The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around *

Journal of Aeronautics, Astronautics and Aviation, Series A, Vol.42, No.1 pp.037-046 (2010) 37 The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around * Wen-Chin Li ** Graduate School of Psychology, National Defense University No. 70, Section 2, Central North Road, Peitou, Taipei, 112 Taiwan, R.O.C. ABSTRACT Human error is often the result of complex interrelationships between humans and computers. Despite the development for new technology in the aviation domain, new types of human errors have been provoked in the rn cockpit. This research applies Hierarchical Task Analysis (HTA) and utilizes the - Human Error Template (HET) method - to predict pilots errors during flight operations when performing a go-around in a large commercial transport aircraft. The method of HET is applied to each bottom level task step of the flight tasks in question based on HTA. A total of 67 pilots participated in this study including 25 captains and 42 first officers. The results show two inappropriate design induced errors and three s of pilots errors with high likelihood committed by pilots during performing go-around, Fail to execute ; Task execution incomplete ; and Task executed too late. Therefore, there is a increasing need to investigate further the impact to flight safety of such errors. The data gathered from this research will help to improve safety when performing a go-around by identifying potential errors on a step-by-step basis and allowing early remedial actions in procedures and crew coordination to be made. Keywords: Design induced human errors, Hierarchical task analysis, Human error identification, Standard operation procedures I. INTRODUCTION Many human error analysis methods focus upon the identification and classification of the errors that operators make at the so-called sharp-end of system operation, such as forgetfulness, inattention, poor motivation, carelessness, negligence, and recklessness involved in accidents (Reason, 1990), and these seek to identify the psychological factors (e.g. inattention, loss of vigilance and carelessness) and organizational influences such as Human Factors Analysis and Classification System (HFACS, Wiegmann and Shappell, 2003). Li and Harris (2006) found that 30% of accidents in military aviation fell within the violations category. These included intentionally ignoring standard operating procedures (SOPs); neglecting SOPs; applying improper SOPs; and diverting from SOPs. This figure was higher in commercial aviation, with almost 70% of accidents including some aspect of a deviation from SOPs (Li, Harris and Yu, 2008). There is a specific regulatory requirement for human-machine interface on the flight deck as an attempt to eradicate many aspects of accidents relevant to human factors at source. However, such rules relating to design can only address the fabric of the airframe and its systems so the new regulation can only minimize the likelihood of error as a result of poor interface design. It cannot consider errors resulting from such factors as poor or inappropriate implementation of procedures, etc. From a human factors viewpoint, which assumes that the root causes of human error are often many and inter-related, the new regulations have only addressed one component of the wider problem. The design of the flight deck interfaces cannot be separated from the aircraft s operating procedures. Complex flight deck interfaces, while potentially more flexible, are also potentially more error * Manuscript received, Sep. 23, 2009, final revision, Nov. 19, 2009 ** To whom correspondence should be addressed, E-mail: w.li.2002@cranfield.ac.uk

38 Wen-Chin Li prone (there are far more opportunities for error). Analysis of aircraft accident investigation reports has suggested that, inappropriate system design, incompatible cockpit display layout, and unsuitable Standard Operating Procedures (SOPs) are major factors causing accidents (FAA, 1996). The roots of human error are manifold and have complex interaction with all aspects of the operation of a rn aircraft. It was not until the last decade that design induced error had been regarded as a factor of key concern for the airworthiness authorities, particularly in the new generation of highly automated aircraft. However, Chapanis (1999) noted that back in the 1940s many aspects of pilot error were really designer error. This was a challenge to the contemporary viewpoint at the time and demonstrated that good design was very important in human error reduction in the flight deck. New generation, rn technology aircraft have implemented automated systems and computerized cockpits. However, human error accidents have become the most significant concern of researchers in the aviation industry. Dekker (2001) has proposed that human errors are systematically connected to features of operators