If you are searching for how to translate engineering tolerances, how to translate measurement uncertainty, or how to preserve plus/minus values, significant figures and specification limits across languages, treat the mathematical relationship as part of the meaning. A tolerance such as ±0.05 mm is not decoration: it defines what counts as acceptable.
Tolerance translation matters in engineering drawings, manufacturing specifications, laboratory reports, quality-control records, scientific papers, calibration certificates, construction documents and product datasheets. A single reversed inequality, missing minus sign or altered decimal place can change a pass condition into a fail condition or enlarge a permitted range by a factor of ten.
This guide gives a practical method for translating tolerances, uncertainty statements and significant figures without changing technical meaning. It explains symmetric and asymmetric tolerances, upper/lower limits, ≤ and ≥ operators, confidence intervals, measurement uncertainty, rounding, unit conversion, significant figures and how to verify the translated result mathematically rather than relying on fluent prose alone.
Why technical precision can disappear in translation
A tolerance states how far a value may vary while remaining acceptable. An uncertainty statement describes limits on what a measurement can confidently establish. These ideas are related but not identical, and translation should not collapse one into the other.
Notation itself carries meaning. The symbols ±, ≤, ≥, <, >, ≈ and % have logical force. Decimal places can signal precision. Parentheses, superscripts and subscripts can distinguish statistical uncertainty, specification limits or dimensional notation.
Technical translation often requires unit handling. If units remain unchanged, the numeric values usually remain unchanged. If units are converted, the tolerance or uncertainty must be converted with the same dimensional relationship and appropriate precision.
The safest workflow treats numbers and operators as data, terminology as technical language, and the final specification as something that can be tested independently of the sentence.
A reliable translation method
1. Identify whether the statement is tolerance, uncertainty or limit
Do not translate all variability language as “margin.” Determine whether it defines acceptance, measurement knowledge, statistical confidence or design allowance.
2. Protect operators and signs
Treat ±, +, −, ≤, ≥, < and > as logic. Verify them character by character after translation.
3. Lock units and dimensional meaning
Identify the quantity and unit before changing anything. Length, temperature, pressure, voltage and concentration require different conversion rules.
4. Preserve decimal precision
Do not drop trailing zeros or decimal places casually. In technical contexts they may communicate reporting precision or specification resolution.
5. Handle asymmetric tolerances explicitly
A +0.10/−0.05 tolerance is not the same as ±0.10 or ±0.05. Keep upper and lower deviations separate.
6. Convert units only when authorised
If the target document requires another unit system, convert nominal value and tolerance consistently, then verify rounding against the relevant standard.
7. Restate the interval as QA
Calculate the minimum and maximum values implied by the translated tolerance. This exposes sign and arithmetic errors quickly.
8. Check terminology against the field
Use domain-appropriate terms for tolerance, uncertainty, accuracy, precision, repeatability, allowance and limit. These words are not interchangeable in technical disciplines.
Twenty recurring tolerance and uncertainty problems
1. Symmetric plus/minus tolerance
This problem appears when a nominal value can vary equally in both directions. A source expression such as 10.0 ± 0.2 mm does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve nominal value, plus/minus sign, tolerance magnitude and unit. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, confirm the translated interval is 9.8 to 10.2 mm. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
2. Asymmetric tolerance
This problem appears when upper and lower deviations differ. A source expression such as 25.00 +0.10/−0.05 mm does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Keep the two deviations separate and attached to the correct direction. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, confirm minimum 24.95 and maximum 25.10. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
3. Upper limit
This problem appears when the value must not exceed a threshold. A source expression such as ≤ 5 ppm does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve the less-than-or-equal operator and threshold. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, state in words: five ppm or less. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
4. Lower limit
This problem appears when the value must meet or exceed a threshold. A source expression such as ≥ 98% does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve the greater-than-or-equal operator. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, state in words: at least ninety-eight percent. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
5. Strict inequality
This problem appears when equality is not allowed. A source expression such as < 10 ms does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Do not change < to ≤ in prose or notation. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, check that exactly 10 ms would still be excluded. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
6. Range specification
This problem appears when acceptable values lie between endpoints. A source expression such as 4.5–5.5 V does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve both endpoints, range direction and unit. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, check that the target does not imply approximately 5 V instead of a bounded range. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
7. Percentage tolerance
This problem appears when variation is relative rather than absolute. A source expression such as ±2% does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Keep percent basis and do not turn it into two percentage points unless context explicitly means that. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, apply the percentage to a sample nominal value to confirm interpretation. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
8. Dimensional tolerance
