Treceți la conținutul principal

Development of a risk assessment system based on pattern matching of behavioural fault models


Accidents are often a deja vu. Not learning from past accidents is leading towards new ones.The larger part of occupational accidents are caused by the mistakes of human operators ,by their fault and negligence.Behavioural fault models could describe, on the basis of accident and incident experience -the occurence of an accident by the mistakes made by the operator, putting into ballance the initial causes of the mistaked actions (lack of knowledge, try to shortcut longer but safer procedures,etc.) and the events that occured because of these causes.Behavioural fault models are perfectly developable using ontologies-and an assessment system based on past knowledge and driven by ontologies could be very usefull to judge the safety at workplace.The paper describes our research in developing such an ontological driven safety assessment system prototype and also the obtained results of running this prototype in Romanian Small and Medium Enterprises. The system starts with the building of a behavioural normal activity model-specific, for example, for the activity of work at heights. By pattern matching this model with behavioural fault models developed from past experiences and also by direct observation of workplace it could be established a quantified degree of safety
Keywords:  Behavioural fault models, risk assessment , human operator


Material loss, incidents and accidents at the workplace had almost always the same causes, the same mode of manifestation and the same pre-event warning which almost always are bypassed. Accidents are happen every day and almost nobody cares or learns something from the accident. A number of accidents happening after the same pattern in Romania have us worry if there are similar fault behaviour patterns of the human operators that were guilty of the accident occurence. If so, is there any possibility to define a human behavioural fault model which could be defined and used ? And what usage could such a model have ? In our research we found that such models are extremely interesting to be used as case studies for an efficient safety training and also by pattern matching real behaviour with the models there could be performed an efficient risk assessment by the human operator point of view.

2. Humans and loss

Information regarding work can be systematically obtained about worker dependant factors that are triggering the undesired event (being it loss, incident or accident), the physiologically and psychologically worker condition at the start/end of work .Also, it is possible to obtain information about pre-accident states, causes of accidents, conditions of their occurrence, erroneous actions and measures of prevent them in a systematic matter.[1]. As known before, humans are the main accident perpetrators. About 75% of the occupational accidents occurred in Romania [2] had peoples as the main cause of the accidents. For the rest peoples are somehow involved,even not directly.To exemplify this assertion table 1 with errors predicted in Japan is shown below.

Table 1: The most common human errors at workplaces in Japan  and their frequency
Type of error
Error of work operation
Error of decision and/or instruction
Error of judgement
Error of pattern recognition
Error of senzory organ

Human errors are given by behavioural models [3] which could start from very simple ones –active (deliberative) and reactive behaviours. The reactive behaviours- reacting at an event occurred in the work context could be categorised into four types:
-instinctive behaviour that follows a simple physical stymulus, state, reaction pattern;
-learned behaviour = instinctive behaviour  within a social context;
-drive controlled behaviour- reactive behaviour triggered by a physical need
-emotionally controlled behaviour- reactive behaviour triggered by an emotional state.
The active behaviours are defined by objectives approached by action plans [4]
These behavioural models were the base of our behavioural fault model (HBFM) which proposes a very simple and complicated question :Why and when could a person make an error at the workplace ? The next chapter shows the basics of the model.

3.The Human Behavioural Fault Model

We have built our model considering it as a calitative one which:
-could be descriptive for repetitive incidents;
-could include fuzziness for better representation of obscure incidents;
-could serve as case study against good and best practice procedures;
-are easy understandable and have distinctive causes for a certain behaviour.
The schema of the model is shown in the figure

Figure 1
Actually, there are behavioural causes till the “move to strike” performed by the worker to do the designed task.
Incidents could occur also in the training phase and in the development of specific capabilities. For example an aprentice could injure himself during the training; the same aprentice could work unguarded during the phase of capabilities development (with the welding machine, for example) and injure himself; the same aprentice if it is not stable on his feet could fell on a machine and hurt himself. 

4.Behaviour clusters-collecting and processing signs of bad behaviour at work.

