The Pain Point: The Bracket That Cracks First

A radiator is mounted to a vehicle, not to a test bench. In service it is shaken by road input, loaded by hose forces, heated and cooled by the engine, and pressed by fan-induced vibration. Almost every radiator failure that is not a corrosion leak is a fatigue crack, and almost every fatigue crack starts at a mounting point or a section change rather than in the core.

The paradox is that light weighting makes this worse. Reducing wall thickness and bracket thickness improves the vehicle's mass budget and the radiator's cost, but fatigue life scales roughly with the cube of thickness in bending. A bracket that is 20 % thinner can lose nearly half its fatigue margin, even though its static strength still passes.

Designers who validate by burst pressure and static load are validating the wrong failure mode.

The Principle: Stress Concentration, Resonance and Thermal Strain

Fatigue starts where geometry concentrates stress. Any hole, notch, sharp corner or abrupt thickness change raises local stress above the nominal value. A radius of 2 mm at a bracket root can reduce the stress concentration factor from around 2.5 to 1.4 compared with a sharp corner - a change that translates almost directly into fatigue life.

Resonance amplifies whatever input exists. A radiator core with side plates and tanks is a plate-like structure with natural frequencies typically in the range excited by engine idle and road input. If a mounting bracket or a hose connection places the structure's natural frequency inside the engine's excitation band, the vibration amplitude rises sharply and fatigue life collapses.

Thermal strain is unavoidable and must be accommodated. The core reaches over 100 °C while the mounting points stay near ambient. The resulting expansion must be allowed to occur somewhere. If every mounting point is rigid, the expansion becomes stress; if the mounts are compliant, the expansion becomes movement.

Hose loads are constant, not cyclic in the obvious sense. A radiator hose under pressure applies a steady force to the inlet neck, and that force is superimposed on vibration. The combination is often more damaging than either alone.

Brackets and tanks interact as a system. A tank wall that is stiff enough on its own becomes a flexible panel once a bracket is bolted through it. The bolt transfers load into a thin plate, and the plate responds by local bending. The result is that the highest stress in the assembly often sits in the tank wall, not in the bracket that was designed for the load.

Corrosion and fatigue multiply each other. A fatigue crack in aluminium needs a stress raiser to start, and a corrosion pit is an ideal one. A bracket root that would survive a million cycles in dry conditions can fail in a fraction of that when road salt has produced pitting at the same location. This is why durability tests that separate corrosion from vibration tend to overestimate life.

Aluminium has no endurance limit at high cycles in the way steel does. Aluminium alloys do not exhibit a true fatigue limit, so design must be based on a finite life target rather than on staying below a threshold stress.

Equipment Classification: Mounting and Structural Components

Side plates and reinforcements. The structural spine of a brazed core, carrying bending loads between the tanks.

Mounting brackets and pins. Rubber-isolated in most automotive applications, rigid in many industrial ones.

Upper and lower tanks with integral mounts. Cast or moulded tanks that combine a pressure boundary with attachment points - the highest-risk combination.

Inlet and outlet necks. Loaded by hose forces, thermal expansion and pressure pulsation simultaneously.

Fan shrouds and fan mounting. Often attached to the radiator, adding mass and excitation.

Vibration isolators. Rubber grommets, bushings and flexible mounts, whose stiffness must be specified as carefully as any structural dimension.

Practical Guide: Designing for the Load History

Define the load case before the geometry. Engine excitation orders, road input spectrum, hose forces and thermal range. Design without this data is guesswork.

Put radii on every bracket root and every boss transition. This is the cheapest fatigue improvement available and it is routinely omitted on drawings.

Isolate in three axes, not just one. A rubber grommet that allows vertical movement but locks horizontal motion transfers the load elsewhere. Specify isolator stiffness in all directions.

Avoid over-constraining the core. Three-point mounting with compliance in the plane of thermal expansion is generally better than four rigid points.

Keep the bracket natural frequency away from excitation bands. Stiff and light is better than heavy and stiff; adding mass to solve a vibration problem often makes it worse.

Support long unsupported spans. A hose neck with no support and a long overhang will fatigue at its root. A small gusset or a tie to the tank wall is usually enough.

Validate by accelerated vibration test with thermal cycling superimposed. Vibration at ambient and thermal cycling without vibration both miss the combined effect.

Check the assembly for preload. A bracket that is forced into position during assembly already carries stress before the vehicle moves.

Keep the hoses from becoming structural members. A hose that is stretched to reach a neck loads that neck permanently and transmits engine movement into the radiator. Route for slack, and support the neck if the straight run is long.

Test the worst case, not the nominal one. The worst case is a fully warm core, a cold ambient bracket, a pressurised system and a rough road, all at once. That combination is what a vehicle delivers on a winter morning after a long climb.

Field Experience: The Four-Point Mount That Failed at 60,000 km

A retrofitted radiator for a bus application used four rigid mounting points, one at each corner of the frame. The original design had used three points with rubber isolation. The change was made for assembly simplicity.

Failures began at around 60,000 km, always at the lower right bracket, always as a crack running from the bolt hole toward the tank corner. Static load testing at three times the calculated service load produced no failure.

Instrumented vehicle testing showed the cause. The four rigid points over-constrained the frame, so the thermal expansion of the core during warm-up was converted into a bending moment at the brackets. Combined with road vibration, the lower right corner saw a stress range more than double the design assumption. The crack initiated at the sharp edge of the bolt hole, which had been drilled without a chamfer.

The fix was to return to three-point mounting with rubber isolators, add a chamfer and a radius at the bracket root, and increase the bracket thickness by 0.5 mm. No further failures were reported in the following two years of service.

The valuable observation was that the original three-point design had been correct not because of engineering elegance but because it allowed the structure to expand. Simplifying it for assembly removed the compliance that made it work.

Summary

Aluminium parts for radiators are fatigue structures first and pressure vessels second. The loads come from the vehicle, the hoses and the thermal cycle, and they concentrate at the small geometric details that drawings tend to leave unspecified. Light weighting is entirely compatible with durability, but only when radii, isolation and load paths are designed deliberately rather than inherited.

Outlook

Two trends will raise the stakes. Higher system temperatures in electrified thermal management increase thermal strain and reduce the fatigue margin of aluminium alloys. And the shift toward thinner, larger cores increases the plate-like behaviour of the structure, moving natural frequencies and making vibration isolation more critical. Expect simulation-driven mounting design to become standard even at moderate volumes, and expect customers to demand vibration-plus-thermal validation rather than separate tests - because that combination is where the real failures occur.