As part of the human factors input to this project, a predictive error analysis was conducted using the task-based risk assessments developed by OLE and DC electrification specialists from Balfour Beatty Rail.
The objective of this exercise was to predict the types of human error, which could occur whilst working in AC, or DC electrified areas.
The method used to conduct this analysis was a predictive form of the technique used to examine the occurrence of error retrospectively, based upon TRACEr Lite, that was applied to the incidents described in the main body of this report. The technique is driven by a task analysis, which in this case was substituted for the risk assessment referred to above.
The process for the assessment is as follows for each task in the task analysis:
(i) Determine the performance shaping factors associated with that task;
(ii) Predict the observable errors that might occur (see below for detail);
(iii) Predict the types of error (perception, memory, decision or action) that might lead to the error as described in (ii);
(iv) For the chosen error type, determine the most likely error mode (a definition of how the error type manifested itself);
(v) Determine the opportunities for recovery from the error described in stages (i) to (iv);
(vi) Where stage (v) indicates that there are recovery opportunities, determine how likely it is that recovery will be successful;
Once this process was complete, references were made to the severity ratings assigned to tasks in the original risk assessment. In order to tie the assessment results to the analysis of incidents reported earlier, each task was checked against the incident data to determine whether human performance of that task had been a causal or contributory factor in any of the incidents that were analysed.
On completion of the analysis, the results were reviewed by electrification specialists from Balfour Beatty Rail over a period of two days to check the feasibility of the errors predicted. An initial meeting was held at The Keil Centre’s Edinburgh office to thoroughly explain the rationale behind the results and ensure that the electrification specialists were comfortable with interpreting the data.
Although this form of analysis is based upon the same model as the methodology used for retrospective analysis in the main body of the report, there are some notable differences that the reader needs to be aware of to avoid confusion.
Firstly, ‘observable error type’ refers to what indication there would be to a third party that an error had been made (for example, missing a step out of a procedure). This is used in addition to the
‘error type’, which describes what happens in terms of the human information processing system of the person making the error.
Secondly, predictive error analysis is also concerned with the opportunities which exist to recover from the error, and how likely successful recovery would be. In order to do this, there needs to be some indication of how the error would manifest itself to a third party (i.e. the observable error type), because it is clearly necessary to be able to detect an error in order to be able to recover from it. Errors, which do not manifest themselves in any way (i.e. they remain inside the head of the person making the error), are clearly more difficult to recover from because detection and recovery
are in the hands of the person making the error. Errors that do manifest themselves such that others (or electronic systems) stand more of a chance of being recovered.
Finally, the term ‘extraneous act’ describes an action that is not required within the task sequence, but which has nevertheless occurred. An extraneous act is not necessarily an incorrect thing to do in itself, but within the context of the task in hand it is inappropriate. An extraneous act can be observed by a bystander, in that an action would be seen that could be recognised as being surplus to requirements for the task.
6.7.1 Results of Predictive Error Analysis
In all, 205 tasks carried out in electrified areas (both AC and DC) were analysed. They comprised tasks involved in inspection of equipment and facilities (i.e. those that do not involve physical contact with energised equipment), those tasks that involve working in close proximity to electrical equipment that may or may not be energised, and those tasks that involve intrusive maintenance of electrical equipment.
The initial results of the analysis revealed that the tasks could be divided into three different groups based on the types of error that could occur when performing the tasks. These groups of tasks were examined to identify any common themes in order to allow them to be identified in this report. The following classification system was adopted:
1. Inspection and Servicing – tasks involving only visual inspection of equipment or servicing equipment;
2. Inspection and maintenance in proximity to electrical source – inspections of components of electrified systems (e.g. conductor rail) and maintenance work in the vicinity of the track (e.g.
vegetation clearance, boundary maintenance);
3. Maintenance – intrusive maintenance of electrical equipment or working in close proximity to energised electrical equipment.
