Visual Inspection Effectiveness

[optomist1]

I'll attempt to ferret out the study supporting this...many years ago. Until then, the often quoted Visual Inspection effectiveness = on a good day an inspector will catch 80% of defects presented

 

[Jim Wynne]

I'll attempt to ferret out the study supporting this...many years ago. Until then, the often quoted Visual Inspection effectiveness = on a good day an inspector will catch 80% of defects presented

There's absolutely no way to apply a number like this that will take into account all situations. There are too many variables, including the number of things to be inspected, the nature of the defects, the abilities of inspectors, the conditions, such as lighting, where inspection is done, and others.

 

[Randy]

It's a subjective goat rope unless you've a method to calibrate the human eye, which can be done to certain degree such as in EPA Method 9.
Method 9 - Visual Opacity | US EPA

 

[Miner]

The linked Sandia Report covers this topic very well.

 

[optomist1]

and to Jim's point above the "study" I am referring to and searching for was pretty specific, i.e. assembly of airborne defense electronics. Specifics and details matter...as some of these often quoted figures can be anecdotal in nature

 

[Djoss]

During some Internal training, I sometime explain the weakness of visual detection, by simple Mathematic example. (In order to reinforce the need to use error proof systems.)

Let's assume for the exercise, that a control visual, operator will catch around 70% of defective part (Still a 30% of defective part left).
To reach a 99,7% of detection probability, we will need to put 5 Visual Control (5 differents operators controling the same control point).

To make this calculation, I'm using the Probability formulae :
P(rejects being identified by at least one inspector) = 1 – (1- e)^n
where e = inspection effectiveness (85% in the example)
n = number of 100% inspections.

Let's comment that Juran also define the "standard" accuracy of Visual control about 85%. (I know that this value could be very variable according to all specific conditions )

Have a look to this good blog, that's explain it better.
Search for the Blog of Harish with 100-visual-inspection-being-human for more info.

Conclusion: If you can: Try to find an automatic error proof system, that's the only way according to me to avoid problems...
KR
Djoss

 

[Jim Wynne]

During some Internal training, I sometime explain the weakness of visual detection, by simple Mathematic example. (In order to reinforce the need to use error proof systems.)

Let's assume for the exercise, that a control visual, operator will catch around 70% of defective part (Still a 30% of defective part left).
To reach a 99,7% of detection probability, we will need to put 5 Visual Control (5 differents operators controling the same control point).

To make this calculation, I'm using the Probability formulae :
P(rejects being identified by at least one inspector) = 1 – (1- e)^n
where e = inspection effectiveness (85% in the example)
n = number of 100% inspections.

The problem with this is that it assumes five apparently cloned operators. You won't find five operators who all have exactly the same skills, visual acuity and experience. Visual inspection should be used only when the risks and variables of each situation are understood and accounted for.