DGPS Problems, Deep Water Drilling, and a New DP Sensor
Differential global positioning systems (DGPS) are extremely popular on DP vessels, as they are usually readily available, relatively cheap, and usually reliable.?It is a technology that is now used almost everywhere - land, air, and sea.?Like anything in the real world, it isn’t perfect and improper dependence on it is responsible for many DP incidents.?It is worth understanding these problems and how they can be avoided, as DGPS is essentially the primary position keeping position reference system on most DP vessels.?That is the problem.?This preference and poor design or operation threaten vessel position keeping redundancy.
Terminology:?DGPS properly only refers to the corrected use of the American satellite global positioning system (GPS) and there are other satellite systems sponsored by other countries.?The proper term for using all of these is DGNSS but the colloquial use of DGPS will be followed in this article as the important failure modes are common to all and I lack appropriate snobbery.?Forgive my terminological inertia.
How It Works:?GPS works a bit like old celestial navigation where the angles of stars relative to the ship at a known time can be used to calculate position.?The satellites circle the Earth faster and send relatively accurate time stamped signals from relatively known satellite locations through variable atmosphere paths.?While most DGPSs don’t measure angles (light speed makes this difficult), receivers know their position from the time it takes the signals to arrive from each satellite (transmit vs. receive time, satellite position model, and geometry).?This uncorrected accuracy can then be refined by differential correction signals that correct for satellite clock and position errors, atmospheric distortion, and some receiver and internal model errors.?The resulting DPGS position is more accurate and reliable than the raw GPS solutions.?The closer the source of the corrections is to the receiver, the less the correction signal delay and the greater the chance of the source experiencing the same atmospheric distortions, so the greater the accuracy and reliability of the corrected position reference.
Why It Fails:?Once you know how something works, you know how it can go wrong.?Obviously anything that blocks or distorts the radio signals threatens the DGPS position reference solution.?Local radio sources and some atmospheric conditions can block the satellite signals.?The distorted atmospheric path to the local receiver may not match that to the source of correction and there can be local sources of reflection that can reduce accuracy.?There is a minimum satellite selection horizon to avoid reflection off the ocean but local equipment or vessels can block satellite signals or provide reflections that the receiver may prefer.?Uncorrected mistakes in the satellite time and position, incorrect corrections, uncorrected or incorrectly corrected movement of the receiver (vibration, pitch/roll/heave), and incorrect local data or models can all cause the calculated position to be wrong.?These can cause the calculated position to jump, drift, or be unreliable.??
Gyro Detour:?The possibility of DGPS jump is why the north speed error correction needs to be turned off in your gyrocompasses while dynamic positioning – the apparent high speed of the position jump can deflect the headings of all the gyros and will do so in a consistent manner.?The higher the latitude, the closer to parallel with lines of longitude (bow North or South), and the faster the apparent speed, the greater the deflection.
Partial Mitigation:?The system engineers know all this and have made considerable improvement over the years but the DGPS systems can still fail.?There are more and better satellites available, using stronger signals over multiple radio frequencies, and improved receivers, using phase detection, to estimate and avoid radio and atmospheric interference.?The DGPSs estimate the quality of each satellite’s data and the overall solution and provide indications of this for the DPO, so he can make informed decisions.?I was originally going to discuss how a DPO could detect faults in different DGPS systems but there is too much variety available for that to be a meaningful discussion.?Check your vendor manuals.?The DGPS also provides limited quality information the DP control system that performs its own repeatability and jump tests and compares the position provided with that from other sources.?Unfortunately, self-monitoring devices may be unable to detect faults due to the problem of self-reference, lack of independent data, incorrect models or assumptions, and software bugs.?This means that the quality information presented to the DPO and the DP system may be wrong without indication and only a check against independent position references can detect the fault.?That is the DP system’s primary means of DGPS position fault detection.?How this is done and the popularity of DGPSs are part of the problem.??
DP Crack:?DGPS are almost like crack to DP operators – it is so cheap, easy to use, easily available, and good that many vessels only really use DGPSs.?Hydro-acoustic position references (HPR) require beacons placed, ladar and radar usually require targets/transducers placed (SceneScan doesn’t), and tautwires require shallow water.?While a flotel can use a gangway as an additional reference and support vessels can position relative to the vessels they support, deep water drill rigs only have access to DGPS and HPR.?4 DGPSs and an HPR is officially a redundant configuration in class rules but it may not be operationally redundant.?This is a common problem.?Worse, the DGPSs often come from the same manufacturer.?Preferably, a vessel would have at least three different position references but this is not practical for a deep water rig and a moving vessel like a pipelayer will be lucky to have two (e.g. beacon on an ROV following an established visible path).?The riser angle is an indicator of position, and keeping that angle small is the goal of the position keeping, but the riser is a dynamic system and only some of its variation comes from rig position.?The problem of an overabundance of DGPSs is complicated by historic advice and DP system position reference system handling.