Each of these classifications will now be examined in more detail to provide:
an indication of the types of errors predicted in each case
an indication of the severity ratings associated with the tasks that make up each class of task whether or not any of the incidents reviewed for the main body of this report involved any of
the tasks included within the classification.
6.7.2 Interpreting the Results of the Predictive Error Analysis
In the pages that follow, the results of the predictive analysis are presented in a series of tables accompanied by explanations of the data. This information should be interpreted in the following way:
1. Review the list of tasks associated with each group. These are located in Appendix E.
2. Review the list of performance shaping factors for each group. These factors could interfere with performance of the tasks in this group.
3. Review the details of predicted errors. This section begins by providing an indication of the types of error that might be observed (e.g. an extraneous act). For each observable error type predicted, there is then a table, which describes how human information processing might break down and result in the predicted observable error. In each table the error type is listed first (e.g. action error) followed by the error mode – how this error might occur (e.g. selection error). Note that in some cases, there are several ways in which the
cases, all feasible ways in which the error could be generated are explained in the table5. The table then goes on to show whether or not the analysis predicts that recovery from the error would be possible, if so the assessed likelihood that recovery from the error would be successful, and a comments field to provide additional explanation.
4. Following the table there is a list of incidents reviewed for this study, which involved relevant tasks from the risk assessment, and a brief explanation of what happened in each case.
6.7.3 Inspection and Servicing List of Tasks
See list at Appendix E.
All tasks in this group received risk rating of “5” or less in the risk assessment (i.e. low risk).
Performance Shaping Factors
The tasks covered by this classification were most likely to be affected by the following factors:
Weather Lighting
Noise and distraction Familiarity with the task Alertness / concentration / fatigue
Details of Predicted Errors
The most likely way that an observer would be able to tell that an error had occurred would be observation of an unrequired and incorrect action (i.e. an extraneous act). The following table describes how this could occur:
Extraneous Act (unintentionally taking action that is not required)
Error
Type Error Mode Is Recovery Possible?
Recovery Success Likelihood
Comments
Action Selection error (unintended
physical action) Yes Low – may not have time to intervene
Action error or incorrect positioning of a hand, or tool, results in contact with energized equipment - special consideration should be given to this task, including isolation
None of these tasks was involved in the incidents that were reviewed for this project.
6.7.4 Inspection and Maintenance in Proximity to Electrical Source List of Tasks
See list at Appendix E. The tasks in this group received a risk rating between 10 and 20 in the risk assessment (i.e. they are moderate to high risk).
5 This means that any of the error types could result in the observable error, not that all of them would occur together to produce the observable error.
Performance Shaping Factors
The tasks covered by this classification were most likely to be affected by the following factors:
Weather Lighting
Noise and distraction Familiarity with the task Alertness / concentration / fatigue
Details of Predicted Errors
As with the inspection and servicing tasks, one of the most likely ways in which errors would manifest themselves would be in extraneous acts (unrequired actions that are also incorrect). In addition to this, errors within this group of tasks are also likely to manifest themselves by operators taking more action than is required to perform the task (e.g. getting too close to live equipment) a type of error expressed as ‘action too much’.
The following tables describe how these situations could occur:
Extraneous Act (unintentionally taking action that is not required) Error Type Error Mode Is Recovery
Possible? positioning of a hand, or tool, results in contact with energized equipment - special
consideration should be given to this task, including isolation
Action Too Much
Error Type Error Mode Is Recovery Possible?
Recovery Success
Likelihood Comments
Perception No Perception Yes
Low to Moderate – more time to intervene if a worker is seen to be getting too close to energized equipment
Going too close to the energized equipment because of failure to perceive proximity to it
Two of the incidents reviewed for this project were related to one of the tasks from this group – manual vegetation clearance.
Adwick in August 2000 involved vegetation clearance but involved a violation by one of the workers rather than a human error.
The Leighton Buzzard incident of June 1985 also involved vegetation clearance, but there was insufficient information in the investigation report to determine the human factors cause.
