Page 48 - 《橡塑技术与装备》英文版2026年2期

HINA R&P TECHNOLOGY AND EQUIPMENT

score, confusion matrix, etc. Depending on the evaluation the nature of the problem and the characteristics of the data,
results, it may be necessary to adjust model parameters or try and model parameters are verified and adjusted to optimize
different feature combinations. Through integrated learning model performance. The predictive model is integrated with
methods, such as online learning or incremental learning, the pyrolysis process control system to achieve integrated
the model can continuously optimize itself based on new optimization. The performance of the predictive system is
data. The classification results output by the model should be continuously monitored, and feedback data is collected to
interpretable and used to guide practical operations. evaluate the long-term effectiveness of the model. The model
1.10 Development strategy for basic theories is updated and optimized based on new data and feedback to
of environmental impact assessment adapt to changes in the process and environmental conditions.
By collecting environmental impact data related to Develop an AI system capable of predicting and
waste plastic pyrolysis, we establish a model framework for optimizing the resource recovery rate during the pyrolysis
environmental impact assessment. We select an appropriate process of waste plastics, thereby enhancing resource
AI model to predict and assess environmental impacts, train utilization efficiency and reducing environmental pollution.
the model using historical datasets, and verify its accuracy and 1.12 Development strategy for basic theory of
generalization ability through methods such as cross-validation equipment maintenance and fault prediction
and leave-one-out. This ensures that the model can provide By collecting historical operational data and real-time
reliable predictions across different datasets. The trained operating status of equipment, select features that have a
model is applied to actual waste plastic pyrolysis processes to significant impact on equipment fault prediction and create
assess environmental impacts under various conditions. The new features. Connect the real-time collected data with the
model can predict direct and indirect environmental impacts prediction model to achieve real-time fault prediction. Integrate
under different pyrolysis parameters, helping decision-makers the prediction system with existing equipment management
optimize the pyrolysis process and reduce environmental systems and workflow systems to ensure smooth data flow.
burden. With technological advancements and the accumulation Deploy the prediction system in the actual environment and
of new data, the model is continuously improved and expanded continuously monitor its performance and effectiveness.
to incorporate more influencing factors, enhancing the accuracy Regularly evaluate the accuracy and efficiency of the prediction
and comprehensiveness of the assessment. The AI model can system and optimize it based on the evaluation results.
assess the environmental impacts of the pyrolysis process, such 1.13 Development strategy of basic theories
as greenhouse gas emissions and pollutant releases. Through of integration and collaboration
predictive modeling, the environmental impacts under different By integrating various AI technologies into every aspect
operating conditions can be quantified, guiding optimization of the entire waste plastic pyrolysis process, efficient and
strategies and reducing negative environmental impacts. environmentally friendly resource recovery and processing
1.11 Development strategy for basic theory can be achieved. Firstly, an integrated data platform needs to
of resource recovery rate prediction be established, which is capable of collecting, storing, and
By collecting historical data on the physicochemical processing data from different sources. Develop AI-based
properties, pyrolysis conditions, product types, and recovery scheduling and optimization algorithms. Build a prediction
rates of waste plastics, as well as any factors that may affect and decision support system. Develop a visual user interface.
the recovery rate, statistical methods or machine learning Establish a continuous learning mechanism to enable the
algorithms are used to determine which features are most system to learn from each operation and continuously optimize
important for predicting the recovery rate. New features, its prediction and decision-making capabilities. In the face of
such as temperature-time interaction features, are created, constantly changing market and environmental conditions, the
or new patterns are discovered through methods such as system automatically adjusts its strategies and parameters.
cluster analysis. Appropriate models are selected based on

·4· Vol.52，No.2