Tolerance Definition in Engineering: Managing Precision in Fabrication

Tolerance in engineering is the permissible variation in a dimension, form, location, or other characteristic of a manufactured part. A dimension on a drawing that reads Ø 25.00 ± 0.05 mm communicates three values: the nominal (target) size of 25.00 mm, the upper limit of 25.05 mm, and the lower limit of 24.95 mm. Any part whose actual dimension falls within this 0.10 mm band is in tolerance and acceptable. Any part outside this band is out of tolerance and must be rejected, reworked, or dispositioned.

The tolerance exists because perfect manufacturing does not exist. Every machining operation, every laser cut, every press brake bend, and every casting leaves geometric variation in the produced part. The engineering tolerance defines how much variation is acceptable for the part to function correctly in its assembly. Set the tolerance too tight, and most parts fail inspection, and manufacturing costs soar. Set it too loose, and parts fail to assemble, bind, or lack the dimensional integrity the design requires.

Tolerance Definition: The Complete Vocabulary

Engineering tolerance has a precise vocabulary defined in ASME Y14.5-2018 and the ISO GPS (Geometrical Product Specifications) standards. Using these terms correctly prevents the misunderstandings between design, manufacturing, and inspection that produce defects and disputes.

Nominal dimension (basic dimension in GD&T): The theoretically exact target value. Ø 25.00 mm is the nominal. In GD&T, the basic dimension is the exact value from which tolerances are applied — it appears in a rectangular box (⏹) on the drawing.

Upper deviation (upper limit): The algebraic difference between the maximum limit and the nominal dimension. For ± 0.05 mm, the upper deviation is +0.05 mm; the upper limit is 25.05 mm.

Lower deviation (lower limit): The algebraic difference between the minimum limit and the nominal dimension. For ± 0.05 mm, the lower deviation is −0.05 mm; the lower limit is 24.95 mm.

Tolerance: The total permissible variation — the difference between the upper and lower limits. For ± 0.05 mm, tolerance = 25.05 − 24.95 = 0.10 mm.

Tolerance zone: The geometric region within which the controlled feature must lie. For a linear dimension, a band of width equal to the tolerance. For a position tolerance, a circular cylinder of diameter equal to the tolerance.

Unilateral tolerance: A tolerance specified entirely on one side of the nominal. Example: 25.00 +0.10 / 0.00 mm — the dimension may only be 25.00 to 25.10 mm; no undersizing is permitted. Used where the design function requires the dimension to deviate only in one direction from nominal.

Bilateral tolerance: A tolerance split about the nominal, either symmetrically (± 0.05) or asymmetrically (+0.08 / −0.02). Symmetrical bilateral is the most common for general dimensions. Asymmetric bilateral is used when the functional requirement allows more variation in one direction than the other.

Types of Engineering Tolerance in Manufacturing

Dimensional tolerance (size control)

The most basic form — controlling the linear size of a feature. Applied to lengths, diameters, widths, depths, and spacing between features. Examples:

• Shaft diameter: Ø 20.00 +0.00 / −0.021 mm (unilateral, controls oversize)
• Plate thickness: 6.00 ± 0.15 mm (bilateral symmetric)
• Hole center spacing: 50.00 ± 0.10 mm (bilateral)

Dimensional tolerances alone cannot fully control the geometry of a manufactured part. A shaft that is 20.00 mm in diameter at one end and 19.99 mm at the other is within a ± 0.01 mm diameter tolerance — but it is tapered, which may violate the design intent. Geometric tolerances are required to control form, orientation, and location separately from size.

Geometric tolerances (GD&T)

Geometric dimensioning and tolerancing (GD&T) per ASME Y14.5-2018 provides a complete language for controlling the shape, orientation, location, and runout of features independently of their size.

