Although absolutes exist in design engineering, combining as integers on the drawing board, or some CAD design program, the real world isn’t nearly as accommodating. Whether due to exotic material properties or machine limitations, tiny imperfections creep into production environments. Basically, no manufacturing process or mechanical industry solution is perfect, which is why they have tolerances.

Dimensionally, the variances referred to will be in the range of plus or minus a few micrometers of error. They’re permissible, up to a point. As long as the variation in physical dimensions doesn’t impact mechanical performance, tolerance is an acceptable part of the manufacturing cycle. Again, the values are invisible to the human eye and are only measurable by engineering instruments.

These minor deviations are rarely problematic in isolation and exclusive to those that utilize condensed assembly builds. However, complex mechanical systems are built out of intricate assemblies, combining thousands of moving parts. Even if each individual part exhibits a tiny amount of dimensional variance, all of these errors will ‘stack’ until they exceed permissible system tolerances. This fact is important enough to have created an entire science around it, leading to the development of specialized tolerance analysis software packages.

Explaining the nature of cumulative system errors

Essentially, one out-of-true tolerance error is permissible in a small mechanical assembly, but if we follow this thought to its logical conclusion, larger systems would suffer as the cumulative effects of enough of these additive errors compromised overall performance, turning a smoothly meshing harmonious whole into a dissonant mess. That, in the proverbial nutshell, is why tolerance stacking can heavily skew the intended fit, alignment or functionality of a mechanical system.

Adopting a know-your-enemy approach to deal with tolerance stacking, let’s see what we can do to understand the finer points of this issue. Consider this two-dimensional example. A jigsaw puzzle is cut by a die cutting machine. Each part is cut to a given tolerance, perhaps as small as a plus or minus 0.05 mm edge deviation. Now, if it’s put together, all parts fit and a picture appears. If this value increases, though, the pieces wouldn’t fit, the picture would distort and the challenge would have gone out of the moment. Fun is the cost in a puzzle; the impact on a large mechanical subsystem is far more critical. It also goes beyond two-dimensional thinking. Assessing three types of tolerance stacking:

Linear stacking

Closer to our jigsaw puzzle example, linear tolerance stacking takes place along a straight line. Consider a powertrain — jigsaws aren’t considered relevant engineering machines — then all the gears in that mechanical power delivery system will exhibit finite tolerance levels, which then accumulate to form a single larger stacked value. It’s plain, old arithmetic. When managed, higher quality builds can be facilitated to assure the calculated stack-up equals an adequate system-wide fit.

Worst-case stacking

This is a step beyond the linear category. The management of system-wide tolerances assumes the maximum possible measurement deviation, again on a single axis, and the cumulative error is considered a worst-case scenario. The maximum, or minimum, measurements entering the tolerance calculating formulas — mostly additive arithmetic — are envisaged as situation extreme, which isn’t a generally viewed circumstance, not unless we’re talking about high tolerance aviation or medical systems.

Geometrical tolerance stacking

If the previous tolerance stacking types work across two-dimensions, then geometric tolerance stacking Sigmetrix.com is the three-dimensional equivalent. It’s also far closer to real-world engineering, using angular measurements, parts flatness, tolerance parallelism and several other dimensional attributes that rely on higher mathematical functions, not simple arithmetical calculations. For example, the powertrain talked about earlier would likely have several gears on a single shaft, which then further reduce or increase power by altering rotational speeds and work done via gear teeth sizing and diameter changes.

Each of these stacking categories addresses unique aspects of dimensional variation, whether through simple additive calculations or more complex geometric relationships. By selecting the appropriate stacking type, engineers can better predict system behavior, reduce assembly risks and ensure optimal performance.

Analyzing techniques for tolerance stacking

Engineering software packages, like the previously mentioned Sigmetrix solution, have been provided to address tolerance stack-up issues. They integrate into popular CAD programs as dimensional analysis tools, enabling engineers to simulate and analyze how tolerances from individual components interact within a machine assembly.

As for the core analysis techniques, worst-case analysis is one tool, but it creates exacting designs based on unrealistic extremes, leading to conservative designs, tight manufacturing specifications and increased production costs. Statistical tools, such as the Root Sum Square (RSS) approach offer a more efficient, and realistic, solution. For example:

RSS=((0.12)+(0.22)+(0.32))

Say that​ RSS calculations, using statistical approximation, create a balance between precision and practicality. When various tolerance values are plugged into the formula, leeway isn’t as tight, costs are reduced and component designs are easier to manufacture using current manufacturing setups.

More complex statistical analysis, perhaps using Monte Carlo sample algorithms, can be performed to gain deeper insights into stack-up variances, but tolerance stacking represents a densely data-driven field. Specialized CAD packages have evolved to incorporate tolerance stacking addons. Like most worthy software solutions, they pay for themselves over time, eliminating inefficient trial-and-error testing, reducing conservative worst-case scenario tight system specs and entirely cutting costly design errors during production.

Closing thoughts on precision-based machinery

Everything up to this point, while not vague, hasn’t been exactly machine-specific, either. A real machine would help add context and tie everything together. For this purpose, we chose the GE Aerospace GE9X aerospace engine. It’s boasted as being the largest commercial aircraft engine ever manufactured. The 134-inch front fan engine has thousands of moving parts, all of which must work in perfect harmony for thousands of hours while under stress.

Just for a minute, picture the 5,000 hours of ground testing before the engine was even classed as airworthy. Special ceramics and high-tensile lightweight alloys were cut and heat treated, and they all left their manufacturing facilities with exacting micro millimeter tolerances attached. Even so, multiplied by however many parts are in that beautifully engineered engine, stack-ups are inevitable. Tolerance analysis keeps the components fitted in place, exactly as designed and as close to absolutely meshed as possible.

Pushed through exhaustive simulations and statistical analysis, the GE9X demonstrates the power of precise engineering in large-scale systems. Its stack-ups are bound to be negligible. The same can be said for smaller though no less precision-based machines, such as those used in medical devices. Their tolerance stack-ups have been calculated to the nth degree.

No matter the fit tightness, no matter the function or component scale, tolerance stacking keeps systems — even those built from ten-thousand interlocking parts — not only viable but flawlessly aligned.