Beruflich Dokumente
Kultur Dokumente
Reliable -
Continuous -
Accurate - position reference systems.
Reliable :
Reliability is of vital importance, to operations where the loss of position reference can
course risk to life and property.
The DP vessels Position Reference Systems (PRS) are specifically designed and provided
for the DP operation purpose and normally independent of the vessel's normal navigation
equipment.
Continuous :
A typical DP system requires positioning update once per second and newer systems
(2006) are updating 4 times per second.
Accuracy :
Normal navigation systems in common use for navigation purpose are not of any value in
DP work. System like Loran-C and GPS do not fulfil the criteria mentioned above.
Loran-C and GPS all range from 15 - 100m or more, while DP operations require better
than 5 meter and if possible better than 1meter accuracy.
Artemis
Hydroacoustic Position Reference (HPR)
Taut Wire
Differential GPS/Glonass
Fanbeam/CyScan/ RADius.
DIFFERENTIAL RADIUS
CORRECTION
ARTEMIS
FANBEAM
TAUT HYDROACOUSTIC
POSITION
WIRE REFERENCE
Each of these systems operates separately and independently of the DP system, and feed
information to the DP by means of an interface. The DP system can handle the different
PRS input, and by pooling / weighting the positions the DP obtain an accurate position.
This process, is a function of the mathematical filtering the system.
Any modern DP system is able to pool position reference data from two or more position
reference systems. If only one position reference system is enabled into the DP then it is
simply calibrated, filtered and used, but if two or more are available, then some form of
pooling is required. In early DP systems without sophisticated mathematical modelling
techniques available, only one position reference input was possible, with the limitations
described in the above paragraph. For situations where two or more position references
are in use, a simple system of pooling was to average the output data from the PRS in
use. This method has disadvantages in that if one PRS drifts, then the averaged position
will also drift. Similarly, if one of three PRS being averaged is lost, then the resolved
position will exhibit a jump to the new average position.
In all modern DP systems the pooling is reliant upon Weighted averaging. Various
methods of weighted averaging are possible. Weighting may be manually achieved, or
automatic. If automatic weighting principles are used, the basis for the weighting may be
Variance or Frequency. With Variance-based weighting, the weighting value will depend
A weighting system based upon this principle (Variance-based) may suffer problems. For
example, a very low value for Variance (thus high weighting) may result from a PRS which
is frozen, or has become a "perfect" position reference. Also, Variance-based pooling is
less useful when there are only two PRS. Further, the data update rate must be taken into
account, since a PRS with a high update frequency may appear to have a higher apparent
Variance than one with a slow update.
Voting can be defeated if two PRS suffer common-mode failure, or if two PRS become
"perfect". The former may occur in the case of the DPO placing a transponder on the
seabed using the Taut wire depressor weight as an anchor point. If the weight drags, then
the voting may reject the third (good) system. A "perfect" reference is one that has failed (a
strange term under the circumstances!) and is giving fixed or frozen data.
When the second and subsequent PRS are selected, the acceptance criteria changes to
10 successive returns within 10m before acceptance into the DP system. Once two or
more PRS are accepted, it is recommended that that the first-selected PRS be deselected,
and re-selected again. This allows a better calibration to take place, since the original
calibration was based on three returns only, while this later calibration is based on ten
returns. In the Kongsberg 70x systems, if such a recalibration is made, the asterisk
marking the actual sensor position might take up a new location (relating to the better
calibration of position), while the reference origin circle will remain in it's original location.
Thus the circle and asterisk may be separated by a small distance. This could be
confusing if the meanings of the symbols are not clear to the DPO.
For any PRS, windows are placed around a representative sample of position returns. The
size of the window relates to the spread, in metres, of the sample of position
measurements. The DP system then determines a value for the radius of the window,
called the INNOVATION. The value of the Innovation is set between 1.5 and 15 metres.
An Innovation of 1.5 indicates that the spread of position fixes from that particular PRS is
1.5m or better. The first stage of Kalman filtering deals with the Innovation values; this
is the Prediction Test, for which the Innovation is the limit. Any returns yielding positions
outside the Innovation window are rejected. This allows outliers (single spurious position
fixes at some distance from the vessel position) to be rejected. Further, any PRS which
has an Innovation value of greater than three times that of the smallest Innovation, is
In the SDP xx equipment, display RefSys page gives a graphic reference of PRS data,
with information colourcoded for each reference.
A circle is shown for each PRS of the radius equal to the Innovation for that PRS. The
Standard Deviation limit is shown as a gray circle centred upon the display centre, which is
the predicted position.
Each PRS is assigned a Weighting value; this is inversely proportional to the Innovation
value, thus the weighting is based on the relative window sizes.
However many PRS are enabled, the weighting values always total 1.0. Within this, the
larger the weighting, the smaller the Innovation or window size. For all PRS the
measurements are filtered. Position reference inputs are sampled once per second. Raw
PRS data is shown on RefSys page as small crosses in the colour corresponding to that
PRS This is unfiltered data so the crosses may exhibit significant movement. Filtering is
applied such that the new filtered measurement is equal to nine times the old filtered
measurements (Northings and Eastings) plus the new measurements, divided by ten. This
The third stage of filtering concerns the statistical mix of the two or more PRS enabled,
in order to provide the calculation of the vessel position. If, for example, three PRS are
engaged; Artemis, HPR and a Taut Wire, then we look separately at Northings and
Eastings. It may happen that the HPR system is giving noisy returns and is close to the
Variance Test reject limit. The Taut Wire is very steady; the Artemis less so. The statistical
mix calculation (for Northing only, for illustration) is as follows:
Thus, from the above we can see that the noisy measurements from the HPR are not
affecting the final position, and that the position is dependant upon measurements from
both Artemis and Taut Wire, with a bias toward the more accurate system.