Each GD&T control specifies a tolerance zone — a geometric region within which the controlled feature must fall. A flatness tolerance of 0.05 mm means the entire controlled surface must lie within two parallel planes separated by 0.05 mm. A position tolerance of Ø 0.20 mm means the axis of the controlled hole must lie within a cylindrical zone of diameter 0.20 mm centered on the true (basic) position.

The datum reference frame (covered in the datum definition article) provides the reference from which orientation and location GD&T controls are measured. Flatness and straightness (form controls) do not require a datum — they measure the feature relative to itself. Position and perpendicularity (location and orientation controls) require datums.

General tolerances

Not every dimension on a drawing receives an explicit tolerance. General tolerances — applied by default to all untoleranced dimensions — are specified in a note referencing a tolerance standard. Two common systems:

ISO 2768: Published by ISO for general-purpose machined parts. Specifies tolerances by tolerance class (f = fine, m = medium, c = coarse, v = very coarse) for both linear dimensions and angular features. For example, ISO 2768-m specifies ±0.10 mm tolerance for dimensions between 30 and 120 mm. A drawing note “Tolerances to ISO 2768-m” applies this default to every untoleranced dimension.

ASME B4.1 (preferred limits and fits): Provides a system of standardized fits (clearance, transitional, interference) with associated limit dimensions. Used primarily for shaft-in-bore and similar mating features where the functional relationship between the two mating dimensions (the fit) is the design requirement.

Fabrication default tolerances (laser cutting): For laser-cut parts, the fabricator may specify default achievable tolerances based on their equipment capability — typically ±0.1 mm to ±0.25 mm for profile dimensions on a well-calibrated fiber laser cutting machine, and ±0.5 mm to ±1.0 mm for bending dimensions after press brake forming. Features requiring tighter tolerances must be explicitly controlled with coordinate dimensions or GD&T.

Tolerance Stackup: Why Individual Tolerances Interact

Individual tolerances do not exist in isolation — they accumulate through an assembly. When a part with dimension A (tolerance ±0.05 mm) fits into a part with dimension B (tolerance ±0.10 mm) to produce an assembly dimension C, the tolerance on C is the worst-case combination of A and B.

Worst-case linear stackup

For a simple chain of dimensions A → B → C where C = A + B:

C_max = A_max + B_max C_min = A_min + B_min Tolerance on C = Tolerance on A + Tolerance on B

This is the worst-case arithmetic stackup — the total tolerance is the sum of all contributing tolerances. It guarantees every possible assembly will be in tolerance, but it produces tight individual tolerances on components to achieve a moderate assembly tolerance.

Example: A laser-cut bracket (dimension A, ±0.15 mm) is bolted to a machine frame (dimension B, ±0.20 mm) to locate a sensor. The sensor clearance gap is dimension C, and must be 2.00 ±0.25 mm. Is this achievable?

Worst-case stackup: ±(0.15 + 0.20) = ±0.35 mm — larger than the ±0.25 mm requirement. The assembly cannot be guaranteed to meet the sensor clearance requirement with these component tolerances. Either tighten the component tolerances, add an adjustable mount, or relax the sensor clearance requirement.

Statistical stackup (RSS method)

When a large number of parts are assembled, it is unlikely that all tolerances simultaneously reach their worst-case values. Statistical tolerance analysis — using root sum of squares (RSS):

Tolerance_C = √(Tolerance_A² + Tolerance_B² + Tolerance_C²)

For the same example: √(0.15² + 0.20²) = √(0.0225 + 0.04) = √0.0625 = ±0.25 mm

The RSS stackup meets the ±0.25 mm requirement exactly — acceptable if the engineer accepts that a small fraction of assemblies (approximately 0.27% at ±3σ) will exceed the tolerance limit. Statistical stackup is used in high-volume production where some percentage of out-of-tolerance assemblies is economically acceptable.

Tolerance Capability in Metal Fabrication Processes

Selecting the correct tolerance for a design requires knowing what each manufacturing process can reliably achieve. Specifying a tolerance tighter than the process’s natural capability ensures defects, not better parts.

