigger signal, the controlled exploration unit immediately activates the disturbance generation mechanism, performing restricted random disturbance operations on the initial control commands output by the core control model. The disturbance algorithm constructs exploratory parameter commands by introducing controlled changes based on the current parameter values; its mathematical expression is: .in, Let j be the j-th control parameter after the disturbance. These are the initial control parameters. The perturbation factor follows a uniform distribution within the interval [-0.2, 0.2] to ensure that the parameter adjustment range is within ±20%. If the perturbation result exceeds the safety threshold, the system automatically executes the boundary constraint correction function to limit the power parameter to below 500 watts and the light intensity parameter to below 100 milliwatts per square centimeter to maintain the operational safety of the reaction system. The generated exploratory parameter commands are sent to the plasma degradation module and photocatalytic degradation module within 5 to 10 minutes after triggering, replacing the original commands of the core control model, enabling them to operate collaboratively in the new control space. During this stage, the controlled exploratory unit continuously collects real-time reaction data and monitors changes in degradation efficiency, inputting the collected dataset and perturbation parameters into the core control model for online retraining. The retraining process minimizes the performance error function. .in, For retraining losses, Number of sampling times To measure the degradation efficiency, To improve model prediction efficiency, the controlled exploration unit outputs updated control commands after retraining. These commands serve as new decision references and are returned to the core control model, enabling adaptive expansion of the parameter space and continuous optimization of the performance surface. Through this closed-loop process, the system gradually improves the synergistic degradation efficiency of plasma and photocatalysis in multiple runs, avoiding the control model from getting trapped in local optima, thereby maintaining high dynamic responsiveness and continuous learning capability.

[0050] V. Product Output and Monitoring Module.

[0051] The product output and monitoring module is responsible for the final collection and closed-loop monitoring of the microplastic degradation process. Its operation focuses on the separation, detection, and information feedback of the final product. The product output and monitoring module includes a product collector, a filtration device, an online analyzer, and an analysis unit.

[0052] like Figure 1 As shown, the final product from the photocatalytic degradation module undergoes solid-liquid separation after entering the product output and monitoring module. The product collector is made of high-strength chemical corrosion-resistant material, and its inner wall is treated with an inert coating to prevent product adsorption and cross-contamination.

[0053] The filtration device employs a multi-stage pore structure to trap unreacted solid residues or photocatalyst particles as the fluid passes through, keeping the solution portion highly clear. The filtered liquid then enters a storage chamber equipped with temperature control and sealing functions to ensure the stability of the product composition and the repeatability of the test results.

[0054] The online analyzer is connected to the collector outlet via a micro-flow path, enabling real-time sampling and analysis of the final product.

[0055] The analysis unit employs gas chromatography-mass spectrometry (GC-MS) or infrared spectroscopy to identify molecular fragment structures and quantitatively calculate product component proportions through separation, ionization, and signal analysis. The system monitors key parameters including degradation rate, carbon dioxide production, and the proportion of small-molecule organic acids; these parameters constitute the core indicators of degradation efficiency. After data acquisition, the signal processing module performs noise suppression and feature extraction to generate a standardized degradation efficiency report. The data output interface handles information transmission, transferring analysis results to an external storage system for long-term archiving and traceability management, and feeding them back to the intelligent response control module to form a closed-loop data system.

[0056] Upon receiving degradation efficiency data, the intelligent response control module performs model correction and long-term trend analysis, adjusting future control strategies based on historical performance curves to ensure system stability and adaptability during multiple rounds of operation. The entire product output and monitoring process achieves continuous operation from reaction completion to performance feedback, ensuring the safe collection of degradation products and quantitative evaluation of conversion effects. The final output data report serves as a crucial basis for verifying the overall performance of the synergistic degradation system and the effectiveness of optimization strategies, thereby guaranteeing that the entire intelligent responsive plasma-photocatalytic synergistic degradation system maintains a highly efficient, stable, and traceable operating state under a closed-loop control structure over the long term.

Claims

1. An adaptive reaction system for intelligent responsive microplastic degradation, characterized in that, The system comprises a sample input and pretreatment module, a plasma degradation module, a photocatalytic degradation module, an intelligent response control module, and a product output and monitoring module, connected in sequence. The sample input and pretreatment module receives microplastic samples and outputs a uniform microplastic suspension to the plasma degradation module. The plasma degradation module receives the microplastic suspension and outputs a reaction mixture containing intermediate degradation products to the photocatalytic degradation module. The photocatalytic degradation module receives the reaction mixture and outputs a final product containing small molecule products to the product output and monitoring module. The intelligent response control module receives real-time reaction parameters from the plasma degradation module and the photocatalytic degradation module, performs intelligent analysis, and outputs control commands to the plasma degradation module and the photocatalytic degradation module. The intelligent response control module incorporates a metacognitive monitor and a controlled exploration unit. The product output and monitoring module receives the final product and outputs degradation product and efficiency data, and feeds the data back to the intelligent response control module.

2. The adaptive reaction system for intelligent responsive microplastic degradation according to claim 1, characterized in that, Metacognitive monitors calculate policy entropy values. The cognitive stability of the model is quantified; among which, The probability of the i-th control instruction type; when the entropy value Below the threshold At that point, the metacognitive monitor determined that the model had fallen into a state of cognitive rigidity.

3. The adaptive reaction system for intelligent responsive microplastic degradation according to claim 2, characterized in that, The metacognitive monitor performs trend analysis on the degradation efficiency sequence in the performance data and calculates the rate of change over consecutive time periods. ;in, Let be the rate of change of degradation efficiency at time t. and These are the degradation efficiency values ​​at adjacent sampling times. The sampling time interval; if When the absolute value of the value remains below 0.01% / second, the metacognitive monitor judgment model also falls into a state of cognitive solidification.

4. The adaptive reaction system for intelligent responsive microplastic degradation according to claim 3, characterized in that, When the model falls into a state of cognitive solidification, the metacognitive monitor generates an exploration trigger signal. Explore trigger signals After being input into the controlled exploration unit as output data, it is mapped into a set of exploratory instructions. ,in, The amplitude is constrained within the safe range based on the system stability boundary function; the parameters output by the controlled exploration unit temporarily cover the initial control commands of the core control model.

5. The adaptive reaction system for intelligent responsive microplastic degradation according to claim 4, characterized in that, Upon receiving the exploration trigger signal, the controlled exploration unit immediately activates the disturbance generation mechanism, performing restricted random disturbance operations on the initial control commands output by the core control model. The disturbance algorithm constructs exploratory parameter commands by introducing controlled changes based on the current parameter values; its mathematical expression is: ;in, Let j be the j-th control parameter after the disturbance. These are the initial control parameters. The disturbance factor is used; if the disturbance result exceeds the safety threshold, the system automatically executes the boundary constraint correction function to limit the power parameter to below 500 watts and the light intensity parameter to below 100 milliwatts per square centimeter; the generated exploratory parameter command replaces the original command of the core control model and is sent to the plasma degradation module and photocatalytic degradation module within 5 to 10 minutes after triggering, so that the two can operate in coordination under the new control space.

6. The adaptive reaction system for intelligent responsive microplastic degradation according to claim 5, characterized in that, The controlled exploration unit continuously collects real-time reaction data and monitors changes in degradation efficiency. The collected dataset and perturbation parameters are then input into the core control model for online retraining. Minimize the performance error function during retraining ;in, For retraining losses, Number of sampling times To measure the degradation efficiency, To improve the model's prediction efficiency; after retraining, the controlled exploration unit outputs updated control commands as new decision references and returns them to the core control model.

Images