Page 49 - 《橡塑技术与装备》英文版2026年2期

SPECIAL AND COMPREHENSIVE REVIEW

1.14 Development strategy for basic theories regression models in the pyrolysis of waste plastics is primarily
of real-time monitoring and quality control manifested in two aspects: prediction and classification,
Through real-time processing and analysis of collected specifically targeting numerical prediction and categorical
data, the real-time monitoring system is integrated with the prediction tasks, respectively.
quality control process to achieve a closed-loop control system. 2.1.1 Analysis of main functions and
By comparing the monitoring data with the set target values i mpl emen tatio n strateg ies fo r key
through a feedback loop, process parameters are automatically technologies of linear regression model
adjusted to achieve optimal operating conditions. Through real- Main function: The regression model is primarily used
time monitoring and AI analysis, product quality during the for predicting continuous variables, such as the yield, energy
pyrolysis process can be monitored, and process parameters consumption, and temperature changes during the pyrolysis
can be adjusted in a timely manner to ensure the stability and process of waste plastics.
high quality of output. Key technology implementation strategy:
(1) Data collection: Collect various parameters during
2 Main functional analysis and key the pyrolysis process of waste plastics, such as temperature,
technology implementation strategy of pressure, time, type of raw material, yield of pyrolysis
AI waste plastic pyrolysis model products, etc.
The pyrolysis technology of waste plastics, empowered (2) Feature selection: Select features that have a
by AI, is advancing towards the goal of higher efficiency, significant impact on the prediction target as input variables.
cleaner operation, and maximized resource recovery and (3) Model Training: Train a linear regression model using
utilization. Firstly, deep learning algorithms are applied to historical data to find the optimal parameters (weights and
optimize the control of the pyrolysis process. By monitoring intercept) that minimize the error between the predicted and
and predicting pyrolysis conditions in real-time, precise actual results.
regulation of reaction parameters is achieved, thereby (4) Model evaluation: Evaluate the predictive
enhancing pyrolysis efficiency and product quality. Secondly, performance of the model through methods such as cross-
reinforcement learning technology demonstrates great potential validation to ensure that the model exhibits good predictive
in simulating the complex dynamic behaviors of the pyrolysis ability even on unseen data.
process. By constructing a virtual experimental platform, (5) Application and optimization: Apply the model to
optimization strategies are continuously adjusted to achieve real-time data to predict yields, energy consumption, etc. under
the best pyrolysis results. Furthermore, integrating artificial different conditions. Adjust the pyrolysis process parameters
intelligence with the Internet of Things (IoT) technology based on the prediction results to optimize pyrolysis efficiency
enables remote monitoring and intelligent maintenance of and resource recovery rate.
pyrolysis equipment, reducing operational costs and enhancing 2.1.2 Main functional analysis and key
equipment reliability. technical implementation strategies of
AI waste plastic pyrolysis technology is developed logistic regression model
around improving efficiency, optimizing processes, enhancing Main function: The logistic regression model is primarily
decision-making capabilities, and prediction. used for classification tasks, such as predicting whether waste
2.1 Main analysis and key technical plastics are suitable for pyrolysis, and the types of products
implementation strategies of linear produced after pyrolysis of different types of waste plastics.
regression model and logistic regression Key technology implementation strategy: several key
model technical steps for application:
The application of linear regression and logistic (1) Data collection: Collect the physicochemical
properties of waste plastics, the types of products before and

Vol.52，2026 ·5·