These are representative ranges — actual capability depends on the specific machine, material, setup, and part geometry. For laser cutting, the laser cutting tolerances guide covers the specific factors (material type, thickness, assist gas, focal length) that move achievable tolerance toward ±0.05 mm or toward ±0.25 mm.

The tolerance-cost relationship: Tolerances tighter than the process’s natural variation require additional process controls — slower cutting speeds, additional inspection steps, more frequent machine calibration, or secondary operations. The manufacturing cost typically increases by 2–10× per order-of-magnitude tightening of tolerance below the process’s natural capability. Tolerancing a laser-cut bracket to ±0.05 mm when ±0.25 mm would function correctly increases cost without improving function.

Tolerance Definition in Laser-Cut and Sheet Metal Parts

What laser cutting controls

A fiber laser cutting machine controls the position and profile of cut features with high accuracy relative to the programmed toolpath. The laser accurately follows the programmed kerf center position — typically ±0.05 mm repeatability on a well-maintained fiber laser. What the laser does not automatically control:

• The kerf width (varies with cutting speed, gas pressure, and material) — this must be compensated in the cutting program
• The perpendicularity of the cut wall (varies with focus position; typically 0.5–2.0° taper)
• The squareness of the part to the sheet surface (perpendicularity of cut edges to the top face)

For holes specifically, the nominal hole position matches the programmed position to ±0.05–0.15 mm. The hole diameter is controlled by adjusting the kerf offset in the cutting program. Holes smaller than approximately 2× material thickness are difficult to cut cleanly and should have their dimensional tolerances specified accordingly.

What press brake bending controls and what it doesn’t

Press brake bending controls bend angle and the position of the bend line relative to the back gauge stop. What bending does not control:

• The overall developed length of the blank (set by the laser cutting program and the K-factor calculation)
• The angle in the final part (affected by springback — dependent on material yield strength and material thickness at the bend location)
• The straightness of the bent flange (affected by material bow and the press brake ram straightness)

Bend angle tolerances in production range from ±0.5° for well-controlled setups with springback compensation to ±2° for standard production. If a part requires ±0.25° angle control, either a precision press brake setup with angle-measuring feedback, or a secondary coining operation is required.

The direct relationship between material yield strength and springback — and therefore between yield strength and achievable angular tolerance — is the reason why our yield in engineering guide and aluminum yield point articles are referenced upstream from press brake process articles. You cannot control the press brake process without knowing the yield strength of the specific material being bent.

Frequently Asked Questions About Tolerance in Engineering

Q: What is tolerance in engineering?

Tolerance is the total permissible variation in a dimension, form, orientation, or location of a manufactured feature. A dimension of 25.00 ±0.05 mm has a tolerance of 0.10 mm — the feature may be anywhere from 24.95 mm to 25.05 mm and still be acceptable. Tolerance acknowledges that perfect manufacturing is impossible and defines how much imperfection is acceptable for the part to function correctly.

Q: What is the difference between tolerance and accuracy?

Accuracy is a measure of how close the actual result is to the target — how close to 25.00 mm the part actually is. Tolerance is the allowed band around the target — how far from 25.00 mm the part may be and still be accepted. A machine can produce parts with high accuracy (very close to the target) but low precision (variable from part to part), or high precision (consistent part to part) but low accuracy (consistently offset from the target). A high-quality manufacturing process is both accurate (on target) and precise (repeatable).

Q: What is a ± tolerance vs. a unilateral tolerance?

A ± (bilateral) tolerance allows variation above and below the nominal in equal amounts: 25.00 ±0.05 mm allows 24.95 to 25.05 mm. A unilateral tolerance allows variation in only one direction: 25.00 +0.10/0.00 mm allows 25.00 to 25.10 mm (no undersizing permitted). Unilateral tolerances are used when the design function requires the dimension to remain at or above (or at or below) the nominal — for example, a minimum clearance that must always be present regardless of other assembly variations.