tools and tasks, and that error has its roots in the surrounding system. The question of human or system failure alone demonstrates an oversimplified view of the roots of failure. The important issue in a human factors investigation is to understand why pilots actions made sense to them at the time the accident happened. It would appear that the human component is now the most unreliable component in the system. Li, Harris and Yu (2008) suggested that to reduce accident rate the paths to failure relating to those organizational influence and human factors must be addressed. The high levels of automation in the new generation airliners have without a doubt offered considerable advances in safety over their forbearers, however new types of error have begun to emerge on these flight decks. As aircraft s reliability and structural integrity have improved over the last 50 years, the number of accidents resulting from such failures has reduced dramatically. However, up to 75% of all aircraft accidents have a human factors component in them. Human error is now the primary risk to flight safety (Civil Aviation Authority, 1998). With regard to checklists and procedures various axioms have been developed over the years. For example, Reason (1988) observed that the larger the number of steps in a procedure, the greater the probability that one of them will be omitted or repeated; the greater the information loading in a particular step, the more likely that it will not be completed to the standard required; steps that do not follow on from each other are more likely to be omitted; a step is more likely to be omitted if instructions are given verbally; and interruptions during a task which contains many steps are most likely to cause errors. Human Error Identification (HEI) techniques are used to identify potential human or operator error in complex, dynamic systems (Kirwan, 1992). A number of different types of HEI approach were identified, including taxonomy based techniques, error identifier techniques, error quantification techniques, cognitive ling techniques and cognitive simulation techniques. Formal error identification techniques implicitly consider both the design of the flight deck interfaces and the procedures required to operate them simultaneously. They can be applied at early design stages to help avoid design induced error during the flight deck design process but they can also be used subsequently during flight operations to diagnose problems with SOPs and provide a basis for well-founded revisions. Formal error identification analysis has been used in the nuclear and petrochemical industries for many years. Most formal error identification methods operate in a similar way. They are usually based on a task analysis followed by the subsequent assessment of the user interfaces and task steps to assess their error potential (Marshall, Stanton, Young, Salmon, Harris, Demagalski, Waldmann and Dekker, 2003). The International Air Transport Association (IATA) analyzed data from 240 member airlines and found about 50% of accidents in 2007 occurred during the phrases of final approach and landing, a period which comprises only 4% of the total flight time. Most pilots are trained that executing a go-around is the prudent course of action when a landing is not progressing normally and a safe outcome is not assured. This is best practice but it isn t always a straightforward decision (Li and Harris, 2008). Knowing how to execute the go-around maneuver and being proficient in its execution are extremely important but still more is required. Pilots must possess the skill and knowledge to decide when to execute a go-around. Many accidents have happened as a result of hesitating too long before deciding to abort the landing. This research analyzed pilots performing a go-around procedure in a large commercial aircraft to identify potential areas for improvement in the design of the SOPs involved. The research aims are to identify the errors in a large commercial aircraft during the go-around and provide a basis for improving software and hardware design, and training to enhance aviation safety. II. METHOD 2.1 Participants A total of 67 Aircraft B pilots participated in this research including 57 national pilots and ten expatriate pilots. There were 25 captains and 42 first officers by job description. The range of pilots age between 25 and 60, there were half of pilots (34 participants) between 31 and 40 years old. There were 62 male and 5 female participants. 2.2 Task Decomposition After the overall task goal was specified (performing a go-around) the next step when undertaking a hierarchical task analysis (HTA)is to break this overall goal down into meaningful sub-goals, which together form the tasks required to achieve the overall goal (Annett, 2005). In the task, Operating Aircraft B safely during the go-around, the overall goal of operating