Bad Advice:?The dream of many a DP operator was to obtain redundant operation from two DGPSs.?If the DGPSs use different satellites and correction services then maybe it would be possible to operate redundantly.?There were even industry guidelines issued for this.?It doesn’t take too much imagination to realize that there are still common failure modes that can affect both DGPSs and industry advice now recognizes that such operations were not redundant and that the best DGPS operation is obtained by using all the valid satellites and corrections possible.?Software and hardware errors can be avoided by using DGPSs from different manufacturers but the DGPSs still use the same principle and have common ways of being wrong.
INS:?One successful improvement was to combine the DGPS with an inertia navigation system (INS).?The INS measures acceleration and integrates that to find changes in position.?This is accurate in the short term but the measurement and integration errors accumulate over time, so the position drifts.?Accumulated measurement errors make sense and integration errors are also easy to understand.?If position is 0.5at2+vt+d and we only know a (acceleration) and t (time) then errors can be expected as the INS can only produce 0.5at2.?While many people expect random numbers to cancel out, the difference between the ideal and the actual tends to increase over time and the numbers may not be random.?Combined with a DGPS, the INS filters out short term position jumps and noise but needs its long term drift corrected by the DGPS.?Such combined DGPS INS systems improve short term DGPS accuracy and can provide short term coverage of a DGPS outage (e.g. atmospheric blocking or local radio interference).?It should be noted that while short term INS failures are generally safer, rarer and more detectable than the DGPS ones, combining the references means that a short term INS fault can disrupt the combined position output.?This is still an improvement in the overall expected accuracy and robustness, but the INS does not add to longer term accuracy because of its tendency to drift and dependence on the DGPS to correct that.?In other words, the combined DGPS INS improves short term fault resistance but will follow a slowly drifting DGPS.?This can be a problem when the DP system already prefers DGPS and/or has many of them.
DP Reference Handling:?There are a number of position reference handling schemes that are used in different DP control systems and over time.?The most popular current scheme prefers fast and consistent position references to ones with slower updates or more variation.?It will automatically remove some faulty references but does so at the cost of automatically following systems with faster updating, and consistent errors.?The next most popular scheme uses manual weighting.?This allows the DPO to make the system weighting reflect redundancy but is more dependent on the DPO to detect and deselect fault position references manually.?Other DP systems require a blend of three independent position reference systems, or follow a single position reference while comparing it against standby systems, so a DPO can deselect a faulty one.?We will consider the first two schemes.
Automatic Weighting:?This scheme means that fast updating, consistent, and wrong DGPS will be preferred over slower updating or messier systems.?If we go back to our 4 DGPS and 1 HPR example, it might place 24% of the position reference weighting on each DGPS and 4% on the slowly updating or noisy HPR.?Unfortunately the four DGPSs have common failure modes and the only independent position reference system is devalued and sidelined by this non-redundant weighting.?It has the advantage that if one of the DGPSs malfunctions then it will be voted out, if consistent, and deweighted, if not consistent.?A fault in the HPR would have little effect on the position.?The weighting of the independent position reference could be improved by buying lots more HPRs or deselecting DGPSs but even a single DGPS would outweigh deep water HPRs due to the slow update rate.?There are several means of improving both HPR fix consistency and update rate but a popular one is combining the HPR with an INS to improve the update rate and consistency so the DP system will weigh it more equally.?Automatic weighting requires more and better equipment to achieve redundant weighting in deep water.?Countering the automatic position reference weighting prevalence of DGPSs requires considerable investment in some circumstances but the lack of it has caused a large number of DP incidents.??
Manual Weighting:?Manual weighting allows redundant position reference weighting to be forced by the operator.?For example, four DGPSs (DGNSSs for the snobs) and one HPR would be weighted 12.5% for each DGPS and 50% for the HPR, as the DGPSs work on the same principle and have common failure modes, while the HPR works on a different principle and has few failure modes in common with DGPS.?In reality, the redundant setup would be more like 2 DGPSs at 25% each and the HPR at 50% with two DGPSs excluded from use because of the MRU or power supplies that they share with the HPR.?A severe thunderstorm can affect both reference types with radio (lightning) and acoustic (thruster) noise but storms can be forecast and this is not an issue for normal operation.?Manual redundant weighting prevents one type of position reference from dominating the position solution without investing a fortune in redundant references, but continued safe operation depends on DPOs reliably detecting position reference problems and deselecting the unhealthy systems when the position references disagree.?This dependency on DPO corrective action causes horror in some, and incorrect operator setup or actions have caused loss of position.
