Cutting Out Defects

As I mentioned a few days ago, I made a presentation at the Potato Processing and Storage Convention in Warsaw last week. The presentation was on a "new" approach to defect control for potato strips, where all the product is sent to ADRs first rather than sorters. Over the coming weeks, I plan to discuss details of that presentation.

To begin, however, let's tell the short story:

It starts with the idea that, in potato strip applications as well as most others, sorters are not perfect. Neither are ADRs. That means that they neither removal ALL the defects that come in, nor do they remove the defects only: they also remove some amount of good product. There has been, and is, of course, an ongoing effort to improve both sides of that: defect removal rates and false rejection rates. But it is my impression that such efforts are nearing an asymptote: that theoretical limit beyond which further improvement is not possible without radical change of the concept.

If you haven't figured it out yet, this blog is a bit about radical change....

Most of the reason that sorters and ADRs do not remove 100% of all defects is that they image product while it is lying on a belt. It is not possible to image the face that is against the belt, so defects that are seen only from that side are missed. About 15-20% of all defects in any given load of strips falls in this category, so the highest possible defect removal is about 80-85% for on-belt systems. Because there are some other, more subtle causes for less than 100% defect removal, I tend to use 80% average removal as the standard number.

Most processors of strips send a flow of defects from the reject stream of the sorter to the accept stream of an ADR, so that the ADR can cut cubes of the defects and retain the rest of the strip for recovery. I will write more in coming weeks on those strip processors who do not use ADR. The ADR functions much like a sorter, removing about 80-85% of its incoming flow.

Problem is, the ADR only sees 80% of the total line incoming defects, so it actually removes only 64% (80% of 80%) of the incoming defects, maximum. See the image from my presentation:

64% doesn't seem like much, does it? And folks try to use various means to get better performance, such as double sorting and recirc (we will also discuss these later). And they usually get somewhat better defect removal, but at significant cost in yield.

And, to be sure, the above model tends to produce A-grade product if all parts stay in tune and incoming defect levels stay below 20%. When (not "if", for most folks) either of those get outside of spec, big problems can occur. Either the product goes out of grade, or the ADR is overflowed with product (with attendant yield and product quality hits, not to mention machine damage/downtime issues), or the line rate must slow down (cuts productivity). So the "if" that started this paragraph is a huge "if".

So some folks that are able to keep their machines in tune, and acquire good potatoes through storage season do just fine with the above plan, thank you! But there are others, and I must say that the number is growing, who do not enjoy such benefits. They process in regions where 20% incoming defect would be considered extremely clean, where 30% and up is the norm. They work in areas where skilled labor to keep systems in tune is difficult and expensive to acquire, and so often they do without- and suffer the consequences: poor product quality, poor yield or both.

So that is the problem... a system that is simply not capable to meet the needs of many new lines in new geographies, even though it is the historical standard of the industry.

Where do we go from here?

Read this blog next week!

Tim

Stumble It!

Bookmark on del.icio.us