The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around 39 Aircraft B Go-Around procedures 1.1 Press TO/GA Switches 1.2 Set flaps lever to 20 1.3 Rotate to go-around attitude 1.4 Verify Thrust Increase 1.5 Gear up 1.6 Select Roll 1.7 Select Pitch 1.8 Follow Missed Approach Procedures 1.1.1 Press TO/GA Switches 1.1.2 Thrust has advanced 1.3.1 Verify TO/GA annunciation 1.3.2 Rotate to proper pitch attitude 1.5.1 Verify positive rate of climb 1.5.2 Place gear lever up 1.7.1 Select Pitch 1.7.2 Verify pitch annunciation 1.7.3 Maintain proper pitch attitude 1.2.1 Command flap 20 1.2.2 Place flap lever to 20 1.4.1 Verify adequate thrust for go-around 1.4.2 Announce go-around thrust set 1.6.1 Select Roll 1.6.2 Verify Roll annunciation 1.6.3 Turn into correct track Figure 1 Standard operating procedures of Aircraft B when performing a go-around as analyzed using hierarchical task analysis Aircraft B safely for go-around was broken down into sub-goals, for example, task 1.1 Press TO/GA Switches; 1.2 Set Flaps Lever to 20; 1.3 Rotate to go-around Attitude; 1.4 Verify Thrust Increase; and 1.8 Follow Missed Approach Procedures. The analysis of each task goal should break down into further sub-goals. This process should go on until an appropriate operation is reached. The bottom level of any branch in a HTA should always be an operation. Whilst everything above an operation specifies goals, operations actually specifically what needs to be done. Therefore operations are the actions to be made by an agent to achieve the associated goal. For example, in the HTA of the flight task Aircraft B safe: operation during the go-around, the sub-goal 1.6 Select Roll Mode is broken down into the following operations: 1.6.1 Select (Operate) Roll Mode; 1.6.2 Verify Roll Mode Annunciation; and 1.6.3 Turn onto Correct Track (Figure 1). 2.3 Research Design Within the eight sub goals for Aircraft B when performing a safe go-around, there were 17 bottom level tasks, shown as the bottom level tasks underlined in figure 1. These bottom level tasks were then broken down into 65 operational items evaluated by all participants. The HET taxonomy is applied to each bottom level task step in a hierarchical task analysis (HTA) of the flight task in question. The technique requires the analyst to indicate which of the HET error s are credible (if any) for each task step, based upon their judgment. There are 12 basic HET error s: Failure to execute, Task execution incomplete, Task executed in the wrong direction, Wrong task executed, Task repeated, Task executed on the wrong interface element, Task executed too early, Task executed too late, Task executed too much, Task executed too little, Misread Information, and Others. A full description of the methodology and all materials can be found in Marshall, Stanton, Young, Salmon, Harris, Demagalski, Waldmann and Dekker (2003). A questionnaire containing these operational items was used to ask participants if they had ever made the reported error (tick ME ) or if they knew of anyone else who had made the error rather than rate the frequency with which they believed the error had occurred (tick OTHER ). It was also hoped that this would help increase the participant s confidence in being able to report errors. If they had made the error themselves but had no desire to admit making the error, they could tick the OTHERS box and the research team would still get a mark that the error had been made during performing go-around. III. RESULT AND DISCUSSION The participants evaluated 17 operational steps when performing the go-around within which each step

40 Wen-Chin Li could comprise of one (or more) of 12 basic different types of human error. Twenty-one pilots had above 10,000 hours flying experience; 18 pilots had between 5,000 and 9,999 hours; 17 pilots had between 2,000 and 4,999 hours; 11 pilots had below 1,999 flying hours. There were 30 instructor pilots and 37 first officers. The operational step 1.1 Press TO/GA Switches contains two sub-goals, 1.1.1 Press TO/GA switches and 1.1.2 Thrust has advanced. There were 8 questions (Q1 to Q8) relating to errors occurring at this stage, each operational step over 40% either by ME or OTHER shown as Table 1. The results show that high percentage of pilots errors relevant to lack of thrust during performing go-around. Takeoff/Go-around (TO/GA) Switches (Figure 2) are designed for activating the Auto-throttle system quickly in an emergency. Pushing either one of the TO/GA switches activates go-around. During the go-around the PF (pilot flying) should press the TO/GA switches or advance the thrust levers manually. PM (pilot monitoring) verifies Auto-throttle system has activated during the go-around and monitors the thrust lever position to check that it has advanced. With the first push of the TO/GA switch, the auto-throttle system activates to establish a 2000 FPM (feet/minute) climb. With