Auto vs Manual Weighting:?Each scheme has their proponents with DP manufacturers and operators that swear by their scheme.?The advantage of redundant weighting is obvious but the advantage of automatically deweighting degrading position reference sensors is also obvious.?The costs are also obvious.?A manual redundant weight requires the DPO to closely monitor the position references so a faulty reference can be deselected.?This requires more work from the DPO and they might be reluctant to deselect a faulty reference and lose redundancy if it affects work (e.g.?forced to stop work, report, evaluate, correct or mitigate).?On the other hand, the automatic weighting takes management of this risk out of the hands of the DPO.?The DPO is out of the loop and monitors the system but has been told that the system knows better than he does.?The system is sometimes wrong when an engaged DPO would have been right, but operators also make mistakes.?It could be argued that disengaged DPOs are less likely to catch automatic DP system weighting errors and that non-redundant weighting puts the system at risk of common errors, but it is not clear if the total risk is greater or less than that of operator errors in a manual system.?Surely, someone must have performed an analysis of this??Presumably, each DP system provider has studies that support their design viewpoint, but I am not aware of them or independent verification.?Based on basic redundancy principles, I used to recommend manual redundant weighting but I need to follow the evidence where it leads.?I just wish that there was some.?Auto weighting causes more DP incidents but it is also used in more vessels.?It is understandably annoying when auto weighting puts 80-90% of the weight on the DGPSs and then the DGPSs walk the vessel off position, but 50% weighting will take the vessel for a slower walk and the DPO might choose incorrectly.?DGPSs can go wrong together while appearing to be healthy.
领英推荐
Rule of Three:?Regardless of the type of DP system that you are using, and its weighting system, redundant operation depends on the DPO reliably detecting and correcting system problems.?Ideally, there should always be three independent reference systems online, so the malfunction in one can be caught by comparison with the other two.?Position reference systems cannot catch all their own errors and independent references are needed to perform that function.?They should work on different principles, with different system reference sensors, power, and interface, so they have different vulnerabilities.?Redundant setup isn’t always possible in some circumstances and that needs to be recognized and acknowledged, so the risk can be planned for and managed.?Even when there are only two position reference types available, there is often a feedback from the task that is being performed that can help decide between disagreeing references.?This could be increasing riser angle, a change in pipe or cable tension, a change in dive bell position, cameras, etc.?Beneath an open sky, with no other references available, DGPS isn’t perfect but it is better than no position references at all.
Deep Sea Drilling:?Unfortunately, it isn’t possible to produce a reliable reference by combining DGPS, HPR, and INS, as both DGPS and HPR are capable of slow drift and INS cannot decide between them.?It can also be difficult for the DPO.?If the DGPSs drift off together and the automatic weighting favors DGPS, then it will look like the HPRs are drifting off instead.?As the DGPSs can drift without alarm or fault indication, this will appear as drift and rejection of faulty HPRs (actually healthy) as the DGPSs were already the de facto position references.?The DPO will need to monitor the riser angles to see if the DP control system was right or if he needs to reverse the situation.?Drift off of faulty HPRs will look identical on the DP desk but will not affect riser angle.?The riser angle needs watched for a while to confirm the situation, as there is natural variation.?Surrounding vessels probably cannot be used as a corrective reference as they are probably suffering the same common HPR or DGPS fault.?Even if the DGPSs and HPRs are equally weighted, most DP systems will tend to follow the average solution until the DP system or DPO is forced to reject one or the other.?This means that if one drifts off then the rig drifts off at half that speed until it either jumps back to the healthy systems or the drifting ones.?From the DP desk, they both look the same as the deselected position references drift off - either because they were faulty, or because they were healthy but the vessel is moving away because it is following the faulty references.?Deep water drillers should aim at achieving redundant position reference weighting and using the riser angle to decide between conflicting HPRs and DGPSs.?It is fair to note that DGPSs appear to drift more than HPRs, according to the annual IMCA DP incident reports, but they are also more commonly used.?While INS can’t help directly, accurate accelerometers can be used in a different way.
