Skip to content
Matmerize

AI Optimization Across the Polymer Value Chain: Synthesis,
Compounding, Processing, and Manufacturing

Optimization is a persistent challenge across the polymer and formulation value chain, from material synthesis and
compounding through processing and product manufacturing. At every stage, teams must select and tune a large
parameter set (monomers, seeds, initiators, chain transfer agents, solvents, surfactants, catalysts, chain
terminators, temperature profiles, and more) against a competing set of targets. Those targets span process safety
and polymerization conversion rate, the physical characteristics of the resulting polymer, and the performance of
the downstream product.
PolymRizeTM, Matmerize’s polymer informatics platform, provides an AI-driven pipeline that lets manufacturers
run this optimization predictively and at scale.

Customer Challenge

A leading global specialty chemicals company relied heavily on physical experimentation to evaluate ingredient
recipes, processing conditions, and compounding routes for early hazard detection, product line optimization, and
portfolio expansion. Three major challenges made the approach increasingly unsustainable:

1. Market pressure

Downstream manufacturers expect higher performance, lower cost, improved sustainability, and
compatibility with existing infrastructure. Meeting those expectations requires shorter development cycles
and faster adaptation of the product portfolio to market demand.

2. Regulatory pressure

Governments worldwide are introducing stringent regulations and tightening requirements on
sustainability, emissions, hazardous substances, and recyclability. The company must continuously
reformulate products and manufacturing processes to maintain compliance without sacrificing
performance or cost competitiveness.

3. R&D throughput

Trial-and-error experimentation across formulation, compounding, processing, & manufacturing
conditions cannot keep pace with compressed product cycles. It slows innovation, drives up cost, and erodes
competitiveness.
The company needed a way to evaluate polymerization recipes, processing conditions, compounding routes and
final product properties against both safety and performance requirements fast enough to shorten development
cycles and improve throughput to reduce cost and strengthen its competitive position.

Solution

Matmerize used its PolymRizeTM platform to develop, deploy, and demonstrate an AI-integrated workflow for
predicting and designing outcomes across all four stages of the value chain. Data needed to flow and sequential
models were built to capture this flow of information as summarized in Figure 1.
Goal Setting: Starting from the company operations to be modeled, the key inputs (e.g. ingredients, recipes,
additive, processing conditions) and the required outputs (e.g. safety indicators, product quality, performance)
were identified for each step - polymer synthesis, compounding, processing and product manufacturing process.
Data Curation & Processing: Historical data covering the identified inputs and outputs, and the data flow
was collected in various formats (structured tables, pdfs, lab notes, printed reports). A data strategy was
developed, and the highly inhomogeneous data was extracted, cleaned, integrated, and processed specifically for
each model to maintain data flow between models.
Figure 1. AI-powered workflow included 4 models for polymer synthesis, compounding, processing,
and final product properties.
Model Development, Optimization, & Validation: Four sequential predictive models, named Model 1,
Model 2, Model 3 and Model 4, were developed to predict the outputs of the four stages: the reactor parameters
(Model 1), the fluid properties (Model 2), optimal processing conditions (Model 3) and the product test results
(Model 4). Multiple AI algorithms available in PolymRizeTM were evaluated and used to train, optimize, and
validate the predictive models for safety indicators, process evolution, and product quality and performance.
Active learning cycles were implemented as new data became available to refine the models and improve data
diversity for better model accuracy.
Interface Development & Deployment: A user-friendly interface integrating the AI models and tailored to
the company’s data and end users (Figure 2) was developed and deployed, allowing technicians to predict safety
indicators and performance across polymer synthesis, compounding, processing, and manufacturing.
Figure 2. (Left) A customized user interface developed to access AI models through an integrated R&D workflow and
(right) model performance validated by subsequent experiments.

Key Outcomes

Cost Savings: The company currently runs 1,000 batches for every 5 recipes that reach the customer, a 0.5%
conversion rate. Of these, 10% are aborted at the first stage due to safety issues, 50% advance to compounding
after polymerization characterization, and only 4 or 5 are ultimately selected after compounding testing. The
AI-integrated approach is expected to double this conversion rate, cutting the number of batches needed in
half. This lowers costs, improves resource utilization, and shortens process optimization by reducing reliance
on costly, time-consuming trial-and-error experimentation.
Accelerated Innovation: Workflows across all four stages have been modeled and can now be evaluated
virtually within minutes, compared to the 1–2 weeks required per batch for pilot-scale experiments. Process
evolution, product quality, and potential hazards, including pressure excursions, temperature runaway, and
gas emissions, are predicted up front, enabling faster and safer process development.
Improved Product Quality & Sustainability: AI enables rapid optimization of polymerization and
processing conditions to simultaneously meet product quality, manufacturing efficiency, and sustainability
targets, including reduced energy consumption, lower waste generation, and improved material utilization.
Regulatory Compliance: The AI-integrated workflow supports the development of formulations and
manufacturing processes that satisfy regulatory, safety, and environmental requirements before manufacturing
or experimentation even begins.
Developed, validated, and deployed at the company's pilot facility, the custom AI models now let the team optimize
and apply predictive capabilities across the polymer value chain rather than discovering outcomes experimentally.
Building on these pilot-scale results, the next phase is to deploy the models across the company's production
facilities.