Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
45 Current standards for RSAs are fairly rigid as they depend only on the type and size of aircraft using the runway. How- ever, numerous factors affect the operations that may lead to aircraft overruns and undershoots. In reality, operations are carried out under varying levels of safety. This study introduces a more comprehensive approach to evaluate the degree of protection offered by a specific RSA and provides a risk-based assessment procedure that is ra- tional and accounts for the variability of several risk factors associated with aircraft overruns and undershoots. In addi- tion this study provides risk models that are based on com- prehensive evidence gathered from aircraft accidents and incidents in the United States and other countries. Informa- tion gathered from these events has been organized into a database that may be used for future studies on airport risk assessment. Moreover, the basic approach utilized in this research can be used to model other types of accidents occurring near airports, particularly for veer-off and takeoff and crash accidents. Existing and new techniques were integrated into a practi- cal approach that uses NOD and normalization procedures. Incorporating NOD allows the user to account for different aircraft flight risk exposure to assess the criticality of risk factors. Although with some limitations, the normalization procedures increased the pool of comparable data for model development. Despite some limitations on availability of information to develop the models presented in this report, the models provide useful information and are integrated in a practical approach that will be useful for RSA planning and airport safety management. Prototype software also was developed as part of this effort and serves to verify how the several models interact and determine risk probability distributions for an RSA subject to specific conditions. Major Achievements Development of Accident and Incident Database One of the accomplishments of this study was the devel- opment of a comprehensive and organized database of acci- dents and incidents. It includes the information gathered for events involving aircraft LDORs, LDUSs, and TOORs. The database has editing and querying capabilities, and includes most parameters associated with the events, including air- craft, airport, runway, operation, causal factors, and conse- quence information used in the modeling process and can be used for future studies on airport safety. The database over- came the lack of depth of many accident databases developed in previous studies. Analysis of the database revealed which anomalies are more frequent for each type of incident and the factors contributing to these events that would be desirable to incorporate in the modeling process. Frequent Anomalies in Aircraft Overruns and Undershoots When possible, the anomalies present during the occur- rences were listed for each accident and incident included in the database. Most of these anomalies were taken from accident in- vestigation report lists of causal and contributing factors; how- ever, other anomalies were listed, even when not included in the report if they were present during the occurrence. These anomalies were summarized by type of event and by severity of the event. The summary provided is useful to help airport operators understand how these factors may lead to accidents, and it helped to identify those factors more fre- quently associated with aircraft overrun and undershoot C H A P T E R 5 Conclusions and Recommendations
46 events. Whenever possible, these factors were incorporated in the modeling approach. Normal Operations Database Another achievement in this research is the availability of a normal operations database for U.S. flight operations. The NOD sample comprising 242,420 flights is a representative sample of 95.8 percent of flights in the United States and cov- ers the great majority of airport types, geographical regions, and operational and meteorological parameters. The devel- opment of this database was crucial in quantifying and char- acterizing accident risk factors, as well as the development of accident frequency models. In addition, the NOD provides an assessment of flight risk exposure and may be used for future studies on airport and runway safety. Development of Risk Models for RSA Each set of models includes one model for frequency and two models for location: one developed from raw distance data and one based on normalized distance data. A rational approach was developed to model severe consequences and serves for any of the model sets. These models incorporate several risk factors identified for each type of accident. These include a range of meteorologi- cal conditions, operational parameters, and existing features and obstacles present in the RSA and its surroundings. In addition the approach allows consideration of multiple risk factors in a single model that accounts for their joint influ- ences on accident likelihood. Consequently, the models developed have substantially improved predictive power, with enhanced sensitivity and specificity, versus existing reg- ulations based on a small number of risk factors. Integrated Practical Approach This study also brings a new rational and probabilistic ap- proach to integrate the frequency and location models for evaluation of the likelihood of an accident with severe conse- quences. The approach accounts for the variability and risk exposure relative to various factors and provides a proba- bilistic assessment of risks. The output of the risk assessment is a probability distribu- tion of risks for an existing RSA subject to specific meteoro- logical and operation conditions. As a result, it is possible to evaluate quantitatively the overall risk of accidents for a par- ticular runway and associated RSA. Prototype Software Running a risk assessment for a specific RSA using the ap- proach developed in this study requires extensive calculations. A sample of NOD for the runway must be representative of the traffic occurring in the airport during at least one year so that it covers different meteorological conditions for the airport. In addition, depending on judgment by the analyst, there may be multiple crash scenarios and these will require modeling for each of these situations. To facilitate these computations, software that incor- porates models developed in this study was developed. The software allows one to enter existing RSA conditions and a sample NOD