Examples:?While the general statements above are true, it is sometimes easier to understand with concrete examples, so let’s look at DGPS and HPR conflict in a deep sea rig?again:
Deep Sea DPO Intervention:?Regardless of weighting system used, it may not be possible to get a clear riser angle indication of position drift until long after a set of position references are rejected.?As the deep sea watch circles are large, this is normally not a critical position keeping problem when a DPO starts with redundant weighting.?He knows that he may have made the wrong decision, and must carefully watching the riser angle for a clear indication of if he made the right decision or if the other set of position references need selected instead.?Drift within the watch circles is not critical, riser angle changes should become apparent as the warning watch circle is approached, and the fault can them be corrected before the riser is endangered.?This may not be an available option if the standing orders with the vessel client unwisely requires work to be shut down when redundancy is lost or when position control is suspect.?With only two reference types and the possibility of loss or malfunction of all references of one type, maintaining full sensor redundancy is not a realistic expectation, as the riser angle is only a useful deciding factor after considerable movement.?Ignoring vessel client requirements, a DPO who has been trained to trust automatic weighting may be in trouble after it automatically rejects one set of references.?He has less reason to keep an eye on the riser angle because he expects the system to have done the right thing despite it lacking information with which to make the decision.?But if he is properly skeptical of the system, he will be able to detect and react to the riser angle indication and correct the problem before critical.?As DGPS drift is uncommon, HPR drift is rare, they work on different principles, and assuming no common system references, power supplies, or interfaces, it is normally safe to assume that HPRs and DGPSs will not drift at the same time.?I may have once read an example of both drifting, but reports and the actual incidents are different, so it is hard to tell.?Another thing to remember about position reference drift is that some of the faults have a limited duration and correct themselves.
Force Control rather than Position:?We can’t always measure position three different ways but we can always measure force, so it is strange that we don’t do so.?Force equals mass times acceleration.?High quality accelerometers are available for use, for example in the INS, and it is possible to determine the mass and mass distribution of the vessel with high accuracy.?Rather than introducing errors by double integrating the measured acceleration, better precision might be obtained by using the acceleration directly.?We complain about not being able to measure the force from current, wave slap, and ice pressure, and wind feed forward is sometimes counterproductive because the measured wind is different than the net wind acting on the vessel, due to variation in wind pressure and direction and loading or crane operation changing the wind sail without changing the model used.?Thruster force output is also an estimate.?High accuracy INS accelerometers provide a means to calculate the net force acting on the vessel, so they can be counteracted.?If the stern of the vessel is not accelerating but the bow is then the controller knows to increase bow thrust while balancing the turning moment with the stern thrusters.?It doesn’t know if the force imbalance is due to wind sail, wave slap, a narrow band of current, ice pressure, or an inaccurate thruster.?It doesn’t know and it doesn’t care, it just balances forces.?A traditional DP system uses a slowly adapting vessel model and position references to estimate and balance forces to maintain position, but if net forces can be quickly and accurately measured then they can be quickly and accurately opposed and this solves a number of problems.
Force Reference System Implementation:?Let’s look at a simple, non-redundant, system as an example - a minimum of four accelerometers should be used – forward (bow) of the center of gravity, aft (stern), port, and starboard.?Forward and aft allow cancelation of pitch and measurement of hull flexing while port and starboard correct roll.?(There is a decent argument for four corners, a better one for three dimensional sensor placement, a need for redundant sensors, and compensation for local vibration sources, but let’s keep to the simple concept for now.)?While the accelerometers cannot directly measure changes in vessel mass, their outputs can be used to accurately determine the vessel’s acceleration and flexing, changes in the location of the center of gravity (as loading changes (cargo, crane lifts, ballast, fuel transfer) and waves lift different portions of the vessel), and the difference of the forces forward and aft.?High accuracy calculation of force needs more than the shipyard estimate of mass and redundant accurate displacement measurements should be used to correct vessel changes as fuel is burnt and the vessel loaded, unloaded, and loads moved (including fuel & ballast).?The interface of the improved, redundant stability/loading measurement system and redundant, corrected accelerometer measurement could be used to create a force reference system that provides several DP advantages.??The drawing below shows the simplified implementation discussed and some of the problems it can address. A fully redundant configuration would different.
Force Reference Advantages:?A good force reference system would put model control to shame and eliminate the problem of dealing with quickly changing, unknown, environmental or mission forces, as the effects can be measured and compensated.?It is not a position reference as it cannot detect position or velocity but it can limit drift in a way that an INS cannot and compensate for forces that are usually unknown.?A well-tuned DP vessel that keeps the velocity of its position hunting low will only drift at that low speed if the position references are lost as the force reference system can only measure and correct the changes in velocity of the vessel (F=ma) but not the initial velocity of the vessel.?This is better than the model control after loss of all position references because it can be sustained much longer and works in changing, as well as constant, environments.?With limited information, current DP control systems can only perform limited reality checks.?These reality checks do not work in all circumstances and the assumptions of vessel movement can sometimes cause problems.?If the vessel is hit by a large wave, the sudden rapid movement that results will often cause loss of all position references, as that is beyond the assumed circumstances, but a force reference system can validate it as real and start correcting for it.?An accurate force reference system could improve on model control, footprint tightness, and system response to unknown forces.?It would probably replace wind compensation as it looks at the net vessel effect rather than applying the wind in three locations to an assumed valid wind model.?An accurate and redundant force reference system could be used like wind feed forward to quickly respond to changes in net force but there are advantages to further integration with the DP control system.
