Units
1. AI, Design Support and CAD Workflows in Mechanical Engineering 2. Material Selection and Ashby Approach 3. Strength and Static Calculation Support 4. Thermodynamics and Fluids Calculation Support 5. Finite Element (FEA) and Simulation Interpretation 6. Predictive Maintenance and Sensor Data 7. Production, Manufacturing Processes and Manufacturability (DFM) 8. Drawing, GD&T and Documentation 9. Standard and Norm Research (Emphasis on Verification) 10. Engineering Calculation Automation with Python 11. Security-Critical Affairs, Ethics, Privacy and Accountability
Unit 9 / 11

Standard and Norm Research (Emphasis on Verification)

Gains:

  • Ability to distinguish standard families and scopes such as TS, EN, ISO, DIN, ASME
  • Standard screening, summarization and compliance checklist creation with AI
  • Ability to apply the discipline of verifying the substances and values cited by AI from the official standard text

In mechanical engineering, no design stands in a vacuum; From pressure vessels to lifting equipment, from bolts to welded structures, everything is subject to standards and norms. A standard (a commonly accepted, documented set of technical rules for a product, process or method) ensures safety, compliance and quality and is often a legal requirement. Standard families such as TS (Turkish Standard), EN (European Norm), ISO (International Organization for Standardization), DIN (German standard) and ASME (American Society of Mechanical Engineers) carry the rules of different geographies and fields. Artificial intelligence (AI) saves time in standard screening, summarization, coverage comparison and compliance checklisting. But this is where the most serious risk of AI in engineering comes into play: AI can produce incorrect or completely fabricated (hallucinatory) standard numbers, item numbers, and limit values. A standard clause or value is never taken as basis for a design, calculation or report without being verified from the official and current standard text. This is the main theme of this unit: using AI as a starting point in standard research, but verifying each attribution from the primary source.

Standard Families and Scopes

Different standards families serve different purposes and geographies. Knowing which family covers your business is the first step:

family

Source / scope

Typical usage

TS / TS EN

Turkish Standards Institute; Turkish equivalent of ENs

Legal/commercial compliance in Türkiye

EN

Europe (CEN/CENELEC); basis for CE

EU market, CE mark

ISO

International; generally valid

Global design, quality (e.g. GD&T, inserts)

RELIGION

Germany; origin of many ISOs

Machine elements, detailed norms

ASME

USA; pressure vessel, boiler, GD&T (Y14.5)

North American projects, petrochemical

The same issue may resonate in more than one family; for example, a screw thread is defined in both ISO and DIN and is often compatible, but not always one-to-one. In a European project, EN/TS EN is taken as basis, in a US project, ASME is taken as basis. Choosing the wrong family will cause the product produced to be incompatible in the target market.

Tip: When asking the AI ​​for a standard, specify the target market and application, such as "A lifting hook to be sold in Türkiye." This leads to the correct standard family (TS EN). However, do not accept the number given by AI without confirming it from the official catalogue.

Hallucination Risk: The Most Critical Confirmation

Standard research is the area where AI hallucination is most dangerous. AI can generate a standard number ("EN 13445-9 clause 7.4.2") that appears real but does not exist; may refer to the wrong clause of an existing standard; or may present an old, outdated version as up-to-date. None of this is the AI's "malicious intent"; The language model generates possible text, it does not guarantee accuracy. The consequence is serious: a calculation based on an incorrect standard value will be rejected in the audit or create a safety risk in the field.

Attention: The only valid way to verify a standard clause issued by AI is to look at the official and current standard text (TSE, ISO, relevant official source). A forum comment, a second-hand blog, or the AI ​​itself saying "sure" is not verification. Standards also have version (year) and effective status; Make sure you are using the current edition.

Step by Step: Standard Research with AI

  1. Define the application and market. What do you produce, to which market?
  2. Have AI produce candidate standards. Possible families and numbers as a starting point.
  3. Summarize the scope. What does the standard cover and what does it not cover?
  4. Verify each attribution from the primary source. Number, version, item, value.
  5. Create a compliance checklist. Items that must be met.
  6. Competent review. The engineer makes and documents the suitability decision.

Candidate standard finding prompt

Role: Mechanical engineer experienced in standards and compliance. Application: A lifting hook with a capacity of 2 tons, to be used in Türkiye. Task: List the standard families and numbers that MAY be relevant for this product (TS EN, ISO etc.). Write the scope for each in one sentence.Rule: Every number and item you give could be a HALLUCINATION; Make it clear that I need to verify all from the official source. Mark if you are not sure.