Q: What tolerance does laser cutting achieve?

A fiber laser cutting machine on a well-calibrated system typically achieves ±0.05–0.15 mm on profile dimensions in flat sheet. Hole position is typically ±0.05–0.15 mm from the programmed center. These values depend on material type, material thickness, cut speed, focal length, and machine condition. Features requiring ±0.05 mm are achievable but require specific parameter control. Features requiring ±0.02 mm or tighter typically require secondary machining operations after laser cutting.

Q: What does ISO 2768 mean on a drawing?

ISO 2768 is an ISO standard that defines default tolerances for linear dimensions and angular features on machined parts that do not have an explicit tolerance called out. A note “Tolerances: ISO 2768-m” applies medium-class default tolerances — for example, ±0.10 mm for dimensions between 30 and 120 mm, ±0.15 mm for 120–400 mm. A note “ISO 2768-f” applies fine-class tolerances: ±0.05 mm for 30–120 mm. Any dimension on the drawing that needs tighter (or looser) control than the default class must have that tolerance explicitly stated.

learn more:

ISO 9001:2026 Update: What CNC and Laser Fabrication Shops Must Know

References

1. American Society of Mechanical Engineers (ASME), ASME Y14.5-2018: Dimensioning and Tolerancing, ASME, New York, 2018. (Complete GD&T standard: tolerance types, GD&T symbols, tolerance zones, datum reference frames, MMC/LMC modifiers, tolerance stackup methodology; the primary US standard for engineering drawing tolerance specification.)
2. International Organization for Standardization, ISO 2768-1:1989 and ISO 2768-2:1989, General Tolerances — Part 1: Tolerances for Linear and Angular Dimensions Without Individual Tolerance Indications; Part 2: Geometrical Tolerances for Features Without Individual Tolerance Indications, ISO, Geneva. (General tolerance classes f, m, c, v for linear dimensions and angular features; the default tolerance system referenced by note on engineering drawings for untoleranced dimensions.)
3. American Society of Mechanical Engineers (ASME), ASME B4.1-1967 (R2018): Preferred Limits and Fits for Cylindrical Parts, ASME, New York. (Standardized clearance, transitional, and interference fit systems; shaft and hole basis fit specifications; design reference for mating dimensional tolerances in shaft-bore assemblies.)
4. Harry, M., and Schroeder, R., Six Sigma: The Breakthrough Management Strategy Revolutionizing the World’s Top Corporations, Doubleday, New York, 2000; and Creveling, C.M., Tolerance Design: A Handbook for Developing Optimal Specifications, Addison-Wesley, Reading, MA, 1997. (Statistical tolerance analysis (RSS method); worst-case vs. statistical stackup comparison; process capability (Cpk) relationship to tolerance and process variation; the engineering management basis for the tolerance stackup analysis presented in this article.)

• Author
• Recent Posts

Dr. Liu Che

Dr. Liu Che holds a Ph.D. in Engineering from Dalian University of Technology and completed postdoctoral research at the Mechanical Engineering Postdoctoral Fellowship of Harbin Institute of Technology. His research focuses on core algorithms for large-scale laser equipment integration, laser cladding remanufacturing, and real-time intelligent monitoring systems for processing. He possesses profound academic expertise in cutting-edge fields such as laser beam spatial shaping and the mechanism analysis of laser processing of special alloys. He has led numerous solutions to technical challenges in cutting and surface modification of high-hardness materials in the heavy machinery industry, demonstrating strong theoretical foundations and outstanding engineering implementation capabilities.

Latest posts by Dr. Liu Che (see all)

• Laser Coating Removal Without Damaging the Substrate - September 30, 2026
• Laser Cleaning for Paint Adhesion: Improving Surface Energy - September 29, 2026
• Common Mistakes in Laser Rust Removal Applications - September 23, 2026