MIQI logoMIQI.
Inspection Technology

Food Metal Detector Testing, Sensitivity and Validation: A Line-Side Guide

Published · By Engineer Cai, Guangdong Miqi M&E Technology Co., Ltd.

A meaningful food metal detector test uses certified ferrous, non-ferrous and stainless-steel test pieces in the actual product, at the least favourable position, and confirms both detection and rejection. This guide separates routine verification from validation and explains product effect.

Food Metal Detector Testing, Sensitivity and Validation: A Line-Side Guide — infographic
Figure: Food Metal Detector Testing, Sensitivity and Validation: A Line-Side Guide (MIQI original)

A food metal detector is not proven by passing a bare test wand through an empty aperture. The challenge must represent the production risk: the correct metal type and sphere size, inside or attached to the real product, at the most difficult position, at normal speed, followed by confirmation that the reject system removed the right pack.

Ferrous, non-ferrous and stainless-steel test pieces

Use controlled test pieces whose material and size are identifiable and traceable under the plant system. Ferrous, non-ferrous and stainless steel respond differently. Austenitic stainless steel is commonly the most difficult of the three for a balanced-coil conveyor detector, but the real hierarchy can change with product effect, frequency, aperture and orientation.

Product effect can be larger than the contaminant signal

Wet, salty, acidic, warm or conductive products can disturb the electromagnetic field and imitate a metal signal. Recipe setup teaches the detector the normal product response so a contaminant can be distinguished from that background. Test the full range of product temperatures, formulations, pack orientations and belt loading rather than tuning on one ideal sample.

Validation, verification and routine monitoring

Do not use the three terms interchangeably
ActivityQuestion answeredTypical evidence
ValidationCan this system and procedure control the defined hazard?Study design, challenge selection, worst-case conditions, documented acceptance
VerificationIs the validated system still working as intended?Scheduled challenge results, reject confirmation, record review
MonitoringIs the process under control at this moment?Operator checks, alarms, observations and production records

A repeatable production test sequence

  • Confirm the correct product recipe, belt speed and reject bin condition.
  • Place each test piece in or on a representative product using the approved method.
  • Challenge leading, centre and trailing positions when the validation identifies positional risk.
  • Run ferrous, non-ferrous and stainless-steel challenges separately.
  • Confirm detection, physical rejection, reject confirmation and alarm behaviour.
  • Record product, recipe, test piece ID, result, operator, time and corrective action.
  • If a test fails, control affected product according to the plant procedure before restarting.

How often should testing be performed?

Testing frequency belongs to the facility hazard analysis, validation, customer requirements and quality system. It may include start-up, product change, defined production intervals, shift change and end-of-run checks, plus any restart after a fault. A machine supplier should not invent a universal interval for every food and jurisdiction.

Does FDA set metal detector critical limits?

Searches for FDA metal detection standards or FDA metal detection critical limits often assume one universal sphere size. FDA guidance discusses metal-inclusion hazards and control strategy, but the facility must establish and validate limits appropriate to its product, process and hazard. A metal detector test pieces standard, customer code or audit scheme may inform the procedure, yet it does not replace product-specific validation. The linked FDA sources should be read directly rather than relying on an unsourced metal detector in food industry PDF.

HACCP is a management system, not a machine certificate. A detector may support a CCP or another control measure, but that designation comes from the facility hazard analysis and documented plan.

Continue your equipment evaluation

Food metal detector models — conveyor, free-fall, foil and combination configurations

Metal detector vs X-ray inspection — choose by contaminant and package

Metal detection: CCP or PRP? — place the device in the plant food-safety plan

Technical references

  1. FDA Fish and Fishery Products Hazards and Controls — Chapter 20: Metal Inclusion
  2. FDA HACCP Principles & Application Guidelines
  3. METTLER TOLEDO — Product Effect in Metal Detection

Related equipment

By Engineer CaiEngineer Cai, MIQI (Guangdong Miqi M&E Technology Co., Ltd.). Talk to us about your line: +86 152 1890 9599 · 897874196@qq.com

Frequently asked questions

Where should a metal detector test piece be placed?+

Use the least favourable validated position. For many conveyor apertures this includes the centre of the opening and relevant leading or trailing product positions, but the correct method must come from the installation validation.

Why is stainless steel harder to detect?+

Many austenitic stainless steels have lower magnetic permeability and conductivity than ferrous metal. Product effect, aperture size, frequency, sphere size and orientation also influence the result.

What is the difference between metal detector validation and verification?+

Validation establishes that the chosen equipment and procedure can control the defined hazard under worst-case conditions. Verification checks routinely that the validated system continues to work.

Can sensitivity be copied from another factory?+

No. Published sphere sizes are only a starting point because product conductivity, aperture, packaging, speed, environment and reject design change achievable performance. Test the real product on the proposed machine.

Further reading

Inspection Technology

Metal Detector vs X-Ray Inspection: Which Food Inspection System Fits?

Metal detection is highly effective for metallic contaminants, while X-ray inspection responds to density contrast and may find glass, stone, ceramic and some dense plastics as well as metal. Packaging, product effect, contaminant risk, product thickness and required secondary checks decide the better technology.

Inspection Technology

What X-Ray Inspection CANNOT Detect: An Honest Limitations Guide

Almost every X-ray inspection page online tells you what the machine finds — metal, glass, stone, bone, dense plastic. Very few tell you what it misses. X-ray imaging works on density contrast: the beam is absorbed differently by different materials, and the detector renders that difference as grey. When a contaminant absorbs roughly as much radiation as the food around it, there is no contrast, and no contrast means no image — regardless of software, AI, or price tag. This guide walks through the physics behind that limit, the contaminant families it affects (hair, paper, low-density plastics, cartilage, string, wood), the product-side conditions that make a detectable object undetectable, and a practical checklist for deciding when X-ray is the right tool, when a metal detector is the better answer, and when you need both. Written by an engineer at a factory that builds all three — which is exactly why we can afford to tell you when X-ray will not save you.

Inspection Technology

How to Validate a Vendor's AI Inspection Claims: The Questions Nobody Wants You to Ask

Almost every accuracy number attached to "AI-powered" food X-ray inspection today is a vendor claim with no independent verification behind it. When you trace the most widely circulated figures back to their source, you land on equipment-maker and SaaS marketing blogs — not peer-reviewed studies, not third-party test reports. Searches aimed specifically at peer-reviewed validation of these numbers return vendor technical documents instead. And no vendor we found discloses the three things that would make an accuracy number meaningful: the test method, the sample size, and the confidence interval. This article is not an argument that AI inspection doesn't work. It is a practical guide to telling a real capability apart from a marketing sentence. It explains what the peer-reviewed literature actually says the hard problem is (training-data annotation, not model architecture), why a demo on a vendor's samples proves almost nothing about your line, and gives you a printable list of questions to put in front of any supplier — including MIQI. Written by Engineer Cai for engineers and QA managers who have to sign off on the purchase.

Ship us your product — free on-machine accuracy test with video proof

Free sample testing · OEM & custom design · 2-year warranty · Worldwide shipping