The Next Person To See Your Product Is Your Customer.....

Fabric Production Image

A typical narrow fabric factory will produce upwards of 1 million metres of fabric a week in diverse qualities and styles. This fabric mostly manufactured on needle looms and crochet knitting machine will have on average 1 defect every 200 metres and this figure is for the premier quality manufacturers. Manufacturing defects are therefore inevitable and it is Fibrescan's mission to detect these defects before your customer does.

First some background on textile inspection.

Reliably detecting textile defects is a difficult task and can prove very expensive if you get it wrong or chose the wrong equipment supplier. Unlike other manufactured goods textiles have a very wide probability distribution of what is tolerated as acceptable fabric; this is the basis to the problems encountered when trying to automate its inspection.                                                                                                                                                                        

Manual inspection is the traditional and fall-back position for inspecting textiles and humans are very good at distinguishing the subtleties that distinguish defective fabric from good fabric. For example, when a machine inspects stretched fabric as opposed to relaxed fabric, it sees two totally different fabrics so it will tend to reject one or the other; however it is an easy judgement for a human inspector to tell that are both examples of good fabric.

"For surface defects in many cases it is not a digital threshold type of decision from a single variable, but rather many analogue inputs with many dynamically variable criteria that determines acceptance. A machine vision system may fall between these extremes." N.Holmes - SME forum 2006.

This sort of problem can lead to one of the biggest headaches when an automated inspection system is introduced - a massive jump in product rejection rates. Parts considered good by humans are now rejected by machines.

Unfortunately human inspectors are both expensive and unreliable (follow this "Find the Fs" link and test yourself)  - they also tire easily making their judgements inconsistent. Studies have shown that a good operator will have an inspection efficiency of just 60 to 70%.

These are the themes that inform the design in every Fibrescan inspection product.
We call it the 95% rule - 95% of your products inspected to 95% inspection efficiency.

We will make our inspection algorithms robust enough to give you these inspection levels.
We will make our systems simple to use.
We will make our systems very very fast.

We will never forget that the next person to see your product will be your customer.

Charles Schofield
Fibrescan Ltd