Patent landscape: who is filing and where
The multi-physics simulation patent dataset reviewed covers more than 40 active and pending filings spanning 2000–2026, with China representing approximately 80% of records — the dominant jurisdiction by a wide margin. Japan, South Korea, and the EU account for most of the remainder, with a smaller set of US-jurisdiction filings. The most frequently cited assignees include China FAW Group, Huazhong University of Science and Technology, Tsinghua University, Chengdu Fujin Power Semiconductor Technology, Xi’an Jiaotong University, and Zhuzhou CRRC Times Electric. International filers include Mitsubishi Electric, Lam Research, and Qualcomm.
The dominant technical approaches fall into four categories: (1) tightly coupled electrical–thermal (electro-thermal) co-simulation, where switching losses from circuit solvers dynamically update thermal boundary conditions; (2) multi-physics finite element analysis (FEA) integrating electrical, thermal, and structural stress fields through tools such as COMSOL and ANSYS; (3) modular finite element model libraries enabling rapid reassembly without full geometry rebuild; and (4) iterative optimization loops that use simulation results to converge chip placement, cooling geometry, and packaging parameters before any hardware is fabricated. The common thread across all categories is substitution of virtual design-evaluate-modify cycles for physical prototype-test-rework cycles.
A review of more than 40 multi-physics simulation patents filed between 2000 and 2026 shows China as the dominant jurisdiction at approximately 80% of records, with key assignees including China FAW Group, Huazhong University of Science and Technology, Tsinghua University, and Zhuzhou CRRC Times Electric.
Electro-thermal co-simulation: closing the feedback loop between circuit and heat
The fundamental reason power electronics thermal design historically required many physical prototypes is that electrical switching losses are temperature-dependent and the thermal response is non-linear. Simulating each domain independently produces systematically wrong answers: circuit simulations that ignore junction temperature rise underestimate losses, while thermal simulations fed with fixed loss values miss the feedback effect. The solution documented across multiple patents is bidirectional, real-time coupling between circuit and thermal solvers.
Standalone thermal network models based on RC analogies are only accurate for linear time-invariant systems. Real power semiconductor thermal systems are non-linear — switching losses change with junction temperature, which in turn feeds back into loss generation. Naive RC-based approaches therefore overestimate thermal performance, which is why bidirectional electro-thermal co-simulation is necessary for accurate design validation.
A Multi-Physics Co-Simulation Method of Power Semiconductor Modules from Shenzhen Union Semiconductor (2023, US jurisdiction) directly addresses this by coupling the circuit simulation software PSpice with the finite element solver COMSOL through a custom indirect coupling interface. The method enables electricity–heat–force co-simulation in a single coordinated workflow with adaptive time-stepping that dramatically shortens simulation time. This patent is the US counterpart of the Huazhong University of Science and Technology Chinese filing, and it explicitly identifies the failure mode of standalone RC thermal models for non-linear systems — a failure mode that forces engineers to verify designs through physical hardware.
The iterative co-simulation concept is extended by a Power Electronics System Electro-Thermal Joint Simulation patent from Chengdu Fujin Power Semiconductor Technology (2022), which implements dynamic suspension of the electrical simulation during thermal steady-state intervals. Because the electrical model requires microsecond time steps while the thermal model requires milliseconds to seconds to reach steady state, naively running both at the finest step is computationally prohibitive. By pausing electrical simulation when the circuit is in steady state and resuming it only when that steady state is disrupted, the method reduces total computation load while preserving accuracy. The patent explicitly states that early-stage electro-thermal co-simulation “can greatly improve power electronics R&D efficiency and effectively avoid product development delays caused by mismatches between electrical parameter design and power device selection.”
“Early-stage electro-thermal co-simulation can greatly improve power electronics R&D efficiency and effectively avoid product development delays caused by mismatches between electrical parameter design and power device selection.”
Zhuzhou CRRC Times Electric’s patent on Thermal Simulation Method for Semiconductor Power Components (2018) addresses the structural inefficiency in conventional FEA workflows: historically, any modification to a power device or heat sink required rebuilding the entire assembly’s geometric and finite element model from scratch. The invention introduces a modular FEM library from which sub-component models are assembled, so that when a designer needs to evaluate a different heatsink geometry, only the relevant sub-model is swapped. This architectural change directly reduces the cost of exploring design alternatives in simulation — each virtual experiment replaces what would otherwise be a new physical prototype. According to WIPO, modular and reusable simulation architectures are an emerging priority in industrial IP strategy for advanced manufacturing.
Chengdu Fujin Power Semiconductor Technology’s 2022 patent resolves the electro-thermal time-scale mismatch by suspending electrical simulation (which requires microsecond time steps) during thermal steady-state intervals (which require milliseconds to seconds), making converged coupled simulation computationally feasible at the early design stage.