6.7.5 Maintenance List of Tasks
See list at Appendix E. All tasks in this group were rated above 20 in the risk assessment (i.e. high risk).
Performance Shaping Factors
This was by far the largest group of tasks identified, covering a range of tasks involving maintenance work on or around electrical equipment. This includes intrusive maintenance of electrical equipment or maintenance in close proximity to energised electrical equipment. Due to the range of tasks involved the factors likely to affect the performance of those carrying out the tasks is extensive, as shown in the following list:
Time pressure; Non-standard activities;
Procedure availability/access/location; Weather;
Lighting; Temperature;
Familiarity with task; Level of experience;
Recency of training; Training quality;
Competence testing; Mentoring quality;
Alertness/concentration/fatigue; Complacency;
Team co-ordination quality; Handover/take-over;
Team maturity; Supervision;
Staff availability;
Details of Predicted Errors
Many of the tasks in this group were more complex than in the previous two groups, and therefore presented more opportunities for error. The following list describes the ways in which errors could manifest themselves in this type of task:
Action Too Much – doing more than is required, e.g. getting too close to an energised electrical source;
Extraneous Act – action that is not required and is incorrect, e.g. unintentionally touching a wrench to the energised conductor rail.
Action too Early – action that occurs at the wrong time, e.g. driving a vehicle off before all of the workers are on board;
Right Action on Wrong Object – the choice of action is correct but the selection of object is incorrect, e.g. cleaning one of several section insulators, but selecting a live one by mistake;
Action in Wrong Order – an action is conducted at the wrong point in a sequence, e.g.
unintentionally starting work before testing has been completed;
Omission – an action that should have been taken is missed out, e.g. failing to conduct live line testing prior to commencing work.
The analysis suggested that there were a number of types of error, which could lead to the occurrence of these observable errors. The following tables describe the predicted errors in more detail, including the likelihood of recovery.
Action Too Much
Error Type Error Mode Is Recovery
Possible? Comments
Perception No
Perception Yes
Low to Moderate – more time to intervene if a worker is seen to be getting too close to energized equipment
Going too close to the energized equipment because of failure to perceive
proximity to it Recovery Success
Likelihood
Extraneous Act (unintentionally taking action that is not required) Error Type Error Mode Is Recovery
Possible? Recovery Success positioning of a hand, or tool, results in contact with
Error Type Error Mode Is Recovery
Possible? Recovery Success
Likelihood Comments
Perception Misperception Yes
Low – could be difficult if the misperceived
information is credible
Memory Forget
information Yes Low – difficult for others to detect a memory failure
Accessing equipment too
Low – difficult to recover is basis of judgment is credible, also difficult for co-workers to detect a decision failure.
Making a judgment about the safety of working around live equipment, which is in some way lacking
E.g. Accessing equipment too early because equipment status information has been misperceived
Right Action on Wrong Object Error Type Error Mode Is Recovery
Possible? Recovery Success
Action taken on a live piece of equipment instead of the intended de-energised or non-electrified equipment due to perceptual confusion between pieces of equipment
on further checks by individual or detection of mistake by co-workers
of equipment instead of the intended de-energised or
Action taken on a live piece of equipment instead of the intended de-energised or
Error Type Error Mode Is Recovery
Possible? Recovery Success
Likelihood Comments
Memory Misrecall Yes Low
Misrecall of work procedure leads to a step in a procedure being taken out of sequence
Omission
Error Type Error Mode Is Recovery
Possible? Recovery Success
Likelihood Comments
Memory Late/missing action Yes
Low to Moderate – reliant on co-worker to spot the omission before harm or last minute realization by the worker
Failure to confirm isolation in place (e.g. live line testing) or check equipment status prior to commencing work. Could be due to memory failure stemming from a late or missing action, information
Failure to check of equipment status, or confirm isolation caused by a poor decision or plan, which does not include provision for standard checks, leading to work commencing with equipment energized
Twelve of the incidents reviewed for this project involved tasks that fall into this group. Of these, seven were classified as errors and the remaining five were violations. Further details of these incidents are provided below:
At East Croydon in September 2002, a worker was electrocuted whilst replacing cable tubing near the rails. The analysis of this incident indicated that this was a violation rather than an error.
