InnerWeb
Skip to main content

Machine Positions from an Old 2D Map: A Case Study

The integrator rebuilds machine positions from the 2D map of the old system and assumes that the approved building plan is more accurate. At the plant described, after using the plan, 44 of 45 markers fell outside the area of their line. Below: what went wrong and which checks to run before saving.

44 of 45
markers outside their line area after the wrong source
Source: migration documentation
0.01 m
maximum conversion error on 64 control points
Source: migration documentation

Starting point: two sources of truth about position

When moving a plant from another system, there are usually two sources of machine positions. The first is the 2D map of the old system, on which someone once placed points by hand. The second is the building plan approved for migration, prepared to reproduce the structure of the facilities. The natural reflex says the plan is newer and official, so it wins.

In this case the reflex failed. The points on the old map turned out to be the real machine positions, while the plan approved for migration was approximate. The difference was not visible on a single marker and only showed up when the whole set was compared.

What happened: 44 of 45 markers outside the line area

After saving machine positions according to the plan, at one plant 44 of 45 markers fell outside the area of their line. Practically the whole layer of positions was unusable: a machine assigned to a line lay on the map next to it, so risks and permits assigned to a location would point to the wrong place.

The same type of error has a second face when buildings from the plan are treated as view records. In the migration described, this erased the only view with a georeferenced calibration, the editor showed no calibration and the orthophoto disappeared. The plant structure has five levels: plot, view (department), building, line, location. A view is not a building, and mixing these levels costs the most.

  • source of points: plan instead of the old system map,
  • effect: 44 of 45 markers outside the line area,
  • additional risk: buildings saved as views erase the calibration.

Markers outside their line area

Markers outside their line areaBefore and after chart: 44 markers outside the line with the plan, 0 after the source correction.01020304050markersApproved plan (Outside line): 44 markers44After source correction (Outside line): 0 markers0Outside lineApproved plan (Inside line): 1 markers1After source correction (Inside line): 45 markers45Inside line
  • Approved plan
  • After source correction
Source: Migration documentation: 45 markers; value after correction equals the expected consistency gate (0)

What to check before saving

Before the script saves anything, the team should answer a few control questions. In this way the error shows up in a trial run, not in production.

Which hall a row belongs to

Location coordinates in old data are sometimes recorded in the coordinate system of a different hall that has no georeferencing. Before use, we check which hall a row belongs to. A row from a hall coordinate system without georeferencing that is converted without this knowledge ends up in the wrong place.

One source for converting meters to coordinates

Converting local meters to geographic coordinates has one source of truth. The migration script must be a 1:1 port of this function and pass verification on control points. In the migration described, the measured maximum error was 0.01 m on 64 points, which is a value to reproduce, not to accept on trust.

Compliance with the line outline

After saving, we count how many markers fall outside the area of their line. The expected value is zero. The figure 44 of 45 is a result that should stop saving at the gate, not end up in a report after the fact.

How to secure a repeat

Structure migration is repeatable: each run builds the structure from scratch. Saving runs in one transaction with a trial mode, that is with rollback, and is preceded by a copy of the structure to an SQL file. In this way a mistake in the source of points costs one run, not manual restoration of data.

For old system maps it is best to put both sources side by side on one board before the first save. Markers from the old map and outlines from the plan are overlaid, and discrepancies are judged by people who know the hall. If the plan is approximate and the 2D map is true, then the map determines machine positions.

Conclusions for the integrator

First: an approved plan is not proof of accuracy. Second: view, building and line are different levels, and an error at a higher level destroys the calibration of the whole area. Third: a script without a trial and gates is a hypothesis, not a migration.

InnerWeb works as an independent system supplementing the client's systems, so the old system can keep working during such a trial. This allows the trial to be repeated as many times as needed before the structure reaches the target environment.

Video

Frequently asked questions

Is the approved migration plan always more accurate than the old map?

No. In the case described, the plan was approximate, and the points on the old 2D map matched the real machine positions.

Can a migration save be rolled back?

Yes, if the save runs in one transaction with a trial mode and is preceded by a copy of the structure to an SQL file.

What does the 0.01 m error on 64 points mean?

It is the measured maximum error of converting local meters to geographic coordinates, checked on control points.

Moving machine positions from another system?

We will compare the old sources with the building plan on a test environment before anything reaches production. Book a talk.

Let's talk