Click for next page ( 17

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 16
16 CHAPTER 4 Methodology Approach The basic goal of this study was to develop a methodology distances. In the great majority of the cases, the pilots imme- to evaluate airfield separations. There are a few different sce- diately stopped the aircraft when the aircraft departed the narios for the analysis of airfield separations, and each requires paved area. The following assumptions can be made regard- a different set of models and a specific procedure for the analy- ing taxiway operations: sis. For example, the evaluation of a separation between a run- way and a parallel taxiway requires a different set of models Aircraft travel at slower speeds relative to runway operations; than a separation between a taxiway and an object. The end of the paved area is a discontinuity that signals to The methodology presented in this report is applicable only pilots that they are off the taxiway; to runways and straight parallel sections of taxiways and taxi- Because the aircraft is traveling slower, the pilot usually has lanes and to straight sections of taxiways and taxilanes when some control and can stop the aircraft almost immediately the separation involves an object. The methodology also after departing the taxiway; assumes that the pilot has full directional control of the aircraft, These three factors combined lead to the assumption that a good visual indication of the taxiway/taxilane centerline, and the location probability distribution can be truncated for no assistance from a marshaller. The following are the types of non-ramp taxiways. separations that may be evaluated with the methodology: Most taxiway and taxilane incidents and accidents occurred Taxiway to parallel taxiway in curved segments or because another aircraft or ground Taxiway to parallel taxilane equipment was located inside the taxiway or taxilane OFA. In Taxiway to object most cases, events with large lateral deviations occurred during Taxilane to parallel taxilane poor weather conditions and situations of low surface friction Taxilane to object (low visibility, rain, and ice). Runway to parallel taxiway or taxilane Modeling of aircraft lateral and vertical deviations on the Runway to object runway involved different phases of flight, and for each phase a specific model was developed or used. For approach and The bases for the developed approach are the random lateral landing, the airborne phase was modeled using the FAA/ICAO and vertical (airborne phase) deviations that may occur during CRM, and for the rollout phase after aircraft touchdown, a normal operations and veer-off incidents. The risk of collision two-part model (frequency and location) was developed and is is related to the probability of large deviations from the nomi- described in ensuing sections. During takeoff, the factors and nal flight path and from the runway, taxiway, and taxilane cen- risk of veer-off are different from the models for landing, and terlines when aircraft are moving on parallel routes. another set of models (frequency and location) was required Despite intensive efforts to identify taxiway and taxilane for this phase of flight. incidents, it was not possible to develop two-part models (fre- quency and location) for taxiway and taxilane veer-offs due Taxiway and Taxilane to the difficulty in obtaining location data for close to 300 Deviation Modeling incidents occurring in straight segments of taxiways and very few relevant incidents occurring on taxilanes. However, it was Initially, the approach for modeling taxiway deviations was noted that taxiway and taxilane incidents due to aircraft devi- a two-part model composed of a frequency model and a loca- ations do not lead to departures from the paved area of large tion model. Taxiway veer-off incident data were collected to

