Design Optimisation/Simulation TRIZ Cases

Controlled oil charging in varied-porosity cores reveals dynamic fault-sand matching beyond geometric connectivity.

Machine learning reuses simulation data from similar circuit designs to reduce EDA computation time while refining power, performance, and area.

Traditional geomechanical simulation is computationally intensive and standard machine learning is data-hungry; GIMI combines physical models with neural networks for reliable forecasts.

Low-quality seismic data limits neural-network training; stochastic velocity models and forward simulations add realistic noise and scattering.

Traditional DDR inversion is slow and prone to local minima; a pretrained neural network predicts subsurface formation properties from measurements and tool parameters.

A physical model and machine learning precalculate abrasive-particle patterns for drill segments, reducing costly tests while targeting wear resistance.

Repair hanging vertices, thin faces, duplicate faces, and concave cells to make depositional-space grids usable for FEM with fewer nodal degrees-of-freedom.

Equal-distance particle placement and minimized velocity variance avoid repetitive calculations for simulation setup.

A QM/MM workflow uses Mining Minima poses and QESP charges to improve binding free energy accuracy without FEP-level computational cost.

Embedding geometry subroutines and form/action features creates reusable, customized multiphysics applications without rebuilding each modeling environment.

Actor-critic reinforcement learning sizes bounded-sliceline grid units for analog placement, reducing chip area and interconnecting wire-length.

Reinforcement-trained placement networks assign chip macro nodes sequentially to improve floorplans while reducing manual intervention and compute demand.

Simulate sensor configurations before MR hardware arrives, compare KPIs, and identify minimum requirements for application performance.

Compare simulation and measurement data, then select relevant metrics to optimize an objective validation measure for model quality.

Complex far-field models can overwhelm memory and runtime; electric-field-based grid thinning preserves accuracy where it matters.

A trained recurrent neural network replaces repeated microstructural simulation in macroscale FEM to predict stress and fatigue indicators faster.

Iterative material redistribution within periodic unit cells cuts computational load while improving thermal efficiency and manufacturability.

This digital twin combines lifecycle data, online monitoring, and multi-physics simulation to regulate additive manufacturing quality in real time.

Material and process variations can cause inconsistent products; thermal profiles, motion programs, and sensor feedback enable adaptive moulding control.

Numerical simulations and physics-based equations train a machine learning model to select aquifers for carbon storage with less computation.

Standardized cognimetric parameters automate psychological assessment and behavioral influence for faster real-time decisions.

Designing geological grids with accurate surface contacts helps add structural information late while reducing overlap artefacts in flow simulation.

Deterministic simulators miss probabilistic nucleation; this approach samples flow-sensitive hotspots and updates pore geometry for CO2 storage prediction.

Graph-based cover properties constrain root-to-leaf paths so one verification run can generate distinct hardware counterexamples for path coverage.

A hierarchical tree of node and data symbols enables efficient bytestream decoding with lossless reconstruction and flexible resolution support.

An APIE links digital content to physical locations and evaluates geolocation and other rules before granting retrieval access.

Historical geometry and simulation metadata train a model to predict core hours before execution, improving resource allocation.

Integrated simulation links subsurface models to facility planning and optimization.

The WAVE tool compares waveforms, models threats, and automates optimization before costly transmit-device implementation.

Selectable optical delays, frequency shifts, and amplitude control emulate coherent LiDAR scenes without costly outdoor setups.

A cloud EDA AI expert learns from user activity, refines design methodologies, and updates local agents for better assistance.

A trained ML model converts text intent into editable parametric 3D objects, reducing complex CAD interaction for less experienced users.

This case combines finite elements with quasi-bonds in crack-prone regions to improve large-scale solid damage simulation efficiency.

Iterative wave-function solving avoids basis functions in high-dimensional quantum models.

GPU random walks on micro-CT pore voxels update NMR relaxation rates, improving surface roughness resolution and pore-size analysis.

Opcode-programmed hardware simulates high-rate network physical layers deterministically.

Micro-CT data and digital rock physics link pore changes to cement mechanics under CO2 or H2 exposure.

A cloud simulator builds panel status from sensor events, serving multiple subscribers without overloading control-panel resources.

A hierarchical variational autoencoder combines global and local LiDAR features to model intensity and replicate sensor noise.

Text, image, video, and CAD inputs become PMTS representations and insights, reducing manual analysis of job inefficiencies.

Machine learning predicts inaccessible internal circuit signals for accurate debugging, reducing repeated modeling and verification effort.

Simulation-guided profiling reveals branch usage, enabling unequal hardware allocation that reduces bulk and improves circuit speed.

This case combines retrieval, parameter extraction, surrogate optimization, and simulation feedback to improve power converter design accuracy.

A native mesh workflow maps scanned anatomy into variable brace geometry, improving contour fit while reducing manual manufacturing effort.

Machine learning uses variance, fault probability, policies, and regions to adapt reporting while reducing unnecessary vehicle data.

Using metal and cell density features, the model prioritizes hard-to-fix DRC violations and reduces redundant ECO fix iterations.

Customer operating and use conditions feed a battery simulator that produces failure distributions and valuations for dynamic risk pricing.

Kernel Ridge models address drill bit metamorphism by comparing drilling GAV data with PVT samples to build reliable adherence curves.

This case generates threat models from infrastructure code, capturing application interactions without relying on manual expert review.

Historical simulation cooperation and line-planning data identify leading collaborators as simulation scale and task complexity grow.