a second push of the TO/GA switch, the auto-throttle system activates in thrust reference (THR REF) to establish full go-around thrust. Failing to press the TO/GA switches may cause aircraft the aircraft to climb without thrust which will have serious consequences. Failing to press the TO/GA switch will not activate go-around thrust and the flight director will also display wrong pitch guidance to confuse the pilots following their decision to go-around with serious consequences. After pressing the TO/GA switches, the PF should check whether the thrust lever is moving forward in case of a system malfunction. Rotation without adding go-around thrust will cause the aircraft to lose airspeed and it is possible to go into stall. performance is determined by thrust and lift. Flap is usually set at 30 for landing. When executing a go-around, retracting the flaps to the 20 position can reduce drag and increase lift during this maneuver. On Aircraft B s flight deck, there is a Flap Gate which is designed to prevent inadvertent retraction of flaps beyond the go-around position. When PF commands flap 20 during the go-around he should speak loudly and clearly and the PM should place the flap lever to 20 immediately. The common errors at this point included unclear commands by the PF causing confusion or delay in the PM s correct response or misunderstanding between the crew members. Step 1.3 Rotate to Go-around Attitude consisted of 1.3.1 Verify TO/GA annunciation and 1.3.2 Rotate to proper pitch attitude. There were including eight questions in the survey (Q15 to Q22) relating to errors occurring at this stage. Each item in which over 40% of respondents suggested that either by ME or the OTHER pilot had made such an error are shown as Table 1. Aircraft B has two primary flight displays (PFDs) presenting all the parameters necessary for flight path control. The PFDs (Figure 3) provide clear go-around information when the pilot presses the TO/GA switch. Go-around initial climb performance depends on sufficient thrust and proper rate of rotation. Late/early rotation or over/under rotation may cause either too much or too little airspeed. Over rotation occurs most frequently during the go-around. When the PF performs the go-around both pilots must verify TO/GA annunciation on the PFDs. If pilots operate too early and over rotation the aircraft (such as pulled back too much on the control column) will cause airspeed drop dramatically leading to catastrophe. TO/GA switches Auto-throttle Disconnect Switches Figure 2 Control Stand Step 1.2 Set Flaps Lever to 20 consists of two further sub-steps 1.2.1 Command flap 20, and 1.2.2 Place flap lever to 20. There were six questions (Q9 to Q14) relating to errors occurring at this stage, with two particular instances having over 40% of respondents suggesting that either ME or OTHER had committed this error (see Table 1). The climb gradient Figure 3 Primary Flight Displays (PFDs) Task 1.4 Verify Thrust Increase consists of the steps 1.4.1 Verify adequate thrust for go-around and 1.4.2 Announce go-around thrust set. N1 (Engine low speed compressor speed) and EPR (engine pressure ratio)

The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around 41 are primary engine indications and are always displayed on the primary EICAS (Engine indication and crew alerting system). There were five questions (Q23 to Q27) relating to errors occurring at this stage. The particularly error prone aspects during this task by either by ME or the OTHER pilot are shown in Table 1. The common errors included wrong EPR or N1 setting however this does not happen when the auto thrust system is being used, it only happens when pilot controls thrust manually. Standard callouts should be loud and clear. The PF should closely monitor that there is adequate thrust for the go-around. When go-around thrust is set, the PM should call go-around thrust set. Good teamwork can help assure flight safety. To little thrust during the go-around will cause airspeed to fall. If airspeed is below the target speed, the pilot should add thrust immediately. Setting the incorrect airspeed at this stage will cause either a stall or a flap over speed. Task 1.5 Gear Up consists of sub-tasks 1.5.1 Verify positive rate of climb and 1.5.2 Place gear lever up. There were five questions (Q28 to Q32) relating to errors occurring at this stage, with each operational step with a frequency of occurrence of higher than 40% either by ME or OTHER being shown in Table 1. Aircraft must attain a positive rate of climb before retracting the gear. The landing gear is controlled by the landing gear lever. When the landing gear lever is moved up, the landing gear begins to retract and automatic braking occurs to stop the wheels spinning. The PM should make sure aircraft attains a positive rate of climb before retracting gear. PF commands gear up and the PM rechecks