June 2026 critical fixes

Bug 121340: Convection coefficient and temperature results differ for models with thermo-mechanical contacts when the finite element method is used and the solution is run in parallel

Fixed version: 2606.2

This fix applies to thermal-mechanical models that use the finite element method and include thermo-mechanical contact, whether run in serial or parallel.

Previously, when calculating the gap distance used to determine whether a contact was open or closed, the solver used the first available connection between an element and its coupling elements. In models with multiple connections, this could lead to inconsistent behavior because the first available connection could differ between CPUs during parallel execution. It could also produce inconsistent results when different connections were simultaneously in different open or closed states.

With this fix, the element gap distance is now calculated as an average across the applicable conductances and connections, providing more consistent behavior in both serial and parallel runs.

An additional correction was made for cases where an initial connection becomes undefined due to structural motion. These undefined connections are now excluded from the averaging process. Previously, the entire element and its associated nodes could be treated as having a zero convection coefficient.

The impact of this change is most noticeable in convection coefficient and temperature results for steady-state simulations in which contact conditions evolve during nonlinear iterations.

Bug 122547: Radiation is automatically deactivated for the wrong footprint and junction elements in PCB component boundary conditions because internal IDs are used instead of user-defined IDs

This fix addresses an issue in the solver pre-processor module where the automatic deactivation of radiation for footprint and junction elements in PCB component boundary conditions could reference incorrect elements in the INPF file.

The issue occurred when user-defined element IDs did not begin at 1. In those cases, the module could use internal element IDs instead of the user-defined element IDs from the model, potentially affecting key results such as view factor sums, net radiative fluxes, and temperatures, among others.

Bug 125718: Models containing enclosure radiation may trigger a memory access crash in the VUFAC module

Fixed versions: 2606.2, 2512.8

This fix addresses a potential crash during the solve process for models that include enclosure radiation.

The crash was caused by an out-of-bounds memory access in the VUFAC module. Because the behavior depends on runtime memory allocation, it does not occur consistently and is not always reproducible.

The correction improves solver stability.

Bug 126469: Bottom side specification of a thermal coupling may incorrectly propagate to subsequent thermal couplings in the INPF file

Fixed versions: 2606.1, 2512.7, 2506.11, 2412.17

This fix addresses a case where the side specification from one thermal coupling could incorrectly carry over to subsequent thermal couplings in the INPF file.

The problem occurred when the bottom side was selected for the primary and/or secondary selection of a thermal coupling. Subsequent thermal couplings with side specification enabled could incorrectly continue using the bottom option, even when the top option was selected.

This behavior has been corrected so that each thermal coupling uses the side specification chosen by the user in the dialog box.

User Story 126735: Implement generalized cylindrical-space stream network generation and wall-to-duct coupling for mixed-direction thermal streams

This improvement addresses a limitation in the legacy cylindrical stream implementation for one-sided and two-sided streams.

The previous implementation only supported cylindrical streams aligned with a principal cylindrical axis, such as pure axial, pure radial, or pure circumferential directions. Mixed-direction streams could fall back to an incorrect planar treatment.

A new unified algorithm now correctly generates duct nodes and couples wall elements to duct elements for cylindrical streams with arbitrary direction vectors. This includes mixed-direction streams such as helical paths (axial and circumferential) and conical paths (radial and axial).

The wall-to-duct coupling now uses proportional area distribution based on overlap in the projected stream coordinate. This replaces the Cartesian nearest-element assignment, which could produce incorrect coupling for curved surfaces.

As a result, solid and fluid temperature results are more accurate for models that use mixed-direction cylindrical streams.

Bug 126839: Incorrect view factors may be computed due to invalid memory access for the deterministic and Monte Carlo ray-tracing calculation methods

Fixed versions: 2512.0, 2506.4, 2412.12

An investigation determined that, under certain conditions, the CPU-based deterministic and Monte Carlo ray-tracing algorithms could incorrectly classify visibility between radiating surfaces. This could result in inaccurate radiative view factors and, consequently, incorrect radiation heat transfer and temperature predictions.

No additional code changes were required specifically for this bug report. The investigation confirmed that the underlying issue had already been resolved by the correction delivered under Bug 114628: Memory access violation and operating system-dependent behavior during Monte Carlo ray tracing calculations on the CPU may cause the VUFAC module to crash. That fix was documented in the README files for tmgver2506 and tmgver2412, beginning with releases 2506.4 and 2412.12, respectively. It corrected the underlying memory access issue affecting the CPU ray-tracing calculations. As a result, the ray-tracing algorithms now correctly determine surface visibility, producing more accurate view factors and improved radiative energy balance.

Users may observe differences in computed view factors and temperatures compared to earlier releases. These differences are expected and reflect corrected solver behavior rather than a regression.