At Oakley in August 2003, a lookout was electrocuted when he joined in with the replacement of insulator pots. This was found to be a violation rather than an error.
At West Croydon in October 2001, two workers were injured whilst maintaining a rail flange lubricator when an uninsulated spanner contacted the energised conductor rail. A human error analysis was performed which suggested that this was an action error (specifically a selection error because it was a physical action) and that this was most likely due to human variability in physical performance (i.e. the same person does not always perform to the same level of precision, their performance will naturally vary). This is in line with the predicted action (selection) error leading to an observed extraneous act. The most effective solutions to prevent such errors of action involve removing the hazard or physically separating the worker from the hazard. In the analysis of this incident, the recommendations covered working under T3 conditions, ensuring properly insulated tools are used, and the use of rail shields.
At Doncaster Belmont in December 2001, a worker died when he climbed on top of a wagon, presumably to check its contents. The worker was alone at the time and so the circumstances behind the accident are unclear. However, a human error analysis suggested that the most likely course of events were that the worker made a poor decision regarding climbing on top of the wagon due to a lack of knowledge. The knowledge that was most likely missing was that he could check the contents from the label (or perhaps where the label was located), and there may have been a lack of knowledge regarding the dangers of overhead lines.
Recommendations covered the provision of training on the hazards associated with AC equipment, the use of situational risk assessments, and familiarity with goods wagon labelling.
At Harlow Mill on the 5th May 2002, a worker received an electric shock when redistributing the load on top of a wagon. This was an error on behalf of the Engineering Supervisor, who had made a poor decision based on a faulty mindset. Recommendations covered effective communication training, guidance on high-risk handovers, checking procedures and audits of COSS briefings.
At Tollerton in May 2001, several workers received electric shocks during the unloading of track by a crane that fouled the overhead line during a renewals project. Investigation of the incident suggested that this was a violation rather than an error.
At Hemel Hempstead in August 2001, a worker received an electric shock when cleaning a section insulator. Two errors were identified, one on behalf of the worker and one on behalf of the person giving the briefing. One was an action error (unclear information on location of live equipment); the other was a perception error where information was misperceived due to confusion. Recommendations covered the conduct of line testing in complex areas and provision of safety communications training.
At Marston Green in July 2003, workers received electric shocks during preparation for work on the OLE. Analysis of the incident revealed that this involved a violation rather than an error.
At Ranskill in October 1998, a worker was killed when he removed the earth end of a long earth before the line end. Analysis revealed that this was a perception error stemming from a misperception that a colleague had removed the line end. Recommendations covered a one-man procedure for removing earths, effective communications training, refresher training and opportunities to develop skills, and auditing of records of working hours.
At Liverpool Street in November 1999, a worker was injured when he made contact with the OLE whilst climbing some scaffolding. Analysis suggested that this was a perception error (misperception of correct location by the contractor) caused by confusion between different locations on the station that looked similar. Recommendations covered coinciding worksite and isolation limits, and regular audits by the site safety authority.
An incident at Dock Junction in February 2002 involved workers trying to dismantle a scaffold under OLE following a shortening of the possession which meant that there was not time to complete the job safely. This was classified as a violation rather than an error.
At Handsworth in March 2002, two workers were burned when they ruptured a buried 132Kv oil-filled HV cable. The analysis suggested that this was due to an error of perception (failure to perceive information), caused by their expectations regarding what was buried at the worksite. Recommendations included planning documentation with enhanced detail and procedures for encounters with buried objects.
At Handsworth in March 2002, two workers were burned when they ruptured a buried 132Kv oil-filled HV cable. The analysis suggested that this was due to an error of perception (failure to perceive information), caused by their expectations regarding what was buried at the worksite. Recommendations included planning documentation with enhanced detail and procedures for encounters with buried objects.