Model More, Wait Less: GT-SUITE v2027

Simulation engineers troubleshoot with screenshots: a plot that looks wrong, a supplier datasheet, an error dialog. Retyping each one as text before asking for help slows the work.

GT-SUITE v2027 adds image support to AI.advisor, the assistant within GT Intelligence Studio, GT’s generative AI product family, trained specifically on GT-SUITE expertise. Engineers paste or drop any screenshot directly into the conversation, and AI.advisor reads the image alongside the question. This supports four common workflows:

• Explain results: Ask what a plot or data table means and get the answer in plain language.
• Model from a spec sheet: Paste a supplier datasheet and ask how to model the component in GT-SUITE.
• Resolve modeling issues: Share a screenshot of the model map, and AI.advisor pinpoints the setup problem.
• Decode error messages: Send the error shown on screen and get the cause and the fix.

Multi-step GT-SUITE modeling tasks, such as building a calibration model or auditing a lubrication circuit, force engineers to repeat the same setup by hand and risk inconsistent results from one user to the next.

AI.modeler, the multi-agent assistant in GT Intelligence Studio, now includes Skills: saved, step-by-step workflows that it activates on demand. A single instruction replaces a manual walkthrough, so every run follows the same expert-guided process. GT-SUITE ships with 24 Skills in three groups:

• Productivity (10 Skills): Automate setup work, such as building a full heat exchanger calibration model from specifications or assembling a refrigerant compressor model from a single object, so engineers spend less time on setup and more on simulation.
• Quality (9 Skills): Audit models before they run, checking plot and results storage, hidden attributes, EngineMap inputs, and lubrication circuits to catch missing data and silent errors early.
• Innovation (5 Skills): Explore more design concepts faster, including cam phasing, injection timing, GT-AutoLion peak power capability, and frequency-domain cranktrain damping for Noise, Vibration, and Harshness (NVH) analysis.

Organizations can also publish custom Skills to a shared network location, so every user in the organization runs the same validated workflow.

This capability has also been made available in GT v2026.2.01

Retraining a metamodel from scratch whenever the underlying data changes consumes time that engineering teams rarely have to spare.

In GT-SUITE v2027, the Machine Learning Assistant now supports metamodel fine-tuning, which adapts a metamodel trained on an original dataset to a new, different dataset. When repeating a project on a new dataset with the Heating, Ventilation, and Air Conditioning (HVAC) refrigerant circuit example model, fine-tuning trained 2–3× faster than training from scratch. Engineers can reproduce this result with the HVAC_Refrigerant_Circuit example under Examples > 0_AI_Machine_Learning. Fine-tuning benefits three scenarios:

• Augmented datasets: When new data extends an existing dataset, engineers fine-tune the existing metamodel to the larger dataset instead of retraining it.
• Repeated projects: When a completed machine learning project must be repeated on a new dataset, engineers start from the existing metamodel rather than from scratch.
• Foundational models: Engineers can train a metamodel on a large, comprehensive dataset, then fine-tune it to a smaller dataset that covers a subset of the factor space. v2027 does not ship any foundational models, so users build this workflow themselves.

Fine-tuning also enables a related capability: evaluating a metamodel on a dataset different from the one it was trained on. Previously, engineers had to retrain the metamodel on the new dataset. Now they can copy and paste the existing metamodel onto the new dataset and evaluate it immediately, provided the new dataset uses the same input factors and responses as the original.

Metamodels trained purely on data can predict behavior that violates physical laws, particularly when training data is sparse.

GT-SUITE v2027 introduces a new static regression metamodel type for physics-informed neural networks (PINNs). Engineers can apply custom governing equations and constraint equations that force the neural network to obey physical behavior, which strengthens existing static regression metamodels where data alone falls short. The PINN metamodel also supports multiple responses, so it can serve wherever a multi-response multilayer perceptron (MLP) is needed.

Engineers define the equations directly in an equation editor on the Create Metamodels page of the Machine Learning Assistant. The editor ties each set of equations to its metamodel.

Heat pump and refrigerant system development typically begins before engineers have selected components, yet system models have traditionally required test data and concrete component data before those models would be built.

GT-SUITE v2027 delivers a dedicated tool for preliminary refrigeration system analysis early in the design cycle. It requires only a handful of inputs, and engineers choose which to provide. For example, inputs could include basic compressor characteristics, a target cooling rate, and boundary conditions. The tool immediately returns the required refrigerant flow rate, capacity, and coefficient of performance (COP). Instead of building a full transient circuit model, engineers can set up a model and obtain results in less than a minute. It supports any refrigerant, so teams can evaluate natural refrigerant alternatives such as CO2 before committing to detailed component and system modeling.

The tool also easily integrates with the native Design of Experiments (DOE) and optimization features to find ideal designs and operating points.

High fidelity thermal component modeling depends on accurately capturing heat transfer between fluids and solids, a coupled problem known as conjugate heat transfer (CHT).