This problem appears when engineering geometry specifies allowable deviation. A source expression such as 50 ± 0.05 mm does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Use target engineering terminology while preserving the dimension exactly. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, compare with drawing callout and inspection criteria. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
9. Geometric tolerance
This problem appears when GD&T-style symbols define form or position controls. A source expression such as position tolerance ⌀0.2 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Protect standard symbols and translate explanatory text only. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, verify symbol, diameter mark, datum references and numeric value. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
10. Temperature tolerance
This problem appears when temperature can be expressed in different scales. A source expression such as 20 ± 2 °C does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Keep the scale unless conversion is required; if converting, convert endpoints rather than only the ± term naively. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, calculate the full acceptable interval before and after conversion. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
11. Pressure specification
This problem appears when multiple unit systems are common. A source expression such as 100 ± 5 kPa does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve unit or convert nominal and tolerance together. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, verify dimensional conversion and rounding. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
12. Electrical rating tolerance
This problem appears when component values have nominal values and tolerances. A source expression such as 10 kΩ ±5% does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Keep resistance unit, multiplier and tolerance. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, calculate allowed resistance range as a check. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
13. Calibration uncertainty
This problem appears when a certificate reports measurement uncertainty rather than product tolerance. A source expression such as U = 0.3 °C, k = 2 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve uncertainty notation and coverage factor; do not translate it as allowed product variation. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, check metrology terminology and certificate legend. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
14. Standard uncertainty
This problem appears when statistical uncertainty uses a particular definition. A source expression such as u = 0.15 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Use correct metrology term and preserve symbol. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, verify whether u and U are distinguished in the source. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
15. Confidence interval
This problem appears when a statistical estimate includes an interval. A source expression such as 95% CI 1.2–1.8 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve confidence level, interval endpoints and statistical label. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, ensure 95% is not translated as measurement accuracy. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
16. Significant figures
This problem appears when reported digits reflect precision. A source expression such as 3.40 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Do not simplify to 3.4 when trailing zero is meaningful. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, compare source reporting precision and field convention. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
17. Rounding instructions
This problem appears when the source specifies how results are rounded. A source expression such as round to nearest 0.01 does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve rounding unit and direction. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, test with sample values above and below the midpoint. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
18. Approximate values
This problem appears when the source signals estimation rather than tolerance. A source expression such as ≈ 100 m does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Keep approximate meaning and do not convert it into an acceptance limit. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, distinguish approximation symbol from ± tolerance. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
19. Nominal values
This problem appears when a stated value is a designation rather than exact measurement. A source expression such as nominal diameter 50 mm does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Use the target technical term for nominal and avoid implying exact measured size. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, check the applicable standard or product definition. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
20. Pass/fail statements
This problem appears when the numeric specification controls acceptance. A source expression such as Pass if deviation ≤0.5% does not merely contain a number; it defines an acceptable range, a measurement limit, a confidence statement or the precision of the reported result. Small changes in signs, units, decimal places or comparison operators can change whether a part passes inspection or a result satisfies a specification.
Preserve operator, threshold and decision logic. The translator should treat the numeric relationship as protected logic. Translate the explanatory language around it, localise notation only when the target standard requires it, and preserve every operator, inequality direction, tolerance sign, unit and significant digit unless an authorised conversion is being performed.
For quality assurance, test one value below, equal to and above the threshold. Then calculate or restate the allowed interval in plain language. If “10.0 ± 0.2 mm” becomes anything other than 9.8 to 10.2 mm, the translation or formatting is wrong. This simple interval check catches many errors that ordinary proofreading misses.
Common failure modes
1. Changing ± into an approximate word
Plus/minus defines a range. “Approximately” is not necessarily equivalent.
2. Reversing ≤ and ≥
This changes the acceptance condition. Restate the condition in words during QA.
3. Dropping a minus sign
Asymmetric tolerances and negative values can become fundamentally wrong.
4. Removing trailing zeros
In technical contexts, decimal places may indicate precision or specification resolution.
5. Converting only the nominal value
When units change, tolerances and limits must change consistently too.
6. Confusing tolerance with uncertainty
Tolerance defines acceptable variation; uncertainty describes measurement knowledge. They serve different purposes.
7. Rounding too early
Premature rounding can shrink or enlarge a specification interval.
8. Proofreading only linguistically
A fluent sentence can still contain wrong mathematics. Recalculate the range.
Worked practice
Practice 1: Symmetric dimension
Situation: 12.0 ± 0.1 mm.
Reasoning: Keep notation exact and verify the interval 11.9–12.1 mm.
Practice 2: Asymmetric shaft tolerance
Situation: 20.00 +0.02/−0.01 mm.
Reasoning: Preserve separate deviations and check endpoints 19.99 and 20.02.
Practice 3: Minimum purity
Situation: Purity ≥ 99.5%.
Reasoning: Translate as at least 99.5%, never at most.
Practice 4: Temperature conversion
Situation: 20 ± 2 °C must be presented in another scale.
Reasoning: Convert the interval endpoints, then express the corresponding target-scale nominal/tolerance only if mathematically justified.
Practice 5: Calibration certificate
Situation: Expanded uncertainty U = 0.4, k = 2.
Reasoning: Preserve metrology terminology, symbol and coverage factor.
Practice 6: Significant zero
Situation: Result reported as 2.50 g.
Reasoning: Do not simplify to 2.5 g if precision is meaningful.
Practice 7: Approximate dimension
Situation: approximately 3 m.
Reasoning: Keep estimation language and do not invent a formal tolerance.
Practice 8: Pass criterion
Situation: Reject if error > 1%.
Reasoning: Check the inequality logically with 0.9%, 1.0% and 1.1% sample cases.
Standards, calculators and AI
Engineering standards, metrology guides, calibration procedures and discipline-specific specifications are the strongest terminology sources. Use the standard that governs the document rather than a generic glossary.
Calculators and simple interval arithmetic are excellent QA tools. Independently compute endpoints, unit conversions and percentage ranges instead of trusting visual similarity.
AI can explain notation but may normalize symbols or round values. Protect numbers and operators in prompts, then verify every technical relationship independently.
How this fits the wider eduKate translation system
Tolerance translation sits where language meets mathematics and measurement. The broader framework is developed in Master Art of Translation | The Complete System for Moving Meaning Between Languages. Vocabulary depth connects to the Vocabulary Learning Hub, while comparison, scope and sentence logic connect to How English Works. The extra discipline here is numerical: translated prose must leave the permitted range and uncertainty exactly where the source placed them.
FAQ
Is tolerance the same as uncertainty?
No. Tolerance usually defines acceptable variation, while uncertainty describes limits on measurement knowledge.
Can ± be translated as approximately?
Not safely. Plus/minus defines a specific range; approximate language can be looser.
Should I keep trailing zeros?
Yes when they communicate technical precision or specification resolution.
How do I translate asymmetric tolerances?
Preserve the separate upper and lower deviations exactly.
Can I convert units?
Only when the project requires it. Convert nominal value, limits and tolerance consistently and verify rounding.
How can I check an inequality?
Restate it in words and test sample values below, equal to and above the threshold.
What about confidence intervals?
Keep the confidence level and endpoints and use appropriate statistical terminology.
Should standard engineering symbols be translated?
Usually no; preserve standard notation and translate the surrounding explanation.
Can AI handle tolerances reliably?
It can assist, but operators, signs, units and arithmetic should be independently checked.
What is the simplest rule?
Treat the numerical relationship as protected logic and verify it mathematically after translation.
Final checklist
- Is this a tolerance, uncertainty, limit or approximation?
- Are all operators and signs unchanged?
- Are nominal value and units correct?
- Are asymmetric deviations attached to the right direction?
- Have decimal places and significant figures been preserved?
- If units changed, were limits and tolerances converted too?
- Can I restate the accepted interval correctly in words?
- Have I distinguished tolerance from measurement uncertainty?
- Did I avoid premature rounding?
- Would the same sample values pass and fail in both source and target versions?
Technical translation is trustworthy only when the same values pass and fail after translation as before it. Preserve operators, signs, units, precision and tolerance structure; convert only with explicit authority; and verify the result by calculation. When the mathematics survives intact, the translated prose can do its job without quietly changing the specification.