Scenario
A compound developer must design or reformulate a plastic to hit thermal (glass transition, heat-deflection) and mechanical (modulus, strength, toughness) targets while controlling cost. The design space — base resin(s), blend ratios, additive package, and filler loading — is large, and compounding each candidate consumes material and time.
Worked example. A recycled-content polyolefin compound must recover the stiffness of virgin material by choosing the right filler and loading without making the part brittle.
Agent workflow
- Read the target property spec and the available resins, additives, and fillers.
- Blend miscibility screen — use solubility-parameter analysis to flag incompatible resin pairs and predict phase behavior of candidate blends.
- Filler reinforcement screen — model stiffness/strength gains versus filler type and loading, and flag the loading where toughness starts to fall.
- Additive package selection — choose stabilizers, impact modifiers, and processing aids that fit the base chemistry.
- Tg screen and mechanical property screen — predict glass transition, modulus, and strength for the ranked candidates.
Demo output
A ranked set of candidate compounds with predicted Tg, modulus, and strength, a miscibility verdict per blend, and the filler loading that best balances stiffness against toughness. The magic moment: the agent finds the filler / loading that restores virgin-like stiffness while staying below the embrittlement threshold.
Deterministic vs LLM
- Deterministic — property predictions, miscibility / solubility-parameter math, and reinforcement modeling.
- LLM — the compound-design narrative and the additive/filler rationale.
Every property number and miscibility verdict is a validated calculation, never an LLM guess. The agent reads and routes; the engine decides.
Data sources
- Native (offline): polymer property prediction; solubility-parameter and miscibility analysis; filler-reinforcement modeling; DOE screening.
- Public reference layer: polymer, additive, and filler property references.
Limits
Screening-level guidance, not a substitute for compounding trials, standardized mechanical testing, or long-term aging. It narrows the compound design space; the extruder and the test bars confirm the winner.
Skills orchestrated by this use case
This scenario routes an AI agent through the following curated skills.
-
Polymer Blend Miscibility Screen
Polymer ChemistryScreen a polymer blend for miscibility and phase behavior by comparing solubility parameters, interaction, and Tg to predict single- vs multi-phase morphology with Paramus property tools.When to use: a user needs an early read on whether two or more polymers will be miscible or phase-separate.Multi-tool WORKFLOW with phase-behavior judgment. -
Additive Package Selection
Polymer ChemistrySelect a plastics additive package by matching stabilizers, plasticizers, and processing aids to the polymer, process, and service environment with Paramus property and compatibility tools.When to use: a user needs to choose or check an additive package for a polymer and its service conditions.Multi-tool WORKFLOW with additive-package judgment. -
Filler Reinforcement Screen
Polymer ChemistryScreen a filled or reinforced polymer for stiffness, strength, and processability trade-offs by relating filler type, loading, and coupling to composite behavior with Paramus property tools.When to use: a user needs an early read on how a filler or reinforcement will change a polymer's properties.Multi-tool WORKFLOW with composite trade-off judgment. -
Polymer Tg Screen
Polymer ChemistryScreen and rank candidate polymers by predicted glass transition temperature (Tg) against a target value or window, orchestrating BigSMILES validation, polymer fingerprinting, and the Paramus BRAIN Tg model.When to use: a user provides repeat units (BigSMILES/SMILES) and asks which best hit a Tg target.Multi-tool WORKFLOW, not one call. -
Polymer Mechanical Property Screen
Polymer ChemistryScreen candidate polymers for mechanical performance (modulus, tensile strength, elongation, toughness proxies) by featurizing structures and predicting mechanical properties with Paramus BRAIN models, then ranking against a performance spec.When to use: a user asks “which polymer is stiffest/toughest?” or wants to meet a mechanical spec.Multi-tool WORKFLOW with a spec-matching decision rule.
Tools it reaches for
- Polymer property prediction (Tg, modulus, strength)
- Blend miscibility / solubility-parameter analysis
- Filler-reinforcement modeling
- DOE screening and ranking
Browse the full deterministic layer in the tool browser.