Time-related ergonomics evaluation for DHMs: a literature review

Transcription

1 356 Int. J. Human Factors Modelling and Simulation, Vol. 1, No. 4, 2010 Time-related ergonomics evaluation for DHMs: a literature review Cecilia Berlin* Department of Product and Production Development, Div. of Production Systems, Chalmers University of Technology, SE , Göteborg, Sweden cecilia.berlin@chalmers.se *Corresponding author Tara Kajaks Department of Kinesiology, McMaster University, 1280 Main Street West, Hamilton, ON, L8S 4K1, Canada kajakst@mcmaster.ca Abstract: Ergonomics problems in production systems are of a multi-causal nature. It has been established in ergonomics literature that time-related factors, including activity duration, repetitiveness, work-rest distribution and muscle reactions to dynamic loads, can influence the occurrence of work-related musculoskeletal disorders (MSDs). In recent years, ergonomic practices have evolved to include the use of digital human models (DHMs) in virtual workstations, resulting in more cost-efficient and proactive evaluations. However, the ergonomic tools provided in DHMs often fail to consider time-related ergonomic factors. This literature review compiles and examines time-related ergonomics terms for the benefit of introducing such concepts into DHMs. The influence of time-scale perspectives and ambiguities regarding how terms have been used are also discussed. Developers of DHMs can benefit immensely from a literary overview of how to consider time-related factors of physical workload. Likewise, the scientific community can benefit from the identification of ambiguities and gaps in ergonomics research. Keywords: ergonomics evaluation; time aspects; dynamic evaluation; digital human modelling; DHM; proactive ergonomics; developers; manufacturing ergonomics; ergonomics simulation; musculoskeletal disorders; cumulative loading; human factors. Reference to this paper should be made as follows: Berlin, C. and Kajaks, T. (2010) Time-related ergonomics evaluation for DHMs: a literature review, Int. J. Human Factors Modelling and Simulation, Vol. 1, No. 4, pp Biographical notes: Cecilia Berlin is a PhD student at Chalmers University of Technology in Göteborg, Sweden, at the Department of Product and Production Development, Division of Production Systems. Her area of research concerns the use of digital human modelling and other ergonomics tools in production systems and organisational contexts. She was involved for three years in the project 4D-ergonomics and received her Licentiate of Engineering in 2009 on the subject of production ergonomics evaluation. She received her MSc in Industrial Design Engineering at Chalmers University of Technology in Copyright 2010 Inderscience Enterprises Ltd.

