IMO
Validation should be conducted on a routine basis to asses the ability of the control to control the identified hazard to the acceptable level. Verification ensures that the validated controls continue to operate as intended. Validation determines what is possible verification ensures it is still possible.
You need to determine the number of samples to be run during validation. This is where it gets a bit tricky. Is 1 pass enough? should it be 30? How about 300 or 10,000. It all depends on your level of confidence and reliability requirements. I often site the Bayes Success Run Theorem as a good reference.
In regards to metal detectors a few things should be validated. Detection capability, operation stability and capability.
Typically I would start with operational stability. Can the system operate at maximum sensitivity with no adverse effect from the environment (interference, ground loops metal free zone violations) or from the inspected product/item.
Products/items that are conductive in nature (high moisture products, metalized film, high temperature) will often require a reduction of sensitivity in order for them to successfully pass through the detector when un-contaminated. Focus at this stage to be setting the detector to successfully inspect contamination free product by passing X# of samples to ensure no false detection. You may need to run a range of product conditions if the product varies during production (moisture, temperature, etc)
Once the product setting has been established and validated we can move on to the detection capability. Metal samples should be passed with product and positioned to travel as close as possible to the absolute center of the detector opening, this is the weakest point for detection. Pass X# of each contamination type to determine the detection capability. This may also require multiple validations for each contamination type as well as validating a range of settings. i.e. Lowest sensitivity to detect samples and highest sensitivity to have an acceptable false reject rate.
Other validation considerations would be for the system itself. items such as reject operation, air fail sensors, bin full sensors and reject confirmation devices should also be validated and verified for proper operation and action.
Hope this helps
RR