Do not overwrite the active configuration when the number, order, length, orientation, or segment reference changes. Create a new version with an exact effective start, close the previous interval, and keep the periods disjoint. A historical result must stay tied to the configuration that applied at the time of measurement. Versioning does not mean automatic rollback.
In brief
- Configuration is part of the measurement model, not just a device description.
- Version validity periods must not overlap or leave an unintended gap.
- Editing the active arrangement should create a new draft version, not change history.
- Activation must be preceded by a check of order, length, axes, references, and the time of the physical change.
- Returning to earlier parameters requires an explicit new decision; we do not promise a full rollback function.
Why configuration changes the result
An automatic chain inclinometer does not send a ready-made truth about the profile. The result comes from combining the readings with the configuration:
- the order of segments from the base upward,
- the length of each segment,
- axis assignment and sign convention,
- angle unit,
- reference measurement,
- the point treated as a fixed anchor,
- the summation rule,
- the time from which the arrangement is physically valid.
Changing the length from 0.5 m to 1.0 m doubles the contribution of the angle to displacement. Swapping the order shifts a local change to a different profile level. A new reference changes all differences. Reversing an axis can change the sign. These are not label corrections. They are changes to the calculation model.
Inclinometer configuration version means an immutable set of parameters and references valid within a defined time interval.
If the application stores only the "current state", opening a measurement from last month later may calculate it with the lengths installed today. The chart will look consistent, but it will no longer reflect the physical arrangement of that time.
The key rule: measurement time selects the version
Assume two half-open intervals:
- version 1: [1 April 08:00, 14 May 13:20),
- version 2: [14 May 13:20, no end date).
The notation [from, to) means the start belongs to the version, while the end does not. A measurement at 14 May 13:19:59 belongs to version 1. A measurement exactly at 13:20:00 belongs to version 2. This way, no boundary instant is assigned to both versions at once.
The start of a version should correspond to the actual physical change or to a consciously defined new state, not to the moment when the administrator had time to click "activate". If a segment was replaced at 13:20 and the configuration was entered at 16:00, the effective start should still reflect 13:20, provided the time of change is reliably documented.
Disjointness means:
- no overlap between intervals,
- an unambiguous version choice for the measurement time,
- explicit handling of any gap,
- preservation of the closed version,
- prohibition of editing a historical version's parameters.
Overwriting versus a new version
| Criterion | Overwriting the active configuration | New version |
|---|---|---|
| Meaning of old data | may change over time | remains tied to the old version |
| Audit | sees the change, but does not always reproduce the result | shows the full set before and after |
| Moment of physical change | easy to lose | becomes the start of the interval |
| Period comparison | ambiguous | explicit |
| Historical alarms | may receive a new context | preserve the context of the period |
| Error correction | risk of another overwrite | new, approved version |
| Report reproduction | difficult or impossible | possible with version and input |
| Accountability | decision hidden in the edit | author, reason and time are recorded |
A log entry such as "length field changed from 0.5 to 0.6" is better than no log, but it is not enough if the system later calculates all dates using 0.6. Change audit and versioned validity solve different problems: the first tells you what was done, the second defines which model should be used for a specific measurement.
Scenario: replacing a segment in an operating string
Illustrative example. A vertical string of eight segments is operating in a wall. Version 1 contains eight 0.5 m sections, axis A oriented toward the excavation, and a reference from the start of works. On 14 May at 13:20 UTC, the crew replaces the damaged fourth segment. The new element is 0.6 m long, and the work changes the geometry of that section and its reference state.
Incorrect approach: editing one field
The administrator changes the length of the fourth segment from 0.5 to 0.6 m in the active record and sets a new zero value. Result:
- results before 14 May are calculated as if the new segment had existed from the start,
- the profile sum above segment 4 changes across the entire history,
- the step caused by the physical replacement can be hidden by the new zero,
- the acceptance report from before the change may not be reproducible,
- it becomes hard to distinguish real movement from a configuration consequence.
Correct approach: version 2
The team creates draft version 2, leaving version 1 unchanged. In the new draft version:
- it enters length 0.6 m for the replaced segment,
- it confirms the order of all eight elements,
- it records the axis orientation after installation,
- it selects the reference from existing correct data after stabilisation,
- it adds a note about the replacement and the time of the physical change,
- it tests the A, B and resultant profile,
- it approves the effective start at 14 May 13:20 UTC.