Explore the full patent landscape for electro-thermal co-simulation in PatSnap Eureka.
Search Patents in PatSnap Eureka →Integrated multi-physics field analysis: beyond electro-thermal coupling
True multi-physics simulation extends beyond electro-thermal coupling to incorporate mechanical stress, fluid dynamics, and electromagnetic fields — each of which influences the others in real power electronic systems. Solder joint fatigue, die-attach cracking, and package delamination under thermal cycling are mechanical consequences of thermal loading that determine long-term reliability, and which can be predicted virtually only if the coupled fields are solved together.
A Power Semiconductor Module Multi-Physics Joint Simulation Method from Huazhong University of Science and Technology (2021) explicitly treats the electricity–heat–force strong coupling problem and implements indirect coupling between PSpice and COMSOL to achieve real-time bidirectional exchange. The patent notes that prior methods — including using fixed loss values as boundary conditions, or fitting loss-versus-time mathematical expressions — all fail to capture the dynamic feedback between electrical behavior and thermal state, producing inaccurate stress field results that cannot be used to predict mechanical reliability.
A Power Electronics Product Fatigue Life Prediction Method from Ningbo Pruely Junsheng Automotive Electronics (2024) extends multi-physics coupling to vibration-thermal interaction — a domain particularly relevant in automotive power electronics. The method builds both a thermal simulation model and a vibration simulation model, then iterates between them to obtain a stable composite temperature field (including heat from both power dissipation and viscoelastic damping in encapsulant materials) and a stress field, from which vibration fatigue life is calculated. Without this kind of coupled virtual analysis, the interaction between vibration-induced heat generation and thermomechanical stress could only be characterized through destructive physical testing.
China FAW Group’s Power Module Simulation Analysis Method (2023) implements what may be the most comprehensive multi-physics workflow in the dataset: a simulation suite spanning electrical, thermal, mechanical (structural force), and fluid flow (flow field) scenarios, all executed on the same digital model of the power module. The method includes structural simplification to remove features that do not affect performance (fillets, ribs, holes), material property assignment including S-N fatigue curves, and multi-domain boundary condition handoffs. Standards bodies such as IEC recognize multi-domain simulation as a prerequisite for reliability qualification in power electronics.
Tsinghua University’s Power Semiconductor Device and Circuit Simulation Platform (2024) provides a unified platform with chip simulation, circuit simulation, and packaging simulation units connected through a coupled simulation unit. The platform explicitly identifies the “electro-magneto-thermal-mechanical (E-M-T-M) multi-physics coupling effect” as the most comprehensive descriptor of real power semiconductor service environments, and argues that breaking down technical barriers between simulation layers is necessary to capture these effects accurately.
Tsinghua University’s 2024 unified simulation platform explicitly names the electro-magneto-thermal-mechanical (E-M-T-M) multi-physics coupling effect as the most comprehensive descriptor of real power semiconductor service environments. Breaking down technical barriers between chip-level, circuit-level, and packaging-level simulation layers is identified as necessary to capture these effects accurately.
Ningbo Pruely Junsheng Automotive Electronics’ 2024 patent builds both a thermal simulation model and a vibration simulation model, iterates between them to converge on a composite temperature field (including heat from viscoelastic damping in encapsulant materials) and a stress field, and calculates vibration fatigue life — replacing destructive physical power cycling tests.
Iterative optimization architectures: replacing hardware loops with simulation loops
The mechanism by which multi-physics simulation reduces physical prototype iterations is not simply “simulate once and build once.” Rather, it involves replacing the physical build-test-modify loop with a simulation-evaluate-optimize loop that can be executed far more quickly and at far lower cost. Several patents in the dataset explicitly describe convergence-based iterative architectures that operate entirely within the simulation environment.
China FAW Group’s Power Module Design Optimization Method (pending, 2025) describes a closed-loop optimization that iterates between thermal simulation and current distribution simulation. In each iteration: (1) thermal simulation determines junction temperature for each chip device; (2) parasitic parameter extraction feeds into dynamic/static current-sharing simulation; (3) a weighted composite metric of maximum junction temperature difference and maximum current difference is computed; and (4) chip positions are adjusted if the metric exceeds a threshold. The loop continues until convergence. The result is an optimized chip layout that accounts for coupled electro-thermal non-uniformities — findings that would require multiple prototype builds to discover empirically.
Nexperia (Shanghai)’s Heatsink Design Method for Vehicle Power Modules (2023) employs a response surface methodology over FEA simulation data to replace exhaustive parametric FEA sweeps. By sampling combinations of heatsink geometry parameters (fin pitch, fin height, corner radius), fitting an explicit surrogate function for both junction temperature rise and pressure drop, and then applying multi-objective optimization to minimize both simultaneously, the method characterizes the full design space with far fewer FEA runs than a brute-force grid search would require. This is mathematically equivalent to compressing hundreds of virtual experiments — and correspondingly many would-be physical prototypes — into a manageable set. Research published by IEEE confirms that surrogate-based optimization is a recognized best practice for multi-objective thermal design in power electronics.