gear. Incorrect commands by the PF may cause misunderstanding between crew members with potentially serious consequences. If the pilot retracts gear with the aircraft having a negative rate of climb, it will trigger a GPWS warning. Task 1.6 Select Roll Mode consists of 1.6.1 Select roll ; 1.6.2 Verify roll annunciation ; and 1.6.3 Turn into correct track. There were nine questions (Q33 to Q41) relating to errors occurred at this stage. Each item in which over 40% of respondents suggested that either by ME or the OTHER pilot had made such an error shown in Table 1. The Mode control panel (MCP, Figure 4) provides a control interface to the autopilot, flight director, altitude alert, and auto-throttle systems. The MCP selects and activates AFDS (Auto flight display system) s (roll and pitch ) and establishes altitudes, speeds, and climb/descent profiles. There are 14 switches or selectors installed on the MCP panel. Most s activate with single push. Roll s (Figure 5) include LNAV (Lateral navigation) and HDG (Heading) switches. The PF uses roll s (HDG or LNAV) to turn the airplane onto the correct track. The PM should monitor these selections closely. When the autopilot is engaged, the PF selects a Roll Mode of either LNAV or HDG SEL. When flying manually, PF calls out LNAV or HDG SEL and the PM selects the commanded roll. Late selection or forgetting to engage LNAV will cause the aircraft to be unable to capture the planned track. This may cause ATC violation. If LNAV is disengaged by mistake, it should be reengaged right away in order to maintain the aircraft s heading in the right direction. The Pilots should closely monitor any change of annunciation. During the go-around aircraft should follow the missed approach procedure or follow ATC instructions. If either LNAV or HDG is selected, the pilot should monitor the flight director commands to be sure that the aircraft intercepts the missed approach course. Mixing up the IAS/HDG bugs on the MCP is the most common mistake made by pilots operating MCP. If the mistake is not detected it may cause the aircraft s airspeed to decrease or the airplane to turn onto wrong heading. Figure 4 Mode Control Panel Figure 5 Autopilot Director Roll and Pitch Control Task 1.7 Select Pitch Mode consists of sub-tasks 1.7.1 Select pitch ; 1.7.2 Verify pitch annunciation ; and 1.7.3 Maintain proper pitch attitude. There were 11 questions (Q42 to Q52) relating to errors occurring at this stage. Each item in which over 40% of respondents suggested that either ME or the OTHER pilot had made such an error are shown can be found in Table 1. In aircraft B one of three pitch s can be selected during the go-around: VNAV (Vertical navigation), V/S (Vertical speed), and FLCH SPD (Flight level change speed). VNAV is the fully automated function and is connected to the FMC (Flight management computer). When the VNAV switch is selected the aircraft will commence climbing or descending automatically. Pushing the V/S switch opens the vertical speed window and displays the current vertical speed. Pitch commands maintain the IAS/MACH window airspeed or Mach. Pushing FLCH SPD switch opens the IAS/MACH window and displays the command speed. The PFDs provide clear and easy-to-read pitch information when the pilot

42 Wen-Chin Li Table 1 The occurred rate of errors break down by detail operational steps for Aircraft B performing go-around Operational Steps of Go-around Procedures Description of Errors Occurred during Go-Around Occurrence rate ME OTHER Operational Step 1.1 Q1. Failed to press TO/GA switch due to pilot s negligence. 20.90% 53.73% Press TO/GA Switches Q4. Accidentally pressed TO/GA switch during normal 29.85% 44.78% approach Q5. Failed to check thrust level 38.81% 56.72% Q8. Thrust lever were not advanced manually when the 29.85% 53.73% auto-throttles became inoperative Operational Step 1.2 Q9. Failed to command flap 20 due to pilot s negligence 25.37% 67.16% Set Flaps Lever to 20 Q14. Forgot to place flap lever to 20 until being reminded 16.42% 50.75% Operational Step 1.3 Q15. Failed to check whether TO/GA was being 44.78% 46.27% Rotate to Go-around Attitude activated Q17. Late rotation, over / under rotation. 46.27% 50.75% Operational Step 1.4 Verify Thrust Increase Operational Step 1.5 Gear Up Operational Step 1.6 Select Roll Mode Operational Step 1.7 Select Pitch Mode Operational Step 1.8 Follow Miss Approach Procedures Q18. No check for primary flight display 26.87% 56.72% Q23. Failed to check go-around thrust setting 53.73% 52.24% Q25. Did not identify and correct speed deviations on time 46.27% 47.76% Q26. Forgot to call go-around thrust set 68.66% 70.15% (1) Q27. Did not identify and correct go-around thrust 35.82% 58.21% deviations on time Q28. Did not check for positive climb indications before 13.43% 55.22% calling gear up Q30. Forgot to put the landing gear up until being reminded 40.30% 59.70% Q33. Did not engage LNAV on time failed to 49.25% 58.21% (3) capture Q37 Failed to check whether LNAV/ HDG was being 31.34% 64.18% activated Q39. Mixed up the IAS/HDG bugs on the MCP 34.33% 49.25% Q42. Did not engage VNAV on time failed to capture 44.78% 62.96% Q46. No check whether VNAV or FLCH was being 38.81% 56.72% activated Q48. Did not monitor the altitude at appropriate time 38.81% 55.22% Q56. Entered the wrong altitude on the MCP and activated 29.85% 41.79% it Q57. Failed to anticipate flight director commands when 16.42% 41.79% intercepting miss approach altitude Q62 Poor or slow instrument scan 43.28% 55.22% Q65. Not using auto-flight system when available and 55.22% 65.67% (2) appropriate. presses any pitch switches. The PF uses pitch s (VNAV, V/S, or FLCH SPD) to attain and maintain the desired pitch attitude. The PM monitors this closely. When the autopilot is engaged, the PF selects one Pitch Mode of either VNAV, V/S, or FLCH SPD. When flying manually, the PF calls out VNAV, V/S, or FLCH SPD and the PM selects the commanded pitch. The common errors identified at this stage were, late or forgetting to engage VNAV on time which will cause the aircraft to be unable to capture the climb path and it may also cause the aircraft level at an improper altitude; pressing the wrong switch (such as THR) will not cause any problem, but it will delay selecting the correct pitch; if VNAV is disengaged by mistake, it should be reengaged right away to get back onto the correct climb path. Pilots should closely monitor the changes of annunciation. If the aircraft deviates from its target altitude, the pilot should correct it immediately. Junior pilots, however, tend to make excessive corrections. Excessive corrections for small deviations in

The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around 43 pitch usually happen when pilots are either not familiar with the automated systems or control the aircraft too roughly. An excessive missed approach altitude will cause serious problems. It is usually caused by wrong data being input into FMC or a wrong altitude setting being made on the MCP. It is important to monitor the altitude to avoid an ATC violation. The final task 1.8 Follow M/A Procedure shows that there was a reported frequency of 50% of respondents reporting Task execution incomplete ; 25% of Task executed in wrong direction ; and 30% of Task executed too late. There were 13 questions (Q53 to Q65) relating to errors occurring at this stage. Each item in which over 40% of respondents reported that either ME or the OTHER pilot had made this type of error can be found in Table 1. Missed approach procedure normally includes an initial heading or track change and a specified altitude to climb to, typically followed by holding instructions at a nearby navigation fix. The pilot is expected to inform ATC by radio of the initiation of the missed approach as soon as possible. At this stage, the PF controls the aircraft using the published missed approach procedure. The PM informs ATC by radio. Before pressing the altitude control selector, the PF should verify that the correct altitude has been selected on MCP. Approach minima are the lowest altitude that the aircraft can fly until the runway is in sight. The PF should decide to make a go-around if runway is not in sight at the approach minima and the PM should call Approaching Minimums to remind the PF to make this judgment. Not being prepared for a go around when approaching Minimums is a serious mistake for pilot. Mis-timing making the go-around decision may cause the aircraft to fly into terrain. On the other hand, pilots can decide to go-around before reaching Minimums. This is a safe operation but making such decision too early will consume time and fuel. Pilots have a high work load during the go-around and it is possible to fail to monitor ATC clearances which can cause a serious problem. Using auto-flight system can reduce pilots workload. It is a good decision to use the auto-flight system when available and it is often appropriate to do this during the go-around. The highest occurring frequency of operational errors reported by pilots performing the go-around task was Q26 Forgot to call Go-around Thrust Set (average 69.41%). This is an error of Fail to execute ; The second highest frequency of reported errors was for task Q65 Not using auto-flight system when available and appropriate (average 60.45%). This is an error of Task execution incomplete ; the third highest frequency of error was for task Q33 Did not engage LNAV on time failed to capture (average 53.73%). This was an error of Task executed too late. Many of the errors observed during the go-around showed an interaction between procedures and the design of the flight deck. They are not simply the product or either poor design or inadequate SOPs alone. For example, the responses to Question 8 suggested that on many