2 Time-related ergonomics evaluation for DHMs 357 Tara Kajaks is a PhD student in the Department of Kinesiology at McMaster University, Hamilton, ON, Canada. Her area of research is in occupational biomechanics and ergonomics, where she studies the use of digital human modelling in proactive ergonomic practices. She received her MSc in Kinesiology from Queen s University in There she investigated occupational risk factors for knee osteoarthritis. 1 Introduction Ergonomics evaluation of production systems (especially manual assembly) is chiefly geared towards identifying risks for MSDs. Due to the fact that there is much well-accepted research regarding evaluation of body postures, the main focus of most ergonomic analysis tools is to identify isolate postures that are considered extreme, unbalanced or physically strenuous. Most analysis methods assume that every posture can be considered as an inert (static) load situation. However, posture and static loading are not the only culprits behind MSDs. Time-related aspects of work behaviour, such as repetitiveness, monotony, and durations of exposure, contribute at least as much, if not more, to the potential risk for work-related injury. A need to shift focus away from pure posture analysis in favour of more time-related ergonomics analysis has been identified (Chaffin, 2005; Chiang et al., 2008; Wells et al., 2007). The ergonomics analysis tools available in most digital human modelling (DHM) software reflect the status of the research field of today; the most widespread DHM software developers to date (Bubb et al. 2006; Laring et al., 2005; LaFiandra, 2009) chiefly provide posture-based analysis tools that assess MSD risks from a static loading perspective (Lämkull et al., 2009). With recent developments in motion capture technology, motion databases and more detailed manikins, movement simulations in DHMs are becoming more relevant and visually convincing. However, the analysis tools offered in most commercial software do not allow for the consideration of time-related components of workload exposure that contribute to MSDs. As DHMs evolve from purely static manikins to more sophisticated models capable of dynamic movements, the inclusion of time-related exposure methods is also needed. The use of DHMs for proactive ergonomic applications is becoming more common as organisations are becoming more aware of the cost-saving and injury prevention benefits associated with such tools (Brazier et al., 2003). Thus, as the field of ergonomics grows to incorporate more advanced time-related exposure methods into mainstream ergonomic practices, these tools must also be built into DHMs. A review of common DHMs is presented by LaFiandra (2008). Examples of such software include Safework Pro and Delmia Human, Jack, AnyBody Technology and Santos. Safework Pro and Delmia Human offer basic static postural ergonomic tools [the NIOSH equation (NIOSH WPG, 1981) and RULA (McAtamney and Corlett, 1993)] as well as Garg s metabolic equation (Garg et al., 1978). Jack is more sophisticated in that it contains a high-level functioning manikin skeleton that allows for many degrees of freedom as well as the calculation of internal joint torques. In addition to the same outputs for Delmia Human, Jack also contains the University of Michigan s 3D static strength prediction program module (Chaffin, 1997). While Jack has the capability of streaming real-time motion data, the analyses must always be performed on static

3 358 C. Berlin and T. Kajaks postures, thus failing to accurately account for time-varying exposures. AnyBody Technology, however, does drive dynamic models and is capable of estimating muscle recruitment patterns using inverse dynamics optimisation techniques (Jung et al., 2009), although the accuracy of these patterns remains to be determined (LaFiandra, 2008). One of the most advanced DHMs used for ergonomic applications is Santos, due to its optimisation-based motion prediction capabilities. This capability makes Santos one of the most likely candidates to develop more advanced time-varying ergonomic exposures (Abdel-Malek et al., 2006). The authors of this article propose that to fully utilise the potential of DHM software with animation capabilities, developers should provide analysis methods that consider the time-related aspects of physical loading during work. Many ergonomics researchers have made advancements in conceptualising time-related factors, in some cases presenting ways to quantify them. However, there is a great diversity of conceptualisations of time-related factors. Therefore, to provide guidance to DHM software developers, this literature review contributes an overview of how time-related ergonomics exposure factors have been addressed and how they may interact. 2 Methods 2.1 Data collection The material for this literature study has been collected primarily via database searches, using a number of combined key terms such as time, ergonomics evaluation, dynamic, repetitive and cumulative. Most sources were found in databases Scopus, Ingentaconnect, Google Scholar, Inderscience, ScienceDirect and similar services. It was decided that emphasis should be placed on literature from 1990 and onwards, since it was presumed by the authors that literature from this limited period would contain more DHM-relevant contributions. If deemed within the scope of interest, additional material was obtained from the reference lists of selected contributions. 2.2 Analysis The contributions were screened after an initial read-through so that the final material was comprised of case studies, literature reviews, experimental descriptions, doctoral theses, proposals of terminology definitions and time-related practitioner methods. Once established, the material was subjected to categorisation using a hermeneutical approach (Smith, 1998). A number of guides for scientific literature reviews were consulted and the one described by Psychology Writing Center (2005) was chiefly adhered to. 3 Results The results of this literature review provide two components; a review of terminology (Section 3.1), and a review of methods (Section 3.2). The selected material highlights a number of recurring concepts which describe how different mechanisms contributing to MSDs vary over time.