Behaviour clustering  is the way to collect and process the human approach optimally. An analysis agent which automatically process behavioural clusters once collected, in the terms of human-system relationship takes into account states of knowledge and behaviour of human operators together with the system possible responsive actions. A Self-Organizing Map could be used [5] as clustering algorithm.Micro models of behaviour, repeated continously are the most important here. They include rational choice models for decision making under uncertainty and risk as well in strategic situations and in collective decision making. Models also incorporate complex assumptions like social orientations and distributional preferences.This clustering and micro-model approach is a little time consuming, not allowing a very fast assessment. However, the repetition of bad (or good) behaviours in time is the key to a safe workplace-so the time lost collecting data could be useful later.

1. K.Yoshino, Construction of human error prediction and causality model and evaluationd study of prediction characteristics, Research Material of the Central Research Institute of Electric Power Industry No. 95901,August 1995, Japan
2.Romanian Labour Inspection Statistics, 2007
3.B.Schmidt, The modelling of human behaviour, Erlangen:SCS Publications, 2000,ISBN 1-565-55211-3
4.M.L.Minsky (ed) Matter, minds and models, in Semnatic Information Processing, MIT Press, Cambridge MA, 1968
5.R.Legaspi et all, Cluster –based predictive modelling to improve pedagogic reasoning, in Computers in Human Behaviour,March  Elesvier 2007


Postări populare de pe acest blog


Analiza cauzelor rădăcină este o metodă extrem de folosită de către managementul de performanță  din firmele dezvoltate. Metoda este considerată ca o metodă primară- care trebuie utilizată în primele faze ale analizei specifice procesului managerial. Ne propunem să prezentăm o metodă de analiză a cauzelor rădăcină care să poată fi aplicată atât pentru managementul calității cât și pentru managementul securității – ținând seama de faptul că în cea mai mare parte, cauzele rădăcină ale problemelor de calitate și problemelor de securitate și sănătate sunt comune. Figura 1 prezintă modul  global de analiză pentru cauzele rădăcină Din figură se poate observa că avem 2 procese distincte: ·         -un proces de identificare- care va fi realizat pe baza metodei cunoscute și ca 5 W ( 5 Why); ·         -un proces de analiză; procesul de analiză urmărește: o   stabilirea cauzelor specifice managementului calității și managementului de SSM; o   ierarhizarea cauzelor identificate;

Figura 1  Structurare…


KPI definitionA key performance indicator(KPI) is a measure of performance, commonly used to help an organization defineand evaluate how successful it is, typically in terms of making progress towards its long-term organizational goals.
–KPIs provide business-level context to security-generated data –KPIs answer the “so what?” question –Each additional KPI indicates a step forward in program maturity –None of these KPIs draw strictly from security data
COBITControl Objectives for Information and Related Technology (COBIT) is a framework created by ISACA for information technology (IT) management and IT governance. It is a supporting toolset that allows managers to bridge the gap between control requirements, technical issues and business risks. COBIT was first released in 1996; the current version, COBIT 5, was published in 2012. Its mission is “to research, develop, publish and promote an authoritative, up-to-date, international set of generally accepted information technology control obj…


Așa după cum s-a văzut dintr-o postare trecută, graful de risc poate fi un instrument util- și nu numai în cazul bolilor profesionale. Vom adapta în continuare  teoria existentă la teoria și practica din România și vom detalia câteva aspecte considerate de interes.
Este interesant de adaptat  graful de risc pornind de la clasicul sistem Om-Mașină folosit în practica de specialitate din domeniul SSM din România. În acest sens, folosind experiența existentă și datele statistice putem dezvolta în mod corespunzător- așa cum se prezintă în continuare în acest material.
Tabelul 1- Atribute folosite în graf Atribut Descriere I (Inițiatori) Operator(O): a. operator pregătit necorespunzător[1] b.operator malevolent [2] c.operator surprins de un eveniment neprevăzut datorat sarcinii[3] d. operator surprins de un eveniment neprevăzut datorat mașinii;[4] e. operator surprins de un eveniment neprevăzut datorat mediului/factorilor naturali. [5] Sarcină(S): a. sarcină incorect formulată- care dete…