Nelson Bros Engineering Scanner / Optimizers


Rod Nelson has been programming and starting trimmer optimizers since his days at Saab Systems, back in 1983.  In the early days, the scanners were not too good, so the optimizer software was full of kludge code that would make the best possible decisions with questionable scan data.  In the words of Wenna Neilsson,  "we don't see so good, but we trim pretty good".

Scanners are a bunch better, today, so less software is needed to correct for bad scan data.


Nearly all sawmill trimmer scanners are Transverse Scanners.  The boards travel laterally thru a scanner that spans the length and thickness of the board.  Trimmer optimizer scanners are starting to appear in planermills.  These can be either Transverse Scanners or Lineal Scanners.  As you probably know by now, NBE is an advocate of Lineal Scanners, so we think a Transverse Scanner in a planermill is a bad idea.  But, since we have not done one in a planermill, you can safely ignore our opinion.  We have also dreamed of lineal scanners for sawmill trimmers, but no mechanical vendor has every shared our dream.

The most common scan heads have multiple lasers being viewed by a line image camera that is synchronized to the scan chain by an encoder.  The lasers are one spaced on 3” or 1” intervals, thus the scan head measures top and bottom contours on those intervals.  Most scanner configurations with the multipoint laser scanners include a thru beam light-curtain with beams every 1/4" or 1/2" to improve width and length resolution. 

The problem with most transverse scanners are:

        They get too much data on the top and bottom face and too little data on the sides.   It is backwards, you want more data on the sides than on the top and bottom face.

        They are complicated with multiple scanner types, requiring lots of spare parts.

        The bottom scanner requires a Klap-Trap of plenums and blowers to keep heads clean.

        They are expensive.

        They are only useful for transverse scanning.  You need another type for lineal scanning.

In 2005, NBE gave up on multipoint, multi-technology transverse scanners and started using the JS-20 heads.  Yes the same heads we use on lineal and snapshot scanners.  Now one scanner can truly be used for all NBE sawmill and planermill applications.  The results were so good we had to brag.  Click the picture below to see our announcement.


Things to consider when selecting the scanner:



To keep from being redundant, we will omit the general BS on optimization that was in the edger optimizer description.  The following are features of the NBE trimmer optimizer.

This meets most customer requirements, but there have been requests for things that the optimizer does not do yet:

Like any vendor, we can do anything, but you should know what exists and what is new.  Then, keep in mind that new tends to be a bad word.

The most important display is the solution screen shown below.  It gives a graphical display of the solution in real-time, plus some production and decision text.   This screen is displayed on the optimizer, supervisor and any monitor computers.  The optimizer's monitor is normally positioned within the operator's view.



The NBE's Supervisor's Solution Screen


NBE's Opinion.   You should know it by now, but in case you did not, NBE has lots of opinions.  Some of them are good.

NBE always stresses:  simplicity, cost and reliability.  Vendors that want to charge a premium for their product tend to stress:  sophistication, data density and recovery percentages.  We all know that recovery, initial cost, maintenance cost and uptime are what determine a project's Return On Investment (ROI), just don't get too excited about the vendor's recovery guarantee.

The normal recovery guarantee is 1% overtrim and 1% undertrim.  This is usually possible with a variety of scanning technologies and optimizers.  I once worked for a vendor that wanted to market our system as having a 1/2% overtrim guarantee.  The final recovery numbers coming back from our startups supported that number.  The marketing guru's got out their calculators and determined that 1/2% could make mills up to 1/2 million dollars per year.  This prompted the sales department to raise the system price $50k, and head out for lunch.  That vendor does not exist anymore.

The recovery numbers are derived from comparing an assessment of what got produced, to an assessment of what should have been produced.  Humans do both these assessments, so they are not exact.  I would be surprised if two graders could come within 1% of each other.   Some vendors are convinced that a good personally during recovery testing is worth 1%.  In one of my previous jobs, it became common knowledge, that one in our service staff could always get a 1% better score.  So, when the techies could only get within 2%, we would bring in the "the master".  That vendor also does not exist anymore.

The problem with a guarantee that gets too close to the uncertainty of human grading is that you can spend all your time trying to get the last 1/2%, but 2 months later, the recovery number has dropped below the guarantee to some value that the mill will maintain.

Mills tend to be stricter during initial recovery tests.  They want to get the best possible recovery and learn the idiosyncrasies of the system.  After they are convinced that the guarantee is going to be met or that the guarantee will never be met, they tend to be less strict.  One reason for becoming less strict is the desire to show a good recovery, thus good payback, thus justifying their decision to spend too much money for an optimizer.  The vendor that wanted to use the recovery numbers from the less strict testing forgot that you first have to get past the initial, stricter testing.   Oops!

Another observation is that boards must be randomly selected for tests.  Selecting boards for certain characteristics is OK for debugging, but not for assessing the machine center's recovery.  Remember, the optimizer makes a decision on every board, not just the waney boards.  Tests with waney boards can sometimes be easier to pass than tests with square edged boards.  If the optimizer misses a marginal face, edge or length wane decision, no one gets excited.  When the optimizer cuts a square edged 16' 2x12 back to 14', everyone gets excited.

At NBE, we believe mills must continually assess machine center recovery.  A 10 board test every day is much better than a 100 board test every month and infinitely better than believing the vendor's assessment or never testing.  All NBE optimizers have tools for saving board data, replaying boards and assessing optimization decisions.

We cannot stress, continual testing enough.  You always learn something about the system in a 10 board test!

One last note before I end this diatribe on recovery. Why do all trimmer optimizers look like they are screwing up?  Assuming that your observations were perfect, a 1% overtrim could result in your detection of a 2 foot overtrim on a 20 foot board (10% overtrim mistake) every 10 boards.  At 90 lugs per minute, you could detect 9 mistakes per minute.  Also a 1% undertrim could result in another 9 mistakes per minute.  For those who are not perfect, you will not detect all of these mistakes, but will falsely detect other mistakes.    Bottom line; only use your observations as motivation to do continual board tests.

Many vendors would argue with my views on recovery guarantees.  Most of them are in sales or will be soon.

If you want some more of NBEs options on optimization, look at  Edger Optimizer.

Conclusion, buy NBE.



NBE Home Page

NBE Article in Timber Processing on Lineal versus Transverse Scanning

Customer List


Email NBE - (360)951-2737  29-Dec-05