• No results found

Why-Because Analysis (WBA) 1051

Detailed description of RCA techniques

C.6 Why-Because Analysis (WBA) 1051

whether A is a necessary causal factor of B. Suppose two events or situations A and B have 1056

been observed. The CT asks whether, had A not occurred, B would also not have occurred.

1057

(Since A did occur, a supposition that A had not occurred is contrary to fact, hence the word 1058

“counterfactual”.) In asking this question all other conditions are assumed to have remained 1059

the same. If the answer is yes: B would not have occurred then A is a necessary causal factor 1060

of B. If the answer is no: B could have happened anyway even if A had not happened ( the CT 1061

fails) then A is not a necessary causal factor of B.

1062

The network of causal-factors is displayed as a Why-Because Graph (WBG), a collection of 1063

“nodes”, boxes, diamonds and other shapes, containing a brief descript ion of the fact, joined 1064

by “edges”, or arrows, where the node at the tail of an arrow is a necessary causal factor of 1065

the node at its head, as determined by the CT.

1066

A WBA is acyclic (contains no loops), so is usually drawn with arrows pointing in the gener ally 1067

upwards direction, as above, or horizontally with arrows pointing generally left -to-right, or 1068

right-to-left.

1069

In order to determine whether sufficiently causal factors are present in the collection of events 1070

and situations presented, the Causal Complet eness Test (CCT) is used. The CCT is applied 1071

to a given event or situation and its collection of necessary causal factors as determined by 1072

the CT. If the CCT is not passed, then the collection of events and situations must be 1073

extended by further factors until it is passed. Suppose A1, A2,...,An have been determined to 1074

be necessary causal factors of B by the CT. Then the CCT is deemed to be passed if, had B 1075

not occurred, one of A1, A2,...,An would not have occurred either (formally, NOT -B is a 1076

necessary causal factor of NOT( A1 AND A2 AND...AND An) as determined by the CT).

1077

When a WBG has been constructed and the CCT is passed for all the events and situations 1078

therein, then the WBG is complete and is deemed to represent a sufficient causal explanation 1079

of the focus event.

1080

Figure C.7 illustrates an example of a WBG for a commercial-aviation runway-overrun 1081

accident.

1082 1083

(11) aircraft hits earth bank

(13) earth bank in overrun path

(16) built by airport authority for radio

equipment

(12) aircraft overuns runway

(15) unstabilized

approach

(18) crew’s actions

(14) braking delayed

(17) wheel braking

delayed

(20) aquaplaning

(25) low weight on each

main gear wheel

(30) excessive speed on

landing (27)

condition of runway surface

(24) runway very wet

(28) weather conditions

(29) amount of water on

runway surface

(34)

crews actions in expectation of windshear

(19) speed brakes & thrust reversers deployment

delayed

(23) braking systems

logical design

(22) actual landing

(21) divergence between consequences of design

& behaviour expected by crew

(26) behaviour expected by

crew

(33)

‘normal’ behaviour expected by crew (32)

company X procedures (31)

crews training at company X

1084

Figure C.7 – Example of a WBG 1085

C.6.2 Process 1086

WBA has the following steps.

1087

1) Determine a collection of facts deemed to be rel evant, under guidance of a stopping rule.

1088

This gives an initial collection C of facts, divided into events, states and situations.

1089

2) Select the focus event (called in WBA the Accident Event) F.

1090

3) Determine intuitively the immediate necessary causal factors of F from amongst the 1091

collection C; check using the CT. Display the results visually as a partial WBG.

1092

4) Determine intuitively the necessary causal factors of those immediate factors; check using 1093

the CT. Extend the WBG with these factors.

1094

5) Proceed to fill out the analysis (to extend the WBG) by testing each fact in C against the 1095

factors already in the WBG.

1096

6) Apply the CCT to determine whether the WBG is complete, or whether factors are missing 1097

from the collection C.

1098

7) Extend C if necessary; incorporate the new facts into the WBG using the CT. If insufficient 1099

information is available, assumptions may be included, providing they are clearly labelled 1100

as such.

1101

8) Finish when the CCT shows sufficient causal factors for each factor, under cognisance of 1102

the stopping rule. If insufficient facts are available, assumptions must be included in order 1103

to allow the CCT to succeed, but clearly labelled as such.

1104

C.6.3 Strengths and limitations 1105

The strengths of WBA are as follows:

1106

 may be performed with a minimum of training (with the use of suitable tools that provide 1107

help on extracting facts from narrative descriptions, an inexperienced analyst can typically 1108

perform a first pass WBA inside two hours);

1109

 the analysis results are easily understandable by third parties ; 1110

 the conceptual background required to perform a WBA is limited (an analyst must be able 1111

to apply the CT, and then the CCT);

1112

 any network of causally-related phenomena may be analysed with a WBA;

1113

 the reasoning behind a WBA may be formally checked using a formal logic;

1114

 can be used together with other methods; say, those providing more structure to the 1115

collection of facts.

1116

The limitations of WBA are as follows:

1117

 the method provides no guidance on the collection of facts to which the tests are applied 1118

e.g. there is no structuring of facts into categories, for example technical, procedural, 1119

human-factors, organisational and legal;

1120

 because facts are not structured, WBA provides limited guidance on corrective action, in 1121

the case recurrence is to be prevented.

1122

C.7 Fault Tree and Success Tree Method

Related documents