Force & Position References:?Going back to our deep sea vessel with only DGPSs and HPRs as position references and a disagreement between the two sets of sensors due to a common fault affecting one set of sensors and causing either DGPS or HPR drift, an INS cannot decide between the two types of sensors but the force reference system knows the actual force acting on the vessel.?The difference between the force required for each sensor groups apparent position change and the known actual force can help decide which group of sensors is wrong.?They can be automatically deselected or deweighted by the DP system.?Without intelligent design, it will not always be able to help but it will improve the odds.?Detection of slow and consistent drift is still a problem as the force reference can only measure changes in velocity but it might be possible to create a heuristic automated test if some vessel movement is allowed to diagnose the problems.?The force measurement system would need deeply integrated with the DP system to allow this and other functions but I can see why vendors would like to sell add on sensor systems for existing DP systems.?Creation and use of force reference systems would solve a lot of problems and add to DP reliability and effectiveness.?Fast, reliable, and redundant direct force measurement trumps current slow model estimates.?It is probably the next major improvement in DP control systems
Heuristic Drift Test:?Kinetic energy is 0.5mv2 and a force reference system knows the mass of the vessel very well while the DP system knows the velocity implied by both the DGPS and the HPR references.?In a more responsive system, we would just perform an impulse test to deplete some of the kinetic energy by applying a known energy and confirm which of the two models are true.?If the vessel was not moving then application of the impulse against the apparent movement would create a movement against the direction of the assumed drift in one reference and slowed movement in the other.?If the vessel was moving the application of an impulse of a known size will confirm the amount of kinetic energy by the change seen in the two sensor systems (I’m generalizing as I don’t know the sensor weighting used in the system), as kinetic energy is proportional to the square of the velocity so a controlled burst of opposing thrust for a short time will have a measurable velocity effect that varies with the hidden speed.?This is harder to perform with a vessel with ramped thrusters of variable calibration.?With sufficiently accurate control and measurement, the system could perform an automated test where the response will identify the drifting set of position references.?To go back to our examples, the vessel moving north at 0.2m/min has one sixteenth of the kinetic energy of the vessel moving south at 0.8m/s and in the opposite direction, so application of a known amount of work to each will have different effects.?By the way, we can accurately measure thruster output, we just choose to not keep calibrated load cells in the thrust bearings.?We assume a lot of things that we can measure in DP because they required considerable effort.?Better performance is possible when better measurement is used with improved automation.
DGPS & Force Reference:?Sometimes vessels have no choice but to operate with a single position reference type.?This is usually DGPS but is sometimes HPR (scintillation).?A force reference system can measure if the DP system is applying net force to maintain an apparent position.?Balanced force is used to maintain position but net force causes acceleration and movement.?While it will still have problems detecting slow drift without clever design (e.g. a practical adaption of the heuristic test given above), the force reference will provide an effective sanity check while operating with only DGPSs or other lone reference system type.?If following the DGPSs requires excessive acceleration (net force to change position) to keep position, it can be rejected and force control defaulted to.?With force control capable of balancing environmental forces for an extended period and a mission load reference such a dive bell, riser, or crane load, an integrated system might allow the DPO to move to cancel velocity or make a position change to make safe or finish work while force control balances the net forces that would otherwise move the vessel.
Conclusion:?DGPS is a good and useful technology but the solution to its occasional problems is not more DGPSs but more independent references.?Too many online DGPSs can threaten vessel redundancy by making a common DGPS error dominant over other sensors.?Redundant DP operation requires maintaining three independent position references.?Where that is impossible, the DPO needs to be able to clearly monitor the vessel task as an aid to monitoring position.?Where operation is dependent on DGPS, leave room in the operational plan for its malfunction and limiting its effect.?DGPS is a popular DP pal, but don’t let it drive drunk.?Even where no other position reference system is available, a force reference system will provide a DGPSs sanity check, as well as improved DP system performance.?Someone needs to build one.?Every DP vessel should have them.
Marine Systems Specialist
3 年well written and informative ! ??