Guangdong Lanrui Technology Group’s Liquid-Cooled Heatsink Microchannel Construction Method with Integrated Bypass Control (2025) introduces a digital twin model that couples fluid domain, solid domain, and shape-memory alloy (SMA) actuator domain through bidirectional conjugate heat transfer, thermomechanical deformation, and fluid-structure interaction. The model is subjected to time-varying heat loads to generate dynamic performance data, and design parameters including actuator geometry are iteratively optimized before any physical microchannel fabrication. The digital twin approach is the formal realization of “replace physical prototypes with virtual experiments.”
China FAW Group’s Intelligent Power Unit Electro-Thermal Performance Analysis Method (2025) employs Saber software to build driver unit electrical models incorporating drive resistance, stray inductance, and protection logic, while simultaneously building a multi-level thermal network model by reverse-engineering the internal layer structure of the power module. The thermal network model is made editable at each hierarchy level, so designers can modify cooling path configurations and instantly observe thermal consequences — enabling virtual exploration of design variants that would otherwise require fabricated prototypes.
Analyse convergence-based optimization and digital twin patents with PatSnap Eureka’s AI-powered search.
Explore Full Patent Data in PatSnap Eureka →Key players and emerging trends in the IP landscape
Several assignees appear repeatedly in the dataset and represent concentrated intellectual property in multi-physics simulation for power electronics thermal design. China FAW Group is the most prolific filer in the automotive power module domain, with at least five patents covering power module design optimization, electro-thermal performance analysis of intelligent power units, DC bus capacitor electro-thermal coupling, power module multi-physics simulation workflows, and random vibration analysis. Their work consistently integrates thermal, electrical, and mechanical physics with automotive-grade robustness requirements.
Chengdu Fujin Power Semiconductor Technology holds two closely related patents on electro-thermal joint simulation that share the key innovation of dynamically suspending electrical simulation during thermal steady-state intervals, directly addressing the time-scale mismatch problem. Huazhong University of Science and Technology and its US-jurisdiction counterpart Shenzhen Union Semiconductor own the foundational PSpice-COMSOL indirect coupling methodology for electricity–heat–force co-simulation, which is cited or mirrored by several downstream patents. Tsinghua University contributes a unified cross-hierarchy simulation platform connecting chip-level, circuit-level, and packaging-level simulation — an architectural contribution targeting the organizational fragmentation of simulation tools that currently prevents full E-M-T-M coupling.
Xi’an Jiaotong University contributes to the power cycling domain with methods that use electro-thermal joint simulation to calculate junction temperature and case temperature curves under realistic load conditions — output that enables accelerated aging test design without physical module destruction. Mitsubishi Electric Corporation holds an EP-jurisdiction patent on Process for Monitoring Thermal Resistances in a Power Electronic System (2024) that demonstrates international focus on thermal impedance measurement methodologies as a complement to simulation, using thermally sensitive electrical parameters to infer layer temperatures non-destructively. Lam Research files on Electromechanical Modeling of Components for Power Box Design (2025), generating electrical twins from mechanical model simulations across load current and ambient temperature ranges — an approach applicable to semiconductor processing equipment power systems where static modeling fails under dynamic operating conditions.
A notable trend across 2021–2026 filings is the integration of machine learning and distributed computing. Guanku Technology (Shanghai)’s Performance Simulation Method for Power Modules (2025) adds a distributed computing platform for large-scale parallel simulation tasks with dynamic resource allocation, enabling design space exploration at a scale that was previously impractical. Northwestern Polytechnical University Shenzhen Institute’s Machine Learning-Based Power Semiconductor Transient Simulation Accuracy Optimization (2025) uses a lightweight convolutional neural network to map low-resolution waveforms to high-accuracy waveforms, resolving the accuracy-versus-computational-cost tradeoff that limits real-time simulation fidelity. As noted by OECD in its analysis of AI adoption in industrial R&D, machine learning integration into simulation workflows is accelerating across advanced manufacturing sectors.
China FAW Group is the most prolific filer in the automotive power module multi-physics simulation domain, with at least five patents in the dataset covering power module design optimization, electro-thermal performance analysis of intelligent power units, DC bus capacitor electro-thermal coupling, multi-physics simulation workflows, and random vibration analysis.
“A lightweight convolutional neural network mapping low-resolution waveforms to high-accuracy waveforms resolves the accuracy-versus-computational-cost tradeoff that limits real-time simulation fidelity — the current frontier of multi-physics simulation for power electronics.”