4 Time-related ergonomics evaluation for DHMs 359 Three literature reviews offer a particularly noteworthy introduction to the problems associated with time-related ergonomics evaluation. First, Kilbom s (1994) review defines the measurable parameters necessary to describe repetitive work, with a pronounced focus on exposure-effect relationships. The major contribution is Kilbom s comprehensive compilation of definitions for the term repetitive and suggested quantifications of it, mostly in terms of cycle times. Later, Li and Buckle (1999) reviewed posture-based methods and the ability of some to perform dynamic recordings of movement in real time. They note that besides posture, other risk factors such as force, frequency and/or repetitiveness of movement, task duration etc. are believed to be important contributors to MSDs, although little is known about the relative importance of each factor. Li and Buckle also emphasise that different risk assessment tools have been developed with very specific types of work in mind and that they can be used wrongly if applied to a different situation. Finally, Wells et al. (2007, p.741) state that the time-related concepts established by ergonomics literature map poorly onto the known quantitative risk factors due to the fact that they lack operational definition. This implies that there is a loss in translation between ergonomists and their engineering counterparts in a design process, because the two groups conceptualise differently. For example, ergonomists aim to minimise MSDs by increasing process variance, thus reducing the likelihood of adopting potentially injurious static postures and excessive cumulative loading. However, engineers seek to minimise process variance in order to design more efficient production systems. Thus, there is a disconnect between the aims of ergonomists and engineers that is driven by modern-day industrial goals, conditions, and tendencies in production systems. Additionally, Wells et al. suggest that modern-day conditions and tendencies in production systems will cause lack-of-variation to surpass peak loads and extreme postures in relative importance as a biomechanical risk factor for MSDs (2007). Their review identifies a number of common concepts, variables and metrics that could facilitate the exchange of time-related information between engineers and ergonomists. 3.1 Time-related evaluation terminology Several authors reviewed here (Kazmierczak et al., 2007; Kilbom, 1994; Wells et al., 2007) have independently brought up the point that time-related terms are used and interpreted differently by different stakeholders depending on the context in which the terms are used and the stakeholder s professional discipline. As shown by this review, several time-related terms used by the research community lack uniform definition. Many concepts have been explored by different lines of research with an unstated, implicit definition. How successfully each concept embodies risk factors is therefore not easy to judge, since inconsequential definitions can lead to losses in translation between literary contributions. Many definitions are closely associated with criteria limits for acceptability, which in turn may be specific to different types of physical work. Kilbom (1994) illustrates this dilemma by presenting tables with numerous definitions and quantitative definition limits of repetitive work. These ambiguities aside, many of the found time-related terms and concepts share certain characteristics, which can be the basis for categorising the terms into functional groups, allowing DHM developers to use terms that are internally compatible in their analysis tool development.

5 360 C. Berlin and T. Kajaks Table 1 Time scales of mechanisms triggering MSD risks Terms Contributor(s) Discomfort Kee and Karwowski (2001) and Rose (2001) Lack-of-variation/ movement sameness Moore and Wells (1992), Neumann et al. (2006) and Wells et al. (2007) Fractions of seconds Endurance time Rose (2001) and Rose et al. (2001) * Fatigue Rose (2001) * Resumption time Neumann et al. (2006) and Rose (2001) * Angular velocity Juul-Kristensen et al.(1997), Marras (1992) and Wells et al. (2007) Angular acceleration Juul-Kristensen et al.(1997), Marras (1992) and Wells et al. (2007) Muscle tenderness Gilad (1995) * Nerve entrapment Gilad (1995) * Acute peak loads Norman et al. (1998) * * * * Exposure variation Veiersted et al. (1990) and Wells et al. (2007) Spinal disc shear and compression Cumulative loading leading to MSDs Time increments (scale and units according to review) Seconds minutes Minutes Hours Weeks Seasons Years Decades EMG gap Action/sub-task Cycle Task Job Work shift Working life * * * * * * * * * * * * Kumar (1990) * * * * * * * Seidler et al. (2001) * * * * * Notes: Cumulative trauma disorder (CTD), repetitive strain injury (RSI), lower back disorder (LBD).