GT-SUITE v2027 adds CHT to GT-Auto-3DFlow, GT’s automated 3D flow solver designed for engineers without 3D CFD experience. The solver computes the flow in 3D and couples it to GT’s native 3D thermal finite element (FE) solver, which calculates temperatures in the solid structures. GT-Auto-3DFlow meshes the 3D flow domain and integrates the thermal domain automatically, so the entire process runs in a single workflow. The result is a solution that is fast, accurate, and easy to use.

This capability extends the existing strengths of GT-Auto-3DFlow, which link 3D airflow to GT’s 2D heat exchangers and provide general-purpose 3D flow for characterizing pressure drop and heat transfer in complex geometries. Typical use cases include cooling of battery packs, e-motors, engines, power electronics, data centers, and more.

Battery performance and aging depend heavily on mechanical effects inside the cell, swelling, internal stress, and gas generation from parasitic reactions, but these effects are difficult to model together with the underlying electrochemistry, and existing tools often treat them in isolation.

GT-AutoLion v2027 extends its electrochemical-mechanical coupling in two directions. First, the model now captures internal gas generation from side reactions that occur during cycling, including the resulting pressure buildup and its effect on cell volume and ion transport kinetics. Second, particle-level mechanics, the radial and tangential stresses that build up in active material particles as lithium ions intercalate and de-intercalate, are now mechanically constrained by neighboring particles and by boundary conditions at the electrode and cell level, rather than treated independently.

Together, these developments let engineers model mechanical behavior consistently across the particle, electrode, and cell scales, giving a more complete and physically consistent picture of how mechanical effects drive performance and degradation over a battery’s life.

In v2027, we have added 2D finite-element-based NVH capabilities to GT-FEMAG, making it a powerful, one-stop multiphysics tool for electric motor design and analysis.

Parametric motor geometries and parametric housing, including lobes, windings and assembly constraints, can be created to accurately capture real-world assembly conditions.

Free vibration analysis predicts eigenfrequencies and mode shapes, while forced frequency response analysis across the entire motor speed range enables the generation of order plots and Campbell diagrams.

A new shrink-fit analysis capability calculates stator and housing stresses and strains, improving simulation accuracy by accounting for assembly-induced effects.

On the 3D side, the NVH workflow has been significantly streamlined through automatic 2D-to-3D mesh extrusion and electromagnetic force mapping onto the meshed stator geometry, eliminating manual intervention. An external system housing can also be coupled to the 3D stator model in GT-SUITE to predict Equivalent Radiated Power (ERP) and acoustic Sound Pressure Levels (SPL).

Building a complete engine model from a blank sheet typically takes hours of setup work — and getting a team productive with the modeling practices behind it can take days of training.

GT-SUITE v2027 introduces Engine Model Builder, a guided workflow that builds a complete engine model in seconds from a small set of user-provided inputs. This removes the need to construct components individually or master advanced modeling techniques up front.

The capability is designed for organizations requiring a rapid engine-modeling solution without significant GT-SUITE in-house experience, including turbocharger suppliers, marine propulsion and shipbuilding companies, and controls engineers looking to begin controller development ahead of detailed engine model availability.

In v2027, Engine Model Builder supports:

• Inline engine architectures (V-engine support is planned for v2028)
• Turbocharged configurations only
• Gasoline and diesel fuels only

Models generated with Engine Model Builder can be further customized using AI.modeler, part of the GT Intelligence Studio generative AI product family, to match individual project requirements, and development will continue in v2028 and beyond as additional architectures and fuel types are added.

Predicting Proton Exchange Membrane (PEM) fuel cell behavior across different layers and operating conditions requires accurate resolution of the electrochemistry, thermal behavior, and water management inside the membrane electrode assembly (MEA). GT-SUITE v2027 introduces SpecsPEMFC-Predictive-Detailed, a new predictive 1D electrochemical template that fully resolves the MEA, including both catalyst layers and the membrane.

The model solves electron and proton transport across every layer, driven by the electrochemical reactions, together with species transport and thermal behavior. It resolves dissolved water in the membrane and the catalyst layers and includes a two-phase flow model for water. Engineers define the discretization of each layer to increase the precision of the results. The model runs transiently and couples with the GT-SUITE 1D flow and thermal solver.

This coupling lets engineers use the same template in stack models, Balance of Plant (BoP) models, and system-level topology and vehicle models. The template suits detailed studies of specific applications, such as evaluating material or operating-condition variations at the model level.

Distillation processes may call for condenser and reboiler configurations that fall outside the standard design, forcing engineers to either accept a built-in model or omit the effect from their simulation entirely.

GT-SUITE v2027 lets the DistillationColumn template represent its condenser and reboiler parts by using external flow components instead of the internal models, opening the door to custom designs that the built-in options can’t capture.  Engineers activate this by setting the Condenser Type and/or Reboiler Type flag to “external,” and the option can be applied to the condenser only, the reboiler only, or both simultaneously giving full flexibility to customize the design and model non-standard configurations.

Metals described as “corrosion-resistant” rely on a thin protective oxide layer, a passive film, that forms on the surface and governs how the material actually behaves in service. When that film grows, depletes, or breaks down unevenly, corrosion resistance can fail in ways that are difficult to predict from bulk material properties alone.

GT-xCHEM, GT’s chemical kinetics and aftertreatment modeling tool, now includes a passivation layer model in the ReactiveDuct template to support passivation-controlled corrosion through a dedicated passive-film reaction framework. The framework tracks dynamic passive-film growth and depletion over time, capturing how the protective layer thickens, thins, or locally breaks down under operating conditions. A reaction-diffusion solver accounts for film diffusion resistance as solvent and ionic species move through the passivation layer, while ohmic resistance from both solid-phase and ionic conductivity is included to reflect the layer’s electrical behavior. Together, these additions let engineers model passivation-controlled corrosion with a level of physical detail that simple bulk-corrosion assumptions cannot capture.

The framework models:

• Dynamic passive-film growth and depletion
• Solvent and ion transport through the passive film
• Film diffusion resistance within a reaction-diffusion solver
• Ohmic resistance from solid and ionic conductivity

To prevent resonance, turbines, compressors, and pumps must avoid speeds at which excitation orders coincide with rotor natural frequencies. Engineers therefore need to map these frequencies across the full operating-speed range early in the design process.

GT-SUITE v2027 adds a rotordynamics option to the 3D frequency-domain analysis (MechFreqAnalysis3D), solving free and forced vibration for assemblies of 3D beams and rigid bodies supported to ground through springs and dampers. The solver includes gyroscopic effects, which split forward and backward whirl frequencies, and offers optional centrifugal stiffening for long, slender components. Engineers choose a fixed reference frame to model asymmetric stator properties, or a rotating frame to model asymmetric rotors.

Campbell diagrams track each mode automatically across the speed range, so engineers can identify critical speeds directly and animate mode shapes at any chosen speed. Engineers apply unbalanced or oscillating forces as boundary conditions on the rotating model to evaluate forced response in the same analysis.

Validation cases cover rigid rotors on flexible supports with isotropic and anisotropic bearings, a flexible rotor compared against an independent finite element model, and a rotating asymmetric cantilever beam, where bending mode frequencies agree with analytical values within 2%.

Electric Vehicle (EV) range limitations mean a simulated route from point A to point B is only useful if it reflects how the vehicle will actually be driven; including where and how long it stops to recharge along the way.

GT-RealDrive now includes an EV Charging Stations capability which searches for and inserts real charging stops directly into a calculated route, combining route planning and charging infrastructure in a single interface. Engineers set a search start location along the route, then define a search corridor distance that controls how far off the main path GT-RealDrive will look for stations. Results can be filtered by connector type, including CHAdeMO, CCS1, CCS2, Type 1 (J1772), Type 2, and Tesla. Each matched station returns its location, supplier name, stop duration, and the number and maximum power rating of connectors available per type. Because GT-RealDrive draws on GT-SUITE’s physics-based vehicle model, energy consumption between stops reflects real driving conditions rather than a simplified range estimate.

The result is a route that accounts for real-world charging infrastructure and connector compatibility, letting engineers validate range, charging strategy, and trip-time assumptions for a specific EV program without manually researching or inserting charging stops.

Modern engineering models routinely span 10–100 files, 1,000-plus parameters, and 20–100 cases built by multiple teams, making it hard to know which file version is current, where data is being reused, and whether changes in one place have rippled through the rest of the project.

GT-SUITE v2027 introduces GT-Project, a new “Project” file (.gtp) that gathers every file tied to an engineering development program, model files, data files, reports, presentations, Process Maps, and even non-GT files, into a single structure. The new Project Explorer panel gives engineers full visibility into where each file is stored and used, with the ability to swap one file for another directly from the panel.

Global and Project Libraries

Library-linked values were previously all-or-nothing, locked in full once tied to a library. GT-Project lets engineers unlock individual attributes or elements within a library-linked object, modify and test the change locally, and then either revert it or push it back to the library once validated.

Existing user object libraries are renamed Global Libraries, reserved for elements meant to be reused across every project in a company. A new Project Libraries tier adds an intermediate scope: data reused only within a single project. Reusable components and subsystems can point to project-specific data, company-wide Global Library data, or both.

Hierarchical, modular models support reuse and parallel work across teams, but that structure has historically made parameter values harder to see and navigate.

Engineers running long simulation and post-processing sessions frequently work in low-light environments, where bright interfaces contribute to eye strain and can pose accessibility challenges over extended use.

GT-SUITE v2027 introduces a Dark Mode theme across all GT desktop applications — GT-ISE, GT-POST, and GEM3D. Users choose between Light, Dark, and System Default, with System Default automatically matching the operating system’s active theme. The selected mode takes effect after an application restart and applies consistently across the interface, reducing glare during long working sessions and giving users more control over how the workspace looks and feels.