Activation closes version 1 just before that moment and starts version 2 from the boundary instant. Earlier measurements keep the old model. Later ones use the new one.
If there is a period between the replacement and obtaining a reliable reference that cannot be interpreted, do not mask it with interpolation. You can keep the raw data and explicitly mark the lack of an approved result for that phase.
What should belong to a version
The minimum scope depends on the solution, but for a vertical string it usually includes:
| Group | Fields requiring versioning |
|---|---|
| Topology | number, order and identification of segments |
| Geometry | length of each segment, counting direction |
| Orientation | A/B axis mapping, signs, angle unit |
| Reference | reference values and the source measurement time |
| Anchor | base point and the assumption of its stability |
| Calculations | equation or calculation profile |
| Validity | start, end, time zone |
| Decision | author, approver, reason and note |
A parameter that affects the result should not remain outside the version simply because it looks like metadata in the interface. Length, axis sign and reference are calculation elements.
Three version states: draft, active and withdrawn
Draft version
A draft version is a candidate for activation. Topology, lengths, orientation and references can be corrected in it, and preview and tests can be performed. It should not affect published results until it is approved.
A draft version can be rejected if it fails review. Deleting an inactive draft is not the same as erasing historical configuration.
Active version
The active version is used for measurements within its interval. After activation, its meaning should remain unchanged. A correction leads to a new draft version, even if only one parameter changes.
At any given time, only one version may apply to the same string. Overlapping intervals would create two different results for the same measurement.
Withdrawn version
A withdrawn version has a closed validity period. It is not "old and unnecessary". It remains necessary to reproduce historical profiles, reports, alarms and decisions.
Withdrawn status does not mean the entire past situation can be restored with one click. The physical string, references and newer data may have changed. If the parameters of the old version are to apply again, the team should create a new version with a new period and approve it against the current state.
Activation is an engineering operation, not an administrative one
Before activation, four types of consistency are needed:
- Physical - the configuration matches what is actually installed.
- Temporal - the start of the version matches the moment of change.
- Computational - lengths, axes and references lead to the correct profile.
- Organisational - the right person approved the impact on reports and alarms.
It is worth previewing the profile on data from just before and just after the boundary. A sudden jump may be the expected result of a replacement, a wrong base measurement, or a sign that the order has been mixed up. Activation should not happen blindly just because the form is complete.
Checklist for activating a new version
- [ ] The reason for the change is described.
- [ ] The exact time of the physical change is recorded in UTC.
- [ ] The number of segments matches the installation.
- [ ] The order is established from the fixed base upward.
- [ ] The length of each segment has a source.
- [ ] A/B axes and signs are confirmed on the plan.
- [ ] The angle unit is unambiguous.
- [ ] The anchor point and its stability are justified.
- [ ] References come from correct existing measurements.
- [ ] It has been determined whether a common reference instant is needed.
- [ ] The draft version does not change published results.
- [ ] Sample A, B and resultant were recalculated.
- [ ] The profile just before and just after the boundary was compared.
- [ ] Every discontinuity was explained.
- [ ] The new version period does not overlap the previous one.
- [ ] The previous version is retained.
- [ ] The impact on alarms and reports was assessed.
- [ ] The responsible person approved the activation.
- [ ] A separate decision was defined for any history recalculation.
How to handle a mistake after activation
First determine whether the error concerns the description, the calculation configuration, or the physical installation. Do not edit a historical version in place. Create a corrected draft version and choose the time from which the correct configuration actually applies.
If the incorrect version generated results in a defined period, an explicit decision is needed:
- keep them with an error flag,
- recalculate that period in a controlled way,
- exclude them from the report,
- reassess alarms and communication.
Creating a new version does not automatically recalculate the entire history. Recalculation should be a separate run with a defined scope, test and atomic publication, so that recipients do not see a mix of results. A broader procedure is described in the article safe history recalculation.
We do not promise a full rollback mechanism for the configuration and all its effects. Versioning allows the context to be reproduced, but reversing a decision may require a new version, recalculation and result review.
How it works in Inclify
Inclify stores automatic chain inclinometer configurations as draft, active and withdrawn versions. Editing an active profile leads to a new version. Validity intervals are half-open and cannot overlap, so measurement time points to one set of lengths, orientation and references.
