Engineering tolerance

• 1. Introduction

  ›

  Engineering tolerance sets the permissible limit of variation allowed in a dimension, a material property, or a measured condition like temperature or humidity. A temperature that runs too hot, or a dimension that runs too wide, is simply called noncompliant, rejected, or out of tolerance once it crosses that limit. Every manufactured part, every material, and every measured condition carries some version of this permitted give, whether the context is mechanical, electrical, or a matter of physical space. But how wide should that give actually be, and who decides where the line falls? Some engineers have gone further still, questioning whether the standard method of drawing that line is even the right way to think about quality.

• 2. Setting the Limits

  ›

  Design of experiments and formal engineering evaluations let engineers test how a proposed tolerance actually behaves before it gets locked into a specification. Scientific principles and professional experience shape that decision too, but no specification, however carefully written, guarantees that production will comply with it.

  Actual production always carries some inherent variation between what goes in and what comes out, and measurement itself introduces error and statistical uncertainty. In a normal distribution of measured values, the outer tails can extend well beyond three standard deviations from the process average. Part of that tail can end up sitting outside the specified tolerance entirely.

  Total Quality Management is one example of the quality management systems that, paired with process controls, are meant to keep actual output within the desired range. A process capability index is the specific measure used to show how tolerances line up against what a process really produces. The choice of tolerance also depends on the sampling plan behind it, including a measure called the Acceptable Quality Level, which raises a real design choice. Should tolerances be strict enough to guarantee near-total conformance, or is some small percentage of out-of-tolerance parts an acceptable cost of doing business? That question about how much imperfection is tolerable is exactly where a very different way of thinking about tolerance enters the picture.

• 3. Beyond the Goal Posts

  ›

  Genichi Taguchi challenged the standard way of thinking about tolerance by comparing it to the goal posts in a football game, where any shot landing between the posts counts exactly the same. In that traditional two-sided model, a measurement sitting just inside the limit is treated as no different from one landing precisely on target.

  Taguchi's alternative holds that the best product is the one whose measurement lands exactly on target, with loss increasing the further a value drifts from it. That relationship between deviation and cost is what's called the Taguchi loss function, or quality loss function, and it is the key principle behind a separate system known as inertial tolerancing.

  Research led by M. Pillet and colleagues at the Savoy University has since driven adoption of this approach within specific industries. Publication of the French standard NFX 04-008 gave manufacturers a formal document to weigh the idea further. That standard turned an argument about loss and deviation into something applicable on the shop floor, where tolerance stops being philosophy and becomes the physical clearance between a bolt and the hole it fits into.

• 4. Fitting Metal to Metal

  ›

  A ten millimeter shaft designed for a sliding fit inside a hole shows how the terminology actually gets used. Basic size is the nominal diameter shared, generally, by both the shaft and the hole around it. Lower deviation measures the gap between the smallest allowed size and that basic size, while upper deviation measures the gap on the larger side. Fundamental deviation is the minimum size difference from basic size; when it's greater than zero, the bolt stays smaller than basic size and the hole stays wider.

  For that ten millimeter sliding fit, the shaft might range from 9.964 to 10 mm, a lower deviation of 0.036 mm with zero fundamental deviation. The matching hole might range from 10.04 to 10.076 mm, a 0.04 mm fundamental deviation and a 0.076 mm upper deviation. That combination produces a clearance fit somewhere between 0.04 mm, called the Maximum Material Condition, and 0.112 mm, called the Least Material Condition. Both parts in this example share the same 0.036 mm tolerance range, known as their International Tolerance grade, though that need not always hold.

  When a drawing specifies no other tolerance, machinists default to a fixed scale keyed to decimal places: plus or minus 0.2 inches at one decimal place, tightening to 0.01 inches at two, 0.005 inches at three, and 0.0005 inches at four.

  International Tolerance grades label each standard with a letter, capitals for holes and lowercase for shafts, plus a number. H7 paired with h6 is a common combination that produces a tight fit. An H7 hole with a 10 mm base dimension may run up to 0.015 mm larger than that base and never smaller. The matching h6 shaft may run up to 0.009 mm smaller than the base and never larger. This standardized approach is known as Limits and Fits and is documented in ISO 286-1:2010. Where a mechanical fit is measured in thousandths of a millimeter, other components state their tolerance an entirely different way, as a percentage of a labeled value.

• 5. Tolerance in Ohms

  ›

  A resistor labeled 100 ohms with a stated tolerance of plus or minus 1 percent is considered acceptable anywhere between 99 and 101 ohms. Critical components can carry tighter demands still, requiring that resistance stay within tolerance across a specified temperature range and over a specified lifetime.

  Most commercially available resistors, capacitors, and some small inductors carry coloured bands that indicate both their value and their tolerance. High-precision components built to non-standard values instead carry the numbers printed directly on them. Low tolerance means only a small deviation from a component's stated value when it's new and operating normally at room temperature. Higher tolerance simply means the component can range further from its labeled value and still count as acceptable. That same logic, an accepted margin around a labeled value, scales up dramatically once the thing being measured is the physical gap between a vehicle and the space it has to fit through.

• 6. Room to Pass

  ›

  In civil engineering, clearance measures the gap between the loading gauge and the structure gauge for railroad cars and trams. For any other vehicle, clearance is the space between its size and the width or height of a doorway, an overpass, or a tunnel it must pass through. The same idea covers the air draft available under a bridge for a boat, and the width of a lock or the diameter of a tunnel for other watercraft. For a waterway, clearance also covers the difference between a vessel's deep draft and the stream bed or sea bed beneath it. That calculation, run against a stream bed or a sea bed, is what keeps a fully loaded vessel from ever finding out the hard way how much clearance it actually had.