ICT-Empowered OSH Risk Management Process
- New Properties of the OSH Risk Management Process in the Context of ICT-Based Smart PPE System Implementation in the Workplace
- Generic Model of the ICT-Empowered OSH Risk Management Process
- Review of Probabilistic Methods for Context Reasoning and OSH Risk Assessment
- New Roles of Smart PPE in the Hierarchy of OSH Risk Controls
New Properties of the OSH Risk Management Process in the Context of ICT-Based Smart PPE System Implementation in the Workplace
Performing occupational risk assessments in the workplace, and then introducing appropriate protective and preventive measures to reduce the risk to an acceptable level, is widely recognised as a key process for an effective OSH management system in the enterprise (ILO 2001; OSHwiki contributors 2019), since the efficiency of this process has a direct impact on preventing accidents at work, occupational diseases and related economic losses for both workers and society. As shown in the previous section, the new functions introduced by smart PPE systems and other workplace wearables make possible faster and more effective detection (in practice, in real or near-real time) and then, as a consequence, a reduction of OSH risk factors in the working environment to an acceptable level. This applies in particular to functions such as measuring and monitoring workers’ exposure to harmful environmental factors (physical, chemical, etc.), monitoring workers’ health by measuring key physiological parameters (e.g. body temperature, heart rate, respiratory rate, etc.), monitoring work comfort (e.g. underclothing temperature and humidity, work position, etc.) and geographical location of workers with regard to potentially dangerous physical objects or high-risk zones, monitoring the current state of PPE protective properties, communicating to workers and/or safety managers about detected hazards and risks, self-adjusting of protective properties of given PPE items, and/or activation of external risk control measures. These functions allow to apply a better approach to OSH risk assessment and management processes, which so far have often been performed in a rather formal manner, which is mainly due to the obligation to meet legal requirements concerning safety and health at work.
In static workplaces, i.e. those where working conditions do not change dynamically, the risk assessment for individual workplaces is carried out in practice no more frequently than every few months, which means that the risk assessment results obtained at a given moment are considered to be averaged over a longer period of time. In addition, it is a common practice to carry out risk assessments for groups of workers, e.g. by grouping together workers exposed to the same or a similar set of harmful agents or by grouping together similar workstations where tasks are performed in the same location and under the same conditions. Therefore, information on the results of risk assessment documented in the enterprise usually indicates a level of risk, which is a generalised value for previously defined groups of workers or for sets of identical or similar jobs.
In the context of this, the use of smart networked PPE systems for OSH risk management make possible the use of this process in real or near-real time, and the personalisation of the outcomes of these process in relation to individual workers. The first of these features is of great importance due to the increasing dynamism of changes in the work environment, resulting from, inter alia, the introduction of new business models and new production concepts, e.g. within the framework of smart factories. On the other hand, the personalisation of OSH risk management makes it possible to focus on the safety and health conditions of individual workers who, being exposed at a given moment to various factors of the working environment, can be protected against the impact thereof in a differentiated manner that is adapted to their needs depending on the type of hazard and the current level of exposure.
Generic Model of the ICT-Empowered OSH Risk Management Process
Taking into account the functions of smart networked PPE systems as well as the concept of integration of these systems with other networks and systems of the industrial IoT, especially in the context of the development of Industry 4.0 concepts, the author proposed a graphic representation of a general model of a cyber-physical system supporting OSH risk management process in smart working environments, which was originally published in Podgorski et al. (2017). Figure 6.4 shows a revised version of this model, which demonstrates the basic functionalities of the system and takes into account new aspects resulting from the development of smart PPE systems, workplace wearables and the industrial IoT.
The hardware layer of the presented model consists of sensors and actuator networks embedded in workplace equipment and smart PPE and wearables worn by workers, as well as other sensors and actuators embedded in machinery and other workplace objects and facilities. Whereas the software layer can be divided into four main parts: (1) a contextual database containing historical and current measurement data collected from all sensors integrated or associated with different physical objects; (2) data fusion and context-aware aggregation module; (3) an inference engine responsible for analysing contextual data and calculating and allocating risk levels to individual workers in real time; and (4) a risk control manager that is responsible for analysing the resources available to control risk, selecting the appropriate preventive and protective measures in view of their functions and potential effectiveness, activating those measures in due time, and monitoring their use.
The proposed model points to the crucial role of context-awareness in ICT- supported processes of hazard detection and risk assessment. The concept of context used for ICT applications can be defined as “any information that can be used to characterise the situation of an entity. An entity is a person, place or object that is considered relevant to the interaction between a user and an application, including the user and applications themselves” (Dey 2001). Whereas context-awareness
FIGURE 6.4 Generic model of a cyber-physical system applied for OSH risk management in smart working environments (modified from Podgorski D. et al. . Towards a conceptual framework of OSH risk management in smart working environments based on smart PPE, ambient intelligence and the Internet of Things technologies. International Journal of Occupational Safety and Ergonomics, 23(1): 1-20)
and context-based reasoning are basic features and functionalities of cyber-physical systems implemented in smart environments facilitated by IoT technologies (Hong et al. 2009; Bettini et al. 2010; Ye et al. 2012). This applies in particular to environments characterised by frequent and dynamic changes that significantly affect the behaviour of users and their interactions with other objects in those environments.
Review of Probabilistic Methods for Context Reasoning and OSH Risk Assessment
As it has already been mentioned in section 6.3.1, the most significant functionality of the ICT-based smart networked PPE systems is the capability of real-time and personalised assessment of occupational risks, taking into account all hazardous factors of a dynamically changing production environment. There are many methods and techniques of work-related risk assessment that have been developed so far - see for example reviews by Marhavilas et al. (2011) and Wang et al. (2015). In general, risk assessment techniques can be divided into two types consisting respectively in the deterministic and the probabilistic (stochastic) approaches (Kirchsteiger 1999; Marhavilas and Koulouriotis 2012). Next, deterministic techniques can by grouped in three main categories: (a) qualitative, (b) quantitative, and (c) hybrid techniques (qualitative-quantitative, semi-quantitative), while stochastic methods are roughly divided into the methods based on two categories: (a) the classic statistical approach, and (b) accident forecasting modelling (Marhavilas and Koulouriotis 2012). It is also claimed that the application of a single risk assessment method will not ensure sufficiently reliable results, so it is advisable to use mixed approaches based on both deterministic and probabilistic methods. At the same time an opinion should be taken into account that probabilistic methods may have an advantage over the deterministic ones, because the former are more cost-effective and their results are easier to communicate to users and other stakeholders (Kirchsteiger 1999).
The literature concerning ICT applications in the field of smart environments and pervasive computing reveal that in the case of relatively simple systems, which consist of up to several sensors and actuators (e.g. location tracking and warning system for workers operating in dangerous zones), the methods for risk assessment and reasoning are usually rule-based and have a deterministic nature. One may find a few examples of such applications in papers by Giretti et al. (2009), Teiser et al. (2010), Yang et al. (2011), Ku and Park (2013), and Fugini and Teimourika (2014). However, in the case of more advanced cyber-physical systems that consist of many sensors monitoring various parameters of users’ health and the environment, and which are used for managing complex scenarios of users’ activities and their relations w'ith smart objects, the use of probabilistic methods or the use of methods based on fuzzy logic is needed (Ranganathan et al. 2004; Haghighi et al. 2008; Singla et al. 2009). In particular this approach is necessary in scenarios where a certain level of uncertainty occurs, which is related to, for example, the recognition and prediction of user behaviour, inaccuracy of sensors, missing information, imperfect observations and inferring on the basis of imprecise and conflicting data. In such cases literature often indicates the practical usefulness of such methods as Baeysian networks, Hidden Markov Models and Dempster-Shafer theory of evidence (Tolstikov et al. 2011; Ye et al. 2012; Perera et al. 2013).
Bayesian networks (BNs) are used to model cause-and-effect relationships between random variables and their conditional dependencies, thus providing a concise representation of their joint probability distribution. Modelling the dependency is carried out by constructing a directed acyclic graph, where nodes represent random variables and the edges correspond to relationships between these variables. BNs have found many practical applications in many domains of science, economy and life, which is reflected in some reviews, e.g. by Kenett (2012) and Weber et al. (2012). Since the BNs are a good tool to calculate the probability of occurrence of various inter-related random events, they have become a frequently used tool for risk estimation, including the risk associated with work (e.g. Leu and Chang
2013). They can also be used for other cause-and-effect analysis in the framew'ork of OSH management, such as analysis of accidents caused by falls from a height
(Martin et al. 2009), or in investigations on relationship between hygiene conditions, ergonomic conditions, job demands, physical symptoms, psychological symptoms, and occupational accidents (Garcia-Herrero et al. 2012). Examples of existing BN application for context reasoning in smart environments include, among others, environmental situation recognition in wireless sensor networks applied for outdoor environment monitoring (Bagula et al. 2010), context reasoning for fall risk assessment of elderly in ambient assisted living (Koshmak et al. 2014), and a real-time monitoring and early warning system that safeguards the well-being of workers exposed to heat stress (Liu and Wang 2017).
Hidden Markov Models (HMM) are a statistical tool in which it is assumed that the modelled system is represented as a Markov chain with hidden states, but with the visible outputs, which are dependent on these states. In turn, a Markov chain is a model of a stochastic process constituting a series of events in which the probability of each event depends only on the outcome of the preceding event. HMM, similarly to BNs, have wide applications in many various domains of science and life, as for example in biology (Choo et al. 2004), speech recognition (Gales and Young 2007), computer vision and pattern recognition (Fink 2014), and many others. In the area of context inferring in smart environments HMM have been applied among others for modelling human behaviour in smart hospitals (Sanchez et al. 2008), and smart homes (Bruckner and Velik 2010; Chahuara et al. 2013). Examples of practical HMM applications in the working environment include location tracking and activity analysis of construction workers (Khosrowpour et al. 2014), predicting the likelihood of work-related musculoskeletal hazards among dental students (Thanathornwong et al.
2014), and real-time location tracking, trajectory prediction, and prevention of potential collisions between workers and construction site objects (Rashid et al. 2018).
The third of the above-mentioned probabilistic method is the Dempster-Shafer theory of evidence (DST), also referred to as the theory of belief functions. DST was initially developed by Dempster (1966), and was then extended, refined and recast by Shafer (1976). This method is regarded as a generalisation of the Bayesian theory of subjective probability, since in traditional probability theory evidence is associated with only one possible event, while in DST, evidence can be associated with multiple possible events. In general, DST consists of the calculation of what is known as the “belief function” in order to obtain degrees of belief by combining all available information (evidence) from different sources. DST has been used to solve problems of processing incomplete or uncertain information in business decision making (Srivastava and Liu 2003) and in many engineering disciplines, such as computer science, construction and production management. In the field of smart environments DST is often used to support data fusion and/or for context reasoning and human behaviour recognition in smart homes (e.g. Zhang et al. 2010; Sebbak et al. 2013), while examples of applying DST for risk assessment and safety management have been presented by, for example, Rakowsky (2007), Zhang et al. (2014), and Nesculescu et al. (2015).
New Roles of Smart PPE in the Hierarchy of OSH Risk Controls
The use of new functions of smart PPE and wearables in combination with ICT applications to support real-time OSH risk management throws a slightly different light on the role of these devices in relation to the classical hierarchy of risk controls, which has been outlined in section 6.2.1. In literature, this hierarchy is often presented graphically, in the form of an inverted pyramid illustrating the relative effectiveness of individual measures for accident prevention and the protection workers’ health. This approach comes from several decades ago and thus reflects a static and periodic approach to risk management resulting from the technological capabilities and organisational practices available in this field at that time. Since the situation in this area in the last decade has changed significantly, mainly due to the development and implementation of the Industry 4.0 concepts and dynamic progress in the development of smart PPE, sensor technologies and wireless ICT networks, it is appropriate to consider the new interpretation of this hierarchy by taking into account the new functions offered by smart PPE systems.
According to the discussed hierarchy elimination and substitution are considered to be the most effective and at the same time the most difficult risk control measures to be applied in existing work processes. These measures are justified mainly at the stage of designing and developing a technology process, a workplace or a service, during which it is still possible to introduce relatively easy and low-cost changes consisting of the elimination of hazard sources or replacement of materials generating hazards with low-emission materials or materials of chemical composition safe for humans. However, in the case of already introduced technological processes, the introduction of risk controls such as elimination or substitution may be unrealistic or economically unjustified, as it may mean the necessity to stop the production process for a longer period of time in order to completely redesign it. Similar difficulties are also to be considered in the case of designing work processes, which are introduced within the framework of digitally transformed smart factories. Therefore, these two highest levels of risk controls, namely elimination and substitution, can also be considered adequate in the case of ICT-based OSH risk management, with the reservation, however, that these activities should be taken into account primarily at the stage of designing new processes.
In traditional non-smart workplaces, the interpretation of the priorities for risk control measures belonging to lower levels of the hierarchy is quite clear. Engineering controls (referred to also as collective protective equipment) should be used first to separate workers from exposure to hazards. If, despite these actions, it is not possible to reduce risk to an acceptable level, administrative controls should be applied (e.g. changes in work procedures, limiting working time at selected positions, additional training or instructing workers, providing warnings to workers on existing hazards, etc.). And only if these earlier actions did not bring the desired effects, should workers be provided with PPE (as so-called the “last resort”), provided that the respective PPE items are appropriately selected and matched to the types of hazards and risks occurring at their workplace.
Such hierarchy of protective measures should be respected in traditional workplaces at the stage of designing workstations and processes, and at the operational stage at the time of planning risk control measures, e.g. those resulting from periodic or ad-hoc risk assessments. However, in workplaces where there are frequent and significant changes in the exposure of workers to various types of risks, strict compliance with these principles may lead to selection of less effective risk controls,
FIGURE 6.5 New functions of smart PPE systems in relation to the classical hierarchy of OSH risk controls (adapted from Podgorski D. . Functions of smart PPE in relation to the Hierarchy of Risk Controls, Linkedln. Posted September 11, 2017) where other, readily available functions offered by smart PPE systems would provide a better level of protection. In such cases, it seems appropriate to consider other priorities for protective measures in line with the interpretation of the risk control hierarchy presented in Figure 6.5.
The proposed interpretation of the hierarchy of risk controls means that if an employer has an access to smart PPE or other wearable devices that have novel functions allowing to control given risks more effectively, e.g. by initiating appropriate engineering controls or performing various types of administrative controls, the use of these smart devices may have a higher priority comparing to the case of traditional non-smart PPE, which should always be considered as the “last resort”.