occasions the thrust levers were not advanced manually when the auto-throttles became inoperative. There could be several reasons for this. For example, when a pilot decides to go-around, the first step is to press the TO/GA switches. This will activate the correct of the autothrust system. However, to control thrust manually, pilots need to press the autothrust disengage switches. Since the TO/GA switches and autothrust disengage switches are next to one another, pilots may accidentally press the wrong switch, which would cause the thrust levers not to advance during the go-around. The qualitative data providing the relevant descriptions and consequences of failing to perform properly the task step relating to task 1.3.2 Rotate to proper pitch attitude can be found in Table 2. The following are some incidents related to the sub-task Press TO/GA Switches ; (1) Pilot re-tried to push the TO/GA switch immediately (to abort the go-around) but aircraft continued the go-around operation; (2) Pilot failed to press the TO/GA switch therefore the aircraft touched down on the runway due to no go-around thrust being delivered and caused a hard landing incident; (3) Aircraft became unstable during the approach due to an unsuccessful go-around. The aircraft adopted an incorrect pitch attitude, either below normal path or climbing at too great a pitch angle attitude; (4) Flight director (F/D) did not display go-around pitch because the autoflight display system (AFDS) was not triggered; it wouldn t provide correct pitch guidance because the pitch annunciation did not change to go-around. However, the error data also show a failure to follow the required procedures in this instance, as in Question 23 ( failed to check go-around thrust setting ) which should pick up the failure of the thrust levers to advance to the appropriate setting. Such confusion of system interface components is not new. Chapanis (1999) recalls his work in the early 1940s where he investigated the problem of pilots and co-pilots retracting the landing gear instead of the landing flaps after landing in the Boeing B-17. His investigations revealed that the toggle switches for the gear and the flaps were both identical and next to each other. He proposed coding solutions to the problem: separate the switches or change the shape of the switches to represent the part they were controlling enabling the pilot to tell either by looking at or touching the switch what function it controlled. This was particularly important especially in a high stress situation. IV. CONCLUSIONS Human factors play a key role in causing air accidents. By analyzing a segment of an in-flight standard operation procedures this may help to identify not only design induced errors but also other human issues in flight operations caused by insufficient training, or poor crew coordination. There are two design-induced human errors identified in this study. The first issue is the design of TO/GA switches and Auto-thrust Disengage switches on Aircraft B which are very close to one another (see Figure 2), hence pilots may accidentally press the wrong switch. The Auto-thrust Disengage switch will disengage auto-thrust system

44 Wen-Chin Li Table 2 The qualitative data containing the descriptions and consequences of the error for sub-task Rotate to proper pitch attitude when performing a go-around. Scenario : Performing a Go-around at XXX International Airport Operational step : 1.3.2 Rotate to proper pitch attitude Error Mode Description Frequency Outcome Frequency Fail to execute Task execution incomplete Task executed too late PF s negligence from surrounding interference ( 2) A/C not rotated when manual fly (1) Pilot s incapability or system failure when A/P engaged (2) Pitch up too late or too fast (3) Panic (5) Distraction. Unanticipated go-around (2) Not enough pitch (3) Under/over rotate or rotate at an improper pitch attitude for go around (1) PF s negligence (2) Did not follow FD pitch (1) Failed to trim to prevent excessive pitch up /failed to trim to reduce forward pressure (2) Distraction. Unanticipated go-around (2) PF s negligence (2) Late rotate when go around thrust set (1) Rotate to proper pitch too slowly (5) Panic (3) Pilot s control input later than pitch change because thrust advanced (2) 15 Not satisfy the go-around climbing rate /Speed up too much (2) Close to TERR (1) A/C did not climb (3) Over speed or under speed (1) No go around pitch (3) Wrong attitude (3) Stall (2) 11 Not enough climb rate or speed too high (2) Not satisfy the go-around climbing rate (2) Climb gradient not enough or lose altitude (1) A/C over pitch which increase pilot s workload (2) over speed or under speed (1) No go around pitch (1) Wrong attitude (2) 13 Not enough climb rate (1) Speed up too much (3) Close to TERR (1) A/C continue to sink (2) Affect go-around performance (2) Wrong attitude (4) 15 11 13 which means