OCR for page 16
17 obtain information on causal and contributing factors (visi- extreme deviations at ANC and 200 most extreme deviations bility, taxiway surface conditions, etc.) to develop a frequency at JFK. The model resulting from the analysis was in the fol- model to estimate the chance of taxiway veer-off incidents lowing general form: occurring under certain conditions. In addition, it was initially -1 c 1 y - thought that lateral deviation information for those incidents p = 1 - exp - 1 + c should be collected to develop location models, particularly n for aircraft that departed the paved taxiway area. With these where two models, it would be possible to estimate the likelihood of an incident occurring with a lateral deviation greater than a y is the specified threshold of exceedence, certain distance from the taxiway edge. Although enough p is the probability estimate of exceeding the threshold y information was available from the data collection exercise to distance from the centerline, and develop frequency models, information on incident location , , n, and c are extrapolation parameters for the model. and deviations was not available. Cohen-Nir and Marchi (2003) conducted another study A couple of observations led to the conclusion that a two- using the data collected from JFK and ANC. In processing the part frequency/location model was not the best approach for JFK deviation data, they identified some problems with the modeling taxiway veer-off incidents and the possibility of air- data collected. Several large deviations that would have put craft collisions with other aircraft and fixed or movable objects. the B-747 outside of the taxiway were recorded in a very short The first observation was that historical taxiway collision period of time. One of the lasers used in the FAA experiment events were not related to taxiway deviations; in almost every had gone out of service, and all subsequent unusually large case, the collision occurred because another aircraft or mov- deviations occurred with only one laser in service and no abil- able object was inside the wing path of the taxiing aircraft. The ity to measure speed, wheelbase, or direction of travel. Because second observation was that when a taxiway veer-off occurred, of the anomalies associated with the data collected from JFK, the aircraft stopped immediately after it departed the paved only the models developed for ANC were used in the airfield taxiway area. separation analysis methodology for this study. Therefore, the combination of frequency and location mod- To avoid collision between two taxiing aircrafts with WS1 els will only help to estimate the probability that the aircraft will and WS2 wingspans, the combined deviations need to satisfy stop off the taxiway paved area, and it will not be possible to the following equation: quantify the risk of wingtip collisions associated with large deviations to evaluate the taxiway separation. In addition, very TTWY-TWY > (WS1 + WS2 ) 2 + d1 + d2 few incidents were found to occur when the pilot had control of the aircraft. For taxilanes, even the frequency model could where not be developed due to the lack of cases with which to build a TTWY-TWY is the required separation between the taxiway database. centerlines and The alternative selected was to use the taxiway deviation d1 and d2 are the deviations of each aircraft, as shown in models developed by FAA/Boeing (Scholz, 2003a and 2003b). Figure 10. The approach and models proved to be in line with the goals for this project. Appendix G provides a summary of the A collision between an aircraft wingtip and a fixed or mov- methodology and results from the FAA/Boeing studies. The able object can be avoided if: first report describes the study and deviation models for John TTWY-OBJ > WS 2 + d F. Kennedy International Airport (JFK), the second describes the study for Ted Stevens International Airport (ANC), and where the third combines the deviation models from both studies TTWY-OBJ is the separation between the object and the taxi- and describes the models for risk of wingtip collision between way centerline, two aircraft and between an aircraft and an object. The colli- WS is the wingspan, and sion risk between two taxiing aircraft can be estimated using d is the aircraft deviation from the taxiway centerline. the mean wingtip distance between the two aircraft, in other words, the wingtip clearance when both aircraft are located This situation is presented in Figure 11. on the taxiway centerlines. One important observation when using the models devel- The deviations at each airport were extrapolated to more oped with data gathered from ANC and JFK is that taxiway extreme deviations as they could happen for significantly higher centerline lights were available in all taxiway sections that were numbers of event exposures, for example, 106 - 109 taxiway monitored. Therefore, probability distributions characteriz- operations. Based on the extreme value limiting assumption, ing those deviations rely on the conspicuity of taxiway center- absolute deviations were extrapolated using the 700 most lines. An important risk control recommendation when using

OCR for page 16
18 Figure 10. Taxiway/taxiway separation. Figure 11. Taxiway/object separation. this methodology is the assurance that taxiway centerlines will Although the data used to develop the lateral deviation mod- be evident under any conditions when using the taxiways. els were collected for the B-747 only, the same models were used to develop risk plots for aircraft belonging to other ADGs. This is a conservative assumption because smaller deviations Taxiway/Taxiway Separation are expected for smaller aircraft. The data presented in Table 4 were converted to a plot based The analysis procedure for taxiway/taxiway separation is on the wingtip separation, which is presented in Figure 12. based on the models presented in the FAA/Boeing study to assess taxiway/taxiway collision probabilities, which are based Table 4. Required separation between taxiway on the wingtip separation of two aircraft located at the center- centerlines. lines of two parallel taxiways (Scholz, 2005). In this approach, lateral deviations were split into two halves, and the devia- tions from the first half were randomly paired with deviations from the second half. In all, 6,157 pairs of d1 + d2 were obtained. For each set of such pairs, the extreme value extrap- olation method was applied to the absolute value of d1 + d2 to obtain estimates of deviation probabilities. To correct for random splitting and pairing effects, the process was repeated 500 times, and a combined estimate with confidence bounds was obtained. Table 4 presents the probability results, which were applied to the procedure to evaluate taxiway/taxiway separations described in this report. Table 4 presents the probability of wingtip collision based on the separation between taxiway cen- terlines and the wingspan distance. To facilitate the analysis, plots were developed for each Aircraft Design Group (ADG) based on the maximum wingspan for the specific ADG.

OCR for page 16
19 1.E-02 1.E-03 1.E-04 Risk per Operation 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 5 10 15 20 25 Taxiway/Taxiway Wingtip to Wingtip Separation (ft) Figure 12. Taxiway/taxiway collision risk. The same taxiway/taxiway separation models are used to ADG. As shown in Figure 11, wingtip separation is the dis- evaluate taxiway/taxilane separations. The estimates of risk tance of the object from the centerline of the taxiway less half for this situation are considered conservative since the speed of the operating aircraft wingspan. Collision happens only if of aircraft is usually slower in taxilanes and deviations are the deviation of the aircraft exceeds this distance to the side expected to be smaller than they are for taxiways. of the taxiway where the object is present. One simplifying The extrapolation of risk for wingtip separations larger assumption is necessary: the wingtip deviation distribution is than 25 ft may lead to inaccurate results and very low risks. the same as the aircraft lateral deviation distribution from the The standard separation for ADG V is 53 ft; based on the taxiway centerline. models, the risk for such a condition is lower than 1.0E-15, Using data from the FAA study, it is possible to estimate or one event in one quadrillion operations. This is no sur- the probability of taxiway-object collision based on the mean prise, as there are no reports of wingtip collisions between wingtip separation; in other words, the separation when both two aircraft on parallel taxiways. In the accident and incident aircraft are located at the respective taxiway centerlines (Scholz, data collected, no record was found for wingtip collisions 2003a). The basic assumption in this case is that lateral devia- between two aircraft in parallel taxiways, and therefore the level tions are similar, independent of the type of aircraft. Although of protection provided by the standards may be considered this cannot be proved using existing data, the assumption is very high. conservative because the data used to model risk were gathered From the point of view of risk and based on the records for large aircraft. of incidents and accidents, the worst credible consequence The resulting model based on taxiway deviation data col- expected for wingtip collisions of two taxiing aircraft is air- lected at ANC is presented in Figure 13. craft damage. In this case, according to the risk matrix recom- mended by the FAA, the risk is acceptable if it is less than one Taxilane/Taxilane Separation in 10 million operations (1.0E-07), the same collision risk probability level used by ICAO in CRM analysis. Aircraft deviation data or studies were not available to develop probability models for taxilane deviations. The data Taxiway/Object Separation collection exercise carried out for this research identified a few taxilane veer-off incidents, and insufficient information was Using the models based on ANC data developed by FAA/ available to develop any lateral deviation models. Records of Boeing (Scholz, 2003a), the wingtip collision risk can be derived taxilane accidents demonstrate that, in most cases, another air- based on the wingtip separation from an object when the air- craft or movable object was parked or located inside the taxi- craft is located at the taxiway centerline. Using this approach lane OFA. Therefore, the occurrence could not be considered provides more flexibility to the analysis for specific aircraft a "taxilane deviation" case, and it could occur independently wingspans, rather than considering the largest wingspan in the of the existing taxilane separation.

OCR for page 16
20 1.E-02 1.E-04 1.E-06 Risk per Operation 1.E-08 1.E-10 1.E-12 1.E-14 1.E-16 0 10 20 30 40 50 60 Taxiway/Object Wingtip Separation (ft) Figure 13. Taxiway/object collision probability based on wingtip separation. FAA established separation criteria for taxilanes and taxi- the wingtip to wingtip separation distances associated with ways and made changes in 1989 (FAA, 1989). These stan- each level of risk for taxiways was adjusted using the ratio cor- dards are based on aircraft design categories. There is no ref- responding to the ADG. For example, a ratio of 0.58 was erence describing the quantitative basis for the criteria, and applied to the model for lateral deviations on taxiways for it is likely that engineering judgment was used to define those ADG V. The resulting model is illustrated in Figure 14. standards. The plot shows the risk trend for taxilane/taxilane separa- Both the FAA and ICAO recognize that aircraft deviations tion of ADG V aircraft based on the mean wingtip separation in taxilanes are usually smaller than those occurring in taxi- (i.e., based on the wingtip distance when both aircraft are ways. Aircraft taxiing in taxilanes are moving at very low speeds, located on the centerlines of parallel taxilanes). The standards and pilots are usually very focused on parking operations in for taxilane separations are considered conservative, given areas where movable objects are common. the lack of recorded taxilane incidents associated with lateral For the approach presented in this report, the ratios of deviations. wingtip separations of taxiway/taxiway to taxilane/taxilane for Figures similar to Figure 14, based on separation between each ADG were calculated. As shown in Table 5, the ratios var- taxilane centerlines, were developed for each ADG based on ied from 0.75 for ADG I to 0.58 for ADG V and VI. the ratios presented in Table 6 and the maximum wingspan These ratios were used to adjust the models used for taxi- for the ADG. way/taxiway separations developed by the FAA. The rationale is that the risk for veering off a taxilane OFA should be simi- lar to or lower than the risk of veering off a taxiway. Therefore, Taxilane/Object Separation Similar to the approach for taxilane/taxilane separations, Table 5. Taxiway/object and taxilane/object mean adjustment factors were applied to the FAA/Boeing models wingtip separations for aircraft design groups. for taxiway/object wingtip separations (Scholz, 2003a). The ADG Distances in ft ratios are the same as those used for the taxilane/taxilane model Item because the standard wingtip separations between two aircraft I II III IV V VI Taxiway/Object and between an aircraft and an object are similar according 20 26 34 44 53 62 Wingtip Separation to AC 150/5200-13 (FAA, 1989). Taxilane/Object 15 18 22 27 31 36 The standard minimum wingtip to object separation for Wingtip Separation ADG V in parallel taxilanes is 31 ft and, based on Figure 15, Ratio 0.75 0.69 0.65 0.61 0.58 0.58 the risk of wingtip collision is lower than 1.0E-09.