Plastic Recycling Feedforward Additive Control Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing plastic recycling processes face challenges in achieving target characteristics for recycled plastic due to variability in input stream composition, leading to inadequate adjustments, increased resource consumption, and waste, as they rely on feedback-based systems that are reactive rather than proactive.
Innovation Solution
A machine-learning based feedforward mechanism that monitors input stream characteristics to proactively determine administration schemes for additives, using sensors and a trained model to adjust processing parameters and additives in real-time, thereby ensuring target characteristics are met in the output.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If feedback-based systems are used to adjust additives, then adjustments can be made based on output stream monitoring, but the adjustments are reactive and inadequate when input stream composition changes by the time feedback adjustments are made
Solution Approach 1:
The system performs preliminary action by monitoring input stream characteristics before the recycling process and proactively determining administration schemes for additives. This feedforward approach predicts the necessary adjustments before the input stream enters the recycling process, eliminating the time delay inherent in feedback-based systems and ensuring target characteristics are achieved even when input composition varies.
2Manufacturing precision
If multiple compounding cycles are performed to achieve target characteristics, then output quality can be improved, but resource consumption and carbon emissions increase
Solution Approach 1:
By proactively determining the optimal administration scheme for additives based on input stream characteristics before processing, the system achieves target output characteristics in fewer compounding cycles. This eliminates the need for multiple iterative cycles, thereby reducing energy consumption and carbon emissions associated with repeated processing.
Solution Approach 2:
The system uses monitoring of input stream characteristics to inform additive administration decisions in real-time. This closed-loop feedforward control ensures that the correct amount and type of additives are applied from the beginning, achieving target quality without requiring multiple compounding cycles and their associated energy consumption.
3Manufacturing precision
If additives are added to achieve target characteristics, then output quality can be improved, but extraneous consumption of additives occurs
Solution Approach 1:
The system applies local quality by determining specific administration schemes tailored to the actual input stream characteristics. Rather than using a uniform additive approach, the system monitors the specific composition and properties of each input stream and adjusts additive types, amounts, and timing accordingly. This precision minimizes extraneous additive consumption while ensuring target characteristics are achieved.
Solution Approach 2:
The system dynamically changes parameters of additive administration (amount, timing, type) based on monitored input stream characteristics. By adjusting these parameters in real-time according to actual input composition, the system achieves target output characteristics with minimal extraneous additive consumption, avoiding both under-dosing and over-dosing scenarios.
Data Source
AI summary
A machine-learning based approach to preemptively adjust additives applied to a stream of plastic to obtain recycled plastic with desired target characteristics. The approach includes obtaining sensor input of an input stream of material into a plastic recycling system and determining, using a machine-learning model, a target set of characteristics of an output material. The machine-learning model is trained to determine an administration scheme associated with a set of additives to be added to the input stream to obtain the target set of characteristics that include at least one of a (i) physical characteristic or (ii) a mechanical characteristic, of the output material. The approach includes generating a control signal configured to affect a control system for adjusting the set of additives within the plastic recycling system in accordance with the administration scheme.