6 Time-related ergonomics evaluation for DHMs 361 Table 2 Definitions of job descriptions associated with time-related physical load exposures Terms Contributor(s) Definition Measure of time-related MSD risk Task Colombini (1998) A specific working activity whose objective is the attainment of a specific operational result. Norman et al. (1998) Jobs are divided into tasks for the purpose of calculating accumulated load (as the peak load of each task x its duration). Kilbom (1994) A task should be defined in terms of the parameters static loads, external force and posture. The engaged body region and duration of exposure should be specified. Cycle Colombini (1998) A sequence of technical, mainly mechanical, actions of relatively short duration, that repeats itself over and over, always the same. Work cycle Gilad (1995) Composed of short, frequent motion elements, non-frequent and forceful motion elements and should be broken down into sub-activities consisting of tasks. Frequency Repetitive movements Accumulated load Body-segment specific repetitive loading Repetitive movements Frequent and forceful motion elements Moore and Wells (2005) The inverse of cycle time (i.e., increased frequency = decreased cycle time). - Moore and Garg (1995) Two approaches: number of efforts per minute, which is rated according to criteria levels, and speed of work which is rated subjectively (on an ordinal scale) based on observation. This definition is specific to the strain index model. Speed of work Bao et al. (2006), Escorpizio and Moore (2007), Kilbom (1994) and Moore and Garg (1995) Rarely uniformly quantified in literature, although it is sometimes mentioned as a risk component. Action frequency Colombini (1998) Measured as the number of (mechanical) actions per time unit (minute, cycle or shift). Duty cycle Moore and Wells (2005) The proportion of stipulated work cycle time actively working, expressed as a percentage of cycle time. Frequent motions, discomfort - Frequent movements Movements and pauses Bao et al. (2006) The percentage of a cycle spent in exertion. Exertion Mathiassen and Winkel (1991) The original definition of duty cycle is only applicable to work oscillating between two load levels, although the term can be expanded to state the relative distribution of load over time. Muscle load

7 362 C. Berlin and T. Kajaks A first useful categorisation is to show the time scale (i.e., units of time length) associated with the MSD-contributing mechanisms (Table 1). The time scale varies in the literature, from very small increments such as gaps (fractions of seconds) in electromyography (EMG) recordings, to the span of an entire working life (years). Some mechanisms have been studied across a range of time increments, while others are particularly associated with one specific time unit. The terms described in Table 2 can be classified by time-related concepts, including: job description, movement characteristics and consequences that affect performance (Figure 1). These concepts are described in greater detail in the following sections. Figure 1 A functional hierarchy of time-related concepts Job description The job description includes some basic time-related inputs that are significant contributors to MSD risks. Terms used to describe jobs are often interrelated and therefore require a clear definition prior to ensuring their proper usage. For instance, the relationship between job, task and cycle needs to be clear in each description. This brings up the meaning of related terms such as frequency, speed of work and duty cycle. A summary of these terms is presented in Table 2. What constitutes a task has been conceptualised in different ways. Broadly speaking, a task can be a measure of accumulated load (Norman et al., 1998), repetitive movements (Colombini, 1998), or repetitive loading (Kilbom, 1994). It is important to note, however,

