How best to use a 3D scanner to test floor flatness compliance.

Laser Scanning for floor flatness specification compliance.

There has been much debate recently about the use of Laser Scanning for checking compliance to floor surface regularity specifications. The debate has been on whether Terrestrial Laser Scanners (TLS) are accurate enough. Sure, they generate pretty pictures and give you a general overview of the floor levelness but are they accurate enough to determine compliance with flatness specifications that have maximum limits that are in the low millimetre range?

No matter what TLS equipment you use, the frequency of measurement, the averaging software developed and the calibration method to tie in each registration, there will be a level of uncertainty (inaccuracy). This is relatively easy to calculate with repeatability tests on a given representative sample of floor. Then using different operators’ different equipment and different environment, you can arrive at a reasonable understanding of a combined uncertainty value.

Specification limit adjustment.

With high quality TLS and thorough workflows, we are seeing this uncertainty value, at best, at around +/- 1mm. Surely this uncertainty value should then be removed from the specified limits to avoid missing false negatives. So, a +/- 3mm elevation limit will become a limit of +/- 2mm say. That is a tightening of the specification by a 1/3rd.

If we now, consider that a flatness specification is a rate in change of elevation difference over a given distance. This involves 3 separate elevation readings, each with a level of uncertainty of say +/- 1mm. What is the level of uncertainty of the rate in change of elevation difference now? +/- 2, +/-3 mm or something else? Again, this should be calculated and taken from the specification limits to determine compliance with a given specification.

False Positives.

A false positive is an error in binary classification in which a test result incorrectly indicates the presence of a condition that is not actually present, while a false negative is the opposite error where the test result incorrectly fails to indicate the presence of a condition when it is actually present.

What does this mean when testing floor flatness?

A false positive will highlight an error on the floor that does not exist. Some would say that this is not a problem. At least we are finding all the errors and this error can be validated by more accurate test methods to determine whether the error does exist. If it does not exist, leave it, and if it does exist, correct it. On a floor, this usually means corrective grinding. The problem here is that an expensive grinding crew might be standing around doing nothing whilst an equally expensive validation process is being carried out to find that most of the identified errors do not actually exist.

A false negative is a bigger problem as the test method has not picked up an actual error and is therefore left untouched. If the flatness requirement is critical and the actual error is later found in operation of the equipment running on the floor, remedial works are then both inconvenient to the end user and costly due to partial shut-down of said operation. In the meantime, the performance of the operation will be impaired, all of which results in a dissatisfied client.

Understanding how best to use TLS effectively to survey a floor to a given specification.

How do we use a TLS with confidence to check floor flatness against a given specification and at the same time deliver the pretty pictures?

First, we must establish a single x, y, z coordinate for a given grid size. Let us say a grid size of 50mm over the whole of the floor. Each point within the point cloud will be allocated to one of the 50mm x 50mm squares. The average z value will be calculated for that square and the centre point of that square will give the corresponding x – y coordinate. Using this set of generated values, we must determine the combined uncertainty of that value.

Establishing a Combined Uncertainty value:

Combined uncertainty is the square root of the linear sums of squared standard uncertainty components. How do we calculate this on a floor?

1.  Survey a reasonable floor area covering at least 4 registrations of the size intending to use on an actual survey. Using the test equipment, frequency of data collection and workflow one intends to use on an actual survey.
2. Survey the same test area at least 5 times with the same equipment and operator. Change operator and repeat. Chane equipment and repeat.
3. Calculate the combined uncertainty of the full set of data from all the surveys. The uncertainty value should be calculated from a full set of data of the whole test area and not just the validation points tying in the registrations.
4.  As the data from a TLS will be an average of a point cloud within a given area, a clear understanding of how that data set has been used to arrive at a single value must be explained. i.e. what averaging algorithms have be used?
5. Once the combined uncertainty of an individual z value has been established, the uncertainty of the rate in change of elevation difference must be determined.
6. This calculated combined uncertainty value should then be taken from the property limits of the specification.
7.  Does an expanded uncertainty need to be considered as part of the specification and if so what coverage factor?

Tightening the specified floor flatness property limits by the combined uncertainty value (expanded uncertainty if specified) will not eradicate false positives. On the contrary, the opposite will happen, and we will experience even more false positives, but all the errors will be amongst this set of data. If the errors are then validated with a more accurate test instrument, the original property limits, without the uncertainty component, can now be used for correction, less the combined uncertainty value of the more accurate test instrument.

By using this method, false negatives should eliminated or as close as.