The vertical profile is calculated from the fixed base upward, with A, B and resultant outputs. Historical calculations keep the version appropriate to their time. Changing the number of nodes or the length does not require a data schema change; it requires activation of a new configuration. Explicit recalculation of an earlier period is a separate operation.
For a correctly configured device, the first valid set of automatic data can create the first active version and the system profile and trend dashboard. Later changes are subject to versioning and audit. Before activating a new version, the administrator should compare a sample of the profile, the time of change and the source physical state. The platform does not offer import of campaigns from a manual probe or a promise of automatic restoration of the previous state. The rules for selecting the reference are described in the guide zero measurement in monitoring.
Withdrawing a version does not automatically reverse decisions or reports already made on its basis.
How versioning supports evidential value
In a dispute or audit, the question is rarely only "what was the value?". Answers are needed to the following:
- which configuration was in force at the time of measurement,
- from which base measurement the change was calculated,
- when and why the segment was replaced,
- who approved the new version,
- whether the report was later recalculated,
- which raw data remained unchanged.
A version with a time interval connects these elements. It does not by itself prove that the measurement was correct, but it makes it possible to reproduce the applied model. This is an important complement to the principles described in the article monitoring data as evidence.
ISO 18674-1 presents monitoring as a process that supports the assessment of soil, structure and works behaviour over time. Maintaining configuration context is a practical condition for comparability. ISO 18674-3 covers the inclinometer method and its application to displacements along a line.
Limitations and boundary cases
Versioning does not fix poor installation, unreliable time or a moving anchor. If a given version was built on the wrong orientation, it remains an auditable but incorrect model. Correction and impact assessment are needed.
Not every descriptive change requires a new model. A typo in a note can be corrected under audit rules if it does not affect the result. However, any change to the number, order, length, axes, unit, reference or calculation rule must be treated as a new version.
GEOKON Model 6140 instruction manual describes a specific in-place string, its installation and profile calculation. It is not documentation for Inclify nor proof of compliance for every device. Project parameters must be approved for the selected sensor.
The Crossrail lesson on SAA shows that changing the assumption of a fixed base had a significant impact on displacement interpretation and required survey data. It is a reminder that not only numbers should be versioned, but also the assumptions that give them meaning.
FAQ
Does every change in segment length require a new version?
Yes, if the length is used in converting angle to displacement. Overwriting would change historical results or their meaning. The new version should take effect from the moment the physical arrangement changed, after the reference, orientation and complete tests of the full engineering profile have been approved.
Can the active configuration be edited?
Parameters that affect calculations should not be edited in place. The correct procedure creates a new draft version based on the active one, lets it be verified, and then activates it from a defined time. The earlier version then moves to history and still reproduces all old results unambiguously.
What does half-open interval mean?
The interval [from, to) includes the starting moment but not the ending one. When a new version starts exactly at 13:20, the previous one ends at that same moment and the measurement at 13:20 belongs only to the new version. This removes boundary conflict and ambiguity in a single calculation.
Does activating a new version recalculate the entire history?
It should not do so as a hidden side effect. The new version defines the configuration for its period. If earlier results need correction, a separate run with scope, tests and controlled publication is launched. This prevents a mix of old and new values in one report.
Can you return to the previous configuration?
You can use the same parameters as the basis for a new version if they match the current physical state and pass approval. This does not mean automatic restoration of old data, references, alarms and reports. Each effect must be assessed separately and the new exact effective time must be recorded.
What should be done if the exact segment replacement time is unknown?
Do not guess without marking it. Collect the work log, device communication and the first data after the change, determine the best documented interval and describe the uncertainty. Data from the transition period may need to be excluded from interpretation. The decision, time boundary and impact on reports should be approved by the responsible engineer.
Sources and further reading
- ISO 18674-3:2017 - Measurement of displacements across a line: Inclinometers, ISO.
- ISO 18674-1:2015 - Geotechnical monitoring by field instrumentation: General rules, ISO.
- Model 6140 Vertical MEMS In-Place Inclinometer String - Instruction Manual, GEOKON.
- The Use of Shape Accel Arrays for Measuring Retaining Wall Deflection, Crossrail Learning Legacy.
- Manual Inclinometer Probe - Data Interpretation Guide, GEOKON.
Next step: carry out the change on a draft version
If you are planning to replace a segment or reorganise an existing string configuration, prepare the node list, work times, references and one data sample. We will help you build a new version and activation criteria without overwriting the historical context. Schedule an inclinometer configuration review.