that the thrust system needs to be operated manually. When the TO/GA switch is pushed, the auto-thrust system will provide full thrust. If te pilot accidentally pushes the Auto-thrust Disengage switch instead, no thrust will be provided. Either error will cause irretrievable consequences. The second issue identified is the design of the Heading knob and Air Speed knob which are located close to each other (see Figure 4). Despite their different shape and nature f operation some pilots reported getting these mixed up. Fortunately, when a pilot mistakenly turns IAS knob to adjust heading, it will be easy to detect because the heading display would not change. Furthermore, if the pilot mistakenly turns the Heading knob to adjust airspeed, it is also easy to detect because of the change of heading. It has to be mentioned that software design, hardware design, and training design may all have an impact on pilots performance. Although many types of human errors occurred in the cockpit were investigated these cannot explicitly be linked to incidents or accidents because of the paucity of the data in the investigation reports. These errors also represent daily issues for pilots as they make these mistakes, which they then have to correct. This research found that there were three types of errors with a high likelihood of occurrence committed by pilots when performing a go-around; Fail to execute ; Task execution incomplete ; and Task executed too late. Therefore, there is a need to investigate further the impact on flight safety of such errors. Many of the errors that were found were the types of errors that most pilots were aware of and have simply had to accept on the flight deck. It is hoped that the implementation of new human factors certification standards and analysis of their associated procedures using a validated formal error prediction methodology will help to ensure that many of these potential errors will be eliminated in the future. ACKNOWLEDGEMENT This project was supported by a grant from the National Science Council of Taiwan (NSC 96-2413-H-606-003-MY3). The author would like to express his appreciation to the National Science Council for providing outstanding contribution to carry out this research. REFERENCES [1] Annett, J., Hierarchical Task Analysis, in, N.A. Stanton, A. Hedge, E. Salas, H. Hendrick, and K. Brookhaus (Eds.) Handbook of Human Factors and Ergonomics Methods. London, Taylor & Francis, 2005. [2] Chapanis, A., The Chapanis Chronicles: 50 years of Human Factors Research, Education, and Design, Aegean Publishing Company, Santa Barbara, CA, 1999. [3] Civil Aviation Authority, Global Fatal Accident

The Analysis of Human Errors in a Large Commercial Aircraft when Performing a Go-around 45 Review 1980-96 (CAP 681), Civil Aviation Authority, London, 1998. [4] Dekker, S., The Re-invention of Human Error, Human Factors and Aerospace Safety, Vol. 1, No. 3, 2001, pp. 247-266. [5] Federal Aviation Administration, Report on the Interfaces between Flightcrews and Modern Flight Deck Systems. Washington DC: Federal Aviation Administration, 1996. [6] Kirwan, B., Human Error Identification in Human Reliability Assessment. Part 2: Detailed comparison of techniques, Applied Ergonomics, Vol. 23, 1992, pp. 371-381. [7] Li, W. C., Harris, D., and Yu, C. S., Routes to failure: analysis of 41 civil aviation accidents from the Republic of China using the Human Factors Analysis and Classification System, Accident Analysis and Prevention, Vol. 40, 2008, pp. 426-434. [8] Li, W. C. and Harris, D., The Evaluation of the Effect of a short Aeronautical Decision-making Training Program for Military Pilots', International Journal of Aviation Psychology, Vol. 18, No. 2, 2008, pp. 135-152. [9] Li, W. C. And Harris, D., Eastern Minds in Western Cockpits: Meta-analysis of human factors in mishaps from three nations, Aviation Space and Environmental Medicine, Vol. 78, No. 4, 2007, pp. 420-425. [10] Li, W. C. and Harris, D., Pilot error and its relationship with higher organizational levels: HFACS analysis of 523 accidents, Aviation Space Environmental Medicine, Vol. 77, No. 10, 2006, pp. 1056-1061. [11] Marshall, A., Stanton, N., Young, M., Salmon, P., Harris, D., Demagalski, J., Waldmann, T., and Dekker, S., Development of the Human Error Template a new methodology for assessing design induced errors on aircraft flight decks. Final Report of the ERRORPRED Project E!1970, London: Department of Trade and Industry, 2003. [12] Reason, J., Human Error. Cambridge University Press: Cambridge, 1990. [13] Reason, J. T., Stress and cognitive failure. In, S. Fisher and J. Reason (Eds), Handbook of Life Stress, Cognition and Health. New York: John Wiley, 1988. [14] Wiegmann, D. A. and Shappell, S. A., A Human Error Approach to Aviation Accident Analysis: The Human Factors Analysis and Classification System, Aldershot, England: Ashgate, 2003.

46 Wen-Chin Li