8 Time-related ergonomics evaluation for DHMs 363 that Kilbom s (1994) definition of task implies that it is clear when a task activity begins and ends. This definition is focused on identifying repetitive work, and emphasises the notion that the repetitive work is concentrated to a specific body segment. In other circumstances, Dempsey and Mathiassen (2006) assert that since modern work is very complex (e.g., in the presence of job rotation, seasonal variations or flexible manufacturing strategies), the usefulness of classical task analysis techniques may be limited in the context of ergonomics evaluation. They also state that such approaches may be useful for identifying peak loads for individual tasks, but are not as successful when cumulative exposures or exposure variation patterns are in focus. Paired with the term task at the lowest level of workplace activities is the term cycle; however, there is ambiguity as to which term is most appropriate for the lowest-level activity that constitutes work. In Colombini (1998), the concept of repetitiveness has been incorporated into the definition of a cycle, stating that tasks consisting of cycles are by definition repetitive. In contrast, tasks are also considered as sub-activities of larger scale motion elements termed work cycles (Gilad, 1995; Laring et al. 2005). Figure 2 Example of a combined breakdown principle as suggested in the literature Similarly, the term frequency holds several definitions. In its most basic form, frequency is defined as the inverse of cycle time (Moore and Wells, 2005). However, Colombini (1998, p.1268) brings up the term action frequency, a more nuanced term that captures situations where workers can voluntarily choose to speed up or slow down their working rhythm, e.g., when producing a fixed number of pieces per shift. The concept of frequency is further complicated by terms such as efforts per minute, speed of work and duty cycle. These first two terms are either rated according to criteria levels or subjectively rated for their inclusion in tools such as the Strain Index (Bao et al., 2009;

9 364 C. Berlin and T. Kajaks Moore and Garg, 1995). In contrast, duty cycle has been used to quantify either the percentage of time spent in exertion (Bao et al., 2006) or the work time spent oscillating between two load levels (Mathiassen and Winkel, 1991). Bao et al. (2009) summarise the situation quite succinctly by stating that the breakdown terms, job, task, work element and exertion, have all been defined inconsistently in literature, and thus it is important in each individual case that each respective line of research clearly defines the terms and how they relate to each other. As an example of a breakdown hierarchy, Bao et al. (2009) describe that during a shift, a job is performed, which can be broken down into several tasks, and from there into several work elements, which are each characterised by exertions. Further characterisations as described by other contributors can add to a clearer definition of this breakdown hierarchy, as suggested in Figure Movement characteristics A DHM needs to be able to generate a model of the work sequence based on the input, and then must be able to visualise and evaluate what is happening. Therefore, there must be time-related, quantitative criteria of acceptability built into the tool. What the DHM can do is to visualise the time-related work characteristics in such a way that the user is made aware of their role as risk factors that may impede on good work performance. What is meant by dynamic work must first be established, as must the related terms monotony and variation. Table 3 summarises terms which imply that there is a variation of motion over time. Dynamic movement is closely related in many contributions to movements of the body s joints, which can be discussed in terms of angular velocity and angular acceleration. In their review of classification criteria for joint angles, Juul-Kristensen et al. (1997) suggest that angular velocity and acceleration combined with force are more important as risk factors than posture. According to Marras (1992), body segments may differ in sensitivity to the two movement components such that, for example, the trunk is more sensitive to angular velocity, while the wrist is more sensitive to angular acceleration. From a production system perspective, Wells et al. (2007) point out that angular velocities and accelerations are closely related to the human operators interactions with the production system and are therefore not so predictable. Some movement-related terms are descriptors of undesirable conditions, such as the concept of monotonous work (see repetitiveness, see Table 4). Both definitions quoted in Table 3 are based on EMG signals and relate monotony to the repeated use of the same muscular structures. An antonym to the above expression is variation, implying that actions vary to such a degree that the strain of one particular muscular load is relieved by switching to other activities, even if they do not involve rest by design. It is commonly suggested that mechanical exposure variation can be used as an intervention strategy against MSDs (Mathiassen, 2006; Möller et al., 2004). However, Wells et al. (2007) argue that little consensus exists regarding metrics for variation, largely due to the fact that variation of exposure can be measured over very different time scales ranging from very short (e.g., EMG gaps < 1 s) to very long (e.g., variations over work seasons) see Table 1. The term exposure variability sometimes also appears in the literature (Mathiassen, 2006; Möller et al., 2004). In general, this term signifies a descriptive statistical measure of biomechanical exposure variation and appears frequently in studies of occupational epidemiology.

10 Time-related ergonomