The industry's recognized authority for design, engineering and application of equipment and services in the global ocean community
Issue link: http://sea-technology.epubxp.com/i/497263
www.sea-technology.com April 2015 / st 49 B MT Scientifc Marine Services (Escondido, California) recently developed and felded an innova- tive watch circle processing system that reduced false alarms caused by GPS errors. The application was an offshore oil and gas platform with stringent positioning requirements. GPS errors can cause a false indica- tion that a platform is out of position, requiring emergency safety measures and stopping production—both un- necessarily. The solution was a watch circle processor based on an adaptive Kal- man flter. The flter estimated the error of each GPS receiver and grad- ually lowered the weight as error in- creased. In this way, even gradually increasing errors were rejected before they had the chance to contaminate the sys- tem's estimated platform position. The watch circle proces- ser system developed and felded by BMT runs unattended in real time and is now operating offshore. Error Rejection Methods There are several approaches to dealing with measure- ment errors in Kalman flters. The frst is to examine the in- dividual measurements before they are used. Outlier mea- surements can be rejected if their measured value is more than a certain number of standard deviations away from their predicted value. While this approach can be effective in rejecting sudden jumps, it can be ineffective in isolat- ing gradual errors. With gradual errors, the erroneous data may contaminate the state estimate before the error can be detected and isolated. Once the position estimate is con- taminated, and perhaps lies between the true and errone- ous positions, it may be diffcult to decide which sensor is responsible for the error. A second approach is to compute the Kalman flter state update using all measurements, and then examine the pro- posed state update before applying it to the estimated state. The state update can be suspended or altered if it seems to be contaminated by outlier measurements—if, for example, its magnitude is beyond a reasonable number of standard deviations. This approach also has its limitations. Once the erroneous data have been incorporated into a state update, it may not be possible to determine which sensor is at fault for the sudden jump in the update. And, as noted earlier, ap- proaches that rely on go/no-go detection of faults may have diffculty handling gradual errors. Finally, the measurement's noise variance can be esti- mated adaptively and noisy measurements can be weighted in inverse proportion to the measurement's estimated noise variance. This approach was chosen for the BMT watch circle pro- cessor system because of two important advantages. First, measurement errors are eliminated before they are allowed to contaminate the flter's state estimate. Second, by em- ploying a continuous approach (as compared to a frm go/ no-go approach), the suspected sensor can be discounted well before it is clearly identifed as bad. GPS Error Rejection Improves Accuracy Of Offshore Platform Watch Circle Kalman Filter-Based Technique Prevents False Alarms By Kevin J. Delaney Adaptively estimating GPS error variances (accounting for platform rotation) allows rejection of GPS error.