Scope summary prompt

Summarize what you know about [standard number]: what product/process it covers, what main topics it covers, what it typically does not cover.Rule: DO NOT give the substance number and limit value, or if you do, mark it as "must be verified". My goal is to find a general direction, for the exact value I will go to the official text.

Verification plan prompt

Come up with a plan to verify the following standard references: [number and clauses given by AI]. For each reference: which official source (TSE, ISOkatalog) should I look at, how to check the version/year, how do I know if it is valid? Make a checklist chart.

Eligibility checklist prompt

A compliance checklist is drafted based on the requirements [of the standard I am verifying]. For each item: requirement, how to prove(calculation/test/document), responsible.Rule: I verified the item numbers; You just install the structure. State that the suitability decision belongs to the competent engineer.

Weak Prompt / Strong Prompt

Weak prompt:

Which standard should this hook comply with? Also give the item number.

AI can confidently make up a number and substance; If the engineer uses this without verifying it, it will be based on the wrong foundation.

Powerful prompt:

List the standard families and numbers that may be relevant for the 2 ton lifting hook to be used in Türkiye as a starting point, and summarize the scope of each in one sentence. Know that I will verify every number and item you give from the official source; Mark that they are not sure and do not produce substance/value.

The second prompt positions the AI ​​as a starting point, includes the hallucination warning, and explicitly leaves the burden of verification to the engineer.

Three Mini Cases (By Numbers)

Case 1 - Fabricated matter. AI says "EN 13445 clause 7.4.2, permissible stress factor 1.5" for a pressure vessel. The engineer looks at the official standard text: the relevant topic is in a different clause and the factor context is not what the AI ​​says. If he used the value as is, the calculation would be wrong. The correct item and value are taken from the official text and corrected. Lesson: the item number is verified no matter how realistic it looks.

Case 2 - Outdated version trap. AI refers to a revised edition of a welding standard years ago; Some acceptance criteria of that edition have changed in the current version. The engineer checks the current version from the TSE catalog and updates the calculation according to the current edition. Since the current version is requested in the audit, this control saves the project. Lesson: even if the number is correct, the version/year is validated separately.

Case 3 - Right start, right direction. An engineer asks AI for a machine in the Turkish market; AI recommends the TS EN ISO 12100 (machine safety risk assessment) family as a starting point. This direction turns out to be correct; The engineer confirms the relevant items from the official text and establishes the risk assessment document. AI saves time here by giving the right research direction, not the value. Lesson: AI's safest contribution is to show "where to look", not to give "exact value".

Common mistakes

  • Using the item/value without verifying: Accepting the number given by the AI without looking at the official text.
  • Not checking the version: Mistaking an obsolete edition for current.
  • Choosing the wrong family: Incompatibility such as using EN as the basis for the US market and ASME for the EU market.
  • Relying on second-hand source: Substituting forum/blog comment as primary standard.
  • Misunderstanding the scope: Relying on a subject that is not covered by the standard.
  • Leaving the suitability decision to AI: Making a "suitable" verdict without engineer approval.

In summary

  • In mechanical engineering, designs are subject to standards such as TS, EN, ISO, DIN, ASME and often legal obligations.
  • The right family of standards is selected based on the target market and application; Wrong family means incompatible product.
  • Standard research is the area where AI hallucination is most dangerous; Numbers, items and values ​​may be fictitious.
  • Each reference must be verified from the official and current standard text; version/year is also checked.
  • AI's safest contribution is to point in the right direction of research; The decision on suitability rests with the competent engineer.

Application task

Choose a concrete product (a lifting element, a pressure component or a machine safety component) and specify the target market. Have the AI ​​list standards families and numbers that may be relevant for this product as a starting point, outlining the scope of each. Then try to verify at least two references (standard number and clause if applicable) from an official/current source; Check if there is a number, what is the current version, does the item say what the AI ​​says. If you find hallucinations or old versions, make a note of them. Finally, draft a brief compliance checklist based on the standard you are verifying and write out why the compliance decision remains with the qualified engineer.

checklist

  • [ ] Target market and application defined; The correct standard family has been identified.
  • [ ] AI output was used only as a starting point and not considered the definitive source.
  • [ ] Each standard number and clause has been verified from the official/current text.
  • [ ] The version/year and validity status of the standard have been checked.
  • [ ] Second-hand sources (forum/blog) are not substituted for the primary standard.
  • [ ] Compliance decision and documentation is left to the competent engineer.
Previous Next