When three or more PRS are deployed, a further reject limit is set and displayed. This is
the Median Limit, and its radius is 6 metres. It's function is to generate rejection of a
jumping PRS measurement through majority voting, and is not affected by the Kalman
filtering.
If a single PRS be deployed then the first and second stage filtering will be carried out, but
all other noise in the measurements will be preserved in the positional calculation. Often a
drilling rig will use two HPR transponders located on the wellhead. One of these will be
active, with the other ready for use as necessary. In these cases it is much better to use
both simultaneously.
If two position references are deployed, one good and one poor, then it is possible for the
relative weightings to be 0.99 and 0.01. Under these circumstances the poor reference will
be frequently if not continually rejected. Another problem is that there is no link between
accuracy and reliability. It may happen that the good PRS is a Taut Wire, while the poor
one is HPR. The Taut Wire carries a large weighting. The depressor weight then starts to
slowly drag through soft mud on the seabed. The DP system knows only that the relative
calibration is no longer correct, thus the system with the lower weighting will be rejected in
this case, the HPR. Thus, with two PRS only, there is a danger that a good PRS be
rejected while a poor or erroneous one be retained and used for positioning. This is a good
argument for the use of three PRS in any operation where positioning is vital or critical. It
must be mentioned here that when using HPR as a PRS the DP system will treat each
transponder as a separate PRS, each with it's own weighting. The DPO, however, must
treat HPR as one PRS only, however many transponders are deployed, as for redundancy
purposes the system operates through a common transducer and transceiver. This will not
Even if three PRS are deployed, it is possible to defeat the redundancy in the system
through poor working practice. One (once) common practice was to deploy the Taut Wire,
and locate a HPR transponder on the depressor weight before lowering. This gives two
PRS on the same downline which is most convenient. The third PRS may be the Artemis
system. The DP accepts all three PRS in the normal way, giving three steady lights on the
console. Then the Taut Wire weight starts to drag: the transponder goes with it, and the
PRS rejected is the Artemis; the only good one! The DP thinks the vessel is on location,
with good HPR and Taut Wire measurements. The reality is that the ship is driving off, and
the (rejected) Artemis is the remaining good system.
The DPO should use caution in his choice of position reference systems. For any
operations requiring system redundancy it is necessary to utilise three position references.
Two PRS are not adequate, since there will arise the question as to which one has failed
when contradictory reference data is received from the two systems. Three systems will
give more security against this possibility, especially if the DP system is programmed to
apply a PRS voting or median check.
Where three PRS are required, the DPO should choose systems which have differing
principles, e.g. HPR, DGPS and Fanbeam: i.e. one acoustic underwater, one radio/satellite
system and one optical laser system. This reduces the probability of Commonmode failure,
where one event may result in the failure of multiple references. Common-mode failure is
more likely to occur in situations where the choice of PRS has included two or more
systems the same, i.e. Artemis and two taut wires. In the latter case, even though the taut
wires are separately located and powered through independent protected supplies, it is
possible for a vessel movement to cause both taut wires to drop out of angular limits
Despite the above comments, the DPO may be obliged to use a less-than satisfactory
combination of PRS simply because a better option is not available. In these
circumstances it is necessary that great care is taken in the deployment and operation of
the available PRS to ensure that they are not for any forseeable reason.
When operations are underway requiring three PRS, and any reduction in PRS input will
result in the vessel operation being suspended, then consideration must be given to the
practice of operating with a spread of four PRS as the norm. This may seem a little
excessive but there is a logical reason. If we consider a deep water drilling operation with
riser connected, working fully redundant with three PRS The rig's procedures will demand
a riser disconnection for any degradation within the positioning capability. This
disconnection represents a considerable cost in terms of lost time. With three PRS,
degraded status is obtained if one of the three is lost. If, however, four PRS were
deployed, then the loss of one of them leaves the vessel operational (not degraded). There
is now no lost time, and slightly less urgency in getting the fourth PRS back on-line.
A freeze test rejects repeated measurements. If the variation in the measured position is
less than a system set limit over a given period of time, the position-reference system is
rejected.
The freeze test is disabled for some position-reference systems (usually GPS and/or
Artemis) due to the resolution in the data from these position-reference systems.
A prediction test detects sudden jumps or large systematical deviations in the measured
position. The limit for the prediction test is a function of the estimated position in the Vessel
Model and the actual measurement accuracy.
A variance test monitors the measurement variance and compares the variance value
with a calculated limit.
A divergence test gives a warning of systematical deviations and/or slow-drift (before the
system is rejected by the prediction test).
A median test detects position measurements that differ from the median position value
with more than a predefined limit. The test is mainly designed to detect slowly drifting
position-reference systems.
Acceptance Limits:
Narrow: Narrow limit. Corresponds to a Minimum Prediction Error circle with a small
radius. Recommended for drill ships/large vessels. For example, drive-off due to thruster
errors will be
detected.
Normal: Medium limit. Systematic drift of the position measurements will increase the
prediction error limit. Example of use is on smaller vessels with moderate position keeping
requirements.
For example transitory position drift-off due to rough sea will be allowed.
Wide: Wide limit. Systematic drift of the position measurements will increase the prediction
error limit. Suitable, for example, for sailing in Mixed/Joystick mode at high speed.
NB! Operator selection of Acceptance Limit is not used in Autopilot and Auto Track (high
speed) modes. Wide is used.