for each type of operation that may challenge the RSA to run the analysis for an existing obstacle. Appendix O describes the prototype software de- veloped for this project and shows some screens, and input and output parameters. Model Limitations Modeling aviation risks always has been a challenge. Acci- dent and incident reports often lack quantitative information for causal factors of overrun and undershoot accidents. The large quantity of causal factors and the scarceness of reliable data limits the accuracy of models developed. Modeling con- sequences still depend on some judgment from the analyst. In this study it was not possible to make more solid recommen- dations on how to treat the obstacle/aircraft interaction and how much energy is required to cause severe consequences during the impact. Most likely, airport operators may need expert help to make correct use of the models developed in this study, but they should be able to use them effectively with the help of a risk assessment professional. The modeling approach on its own is not enough to allow anyone to perform a risk analysis for RSAs. The assessment of risk associated with overrun and undershoot events has to be placed in some context, a process to be followed by airport operators that will support their risk management decision making. A review of existing databases has shown that deviation data for air operations are generally sparse, particularly for reporting of relevant details on incidents (more detailed data generally are available for accidents). Collection of NOD was a challenge, particularly to counteract the impossibility of accessing the flight operation quality assurance (FOQA) data- base from the airlines. The models and approach developed in this study can be helpful for airport planning and safety management RSA improvement actions. However, under no circumstances should the models developed in this research be used to as- sess real-time operational risks. There are several factors not accounted for in the models and their accuracy is not ap- propriate for decision making during actual operations or emergency situations.
47 Recommendations for Future Work Improved Normal Operations Data Collecting normal operations data for parameters not cov- ered in this study would greatly help to improve the approach developed in this study. Some of the factors, although con- sidered important to modeling aircraft overrun and under- shoot events, were not available from the NOD sources used in this study. Of particular importance are: the runway criti- cality, the presence of tailwind and the runway friction dur- ing the operation. Information on these factors could not be obtained from the ETMSC and ASPM databases. Incorporation of these factors would certainly improve the predictive performance of the models and enhance the over- all accuracy of the risk assessment. Airline flight data recorder (FDR) information is especially relevant to obtain informa- tion on runway criticality but access to such data will require the cooperation of airlines and industry organizations. Even if data from FDRs are not available, the runway crit- icality factor may be incorporated in the models using the basic distances required for operation of each aircraft type. Although this improvement was identified in the course of this research, time and financial resources constraints did not allow the incorporation of such factors in the present models. Availability of Information for Accidents and Incidents Accident investigation and incident reports seldom contain the comprehensive information required to improve model- ing. Even when there are standard forms and reports, many of the existing fields are not filled by the reporter. In addition, some important parameters associated with events relevant to this study are not available in the standard forms, or the in- vestigator is not aware of their importance for modeling risks. Of particular interest is the runway distance required for land- ing or takeoff. Without such information or the parameters (e.g., actual weight at crash) required to compute the distances of interest, it becomes difficult to assess the runway criticality. Precise wreckage path, location exiting the runway, and final wreckage location are rarely reported, even for accidents. Details on the type of obstacle, its dimensions and location are seldom reported. Therefore, it is important to develop more comprehensive guidance related to the parameters required to improve risk modeling in the future. A list of sug- gested parameters to report is presented in Appendix N. It should be noted that the availability of information on accidents and incidents will only be helpful if these parame- ters also are available in NOD. Development of Comprehensive Software The prototype software developed for this project is fairly simple and allows risk assessment of RSA using sample NOD for different types of operation challenging a specific RSA. However, there are limitations, as resources constrained the development of more comprehensive software. The risk as- sessment is performed for each type of accident separately and does not integrate the results to obtain the overall risk. Moreover, the analysis system is capable of evaluating one simple and rectangular-shaped RSA or one existing obstacle. Although the analysis of other shapes and multiple obstacles is possible, each comprises one single analysis that will re- quire integration to be performed manually. Data normalization is still limited to one type of terrain. Ideally, the analysis should include multiple sections with different terrain conditions and possibly the capability to evaluate risks when Engineered Material Arresting Systems are available in the RSA. Development of Onboard Real-Time Models One of the products of this research is a summary of fac- tors or anomalies frequently present during aircraft overruns and undershoot. Some of these factors may greatly help pilots making good decisions about attempts to go around if the information and risk processing capabilities are provided in real-time. If onboard computers could process real-time in- formation, like approach speed, height above threshold, touchdown location, wind, and braking capability based on actual aircraft deceleration to estimate remaining distance re- quired compared to distance available, an effective warning system could be provided by aircraft manufacturers to advise pilots on the best emergency procedure (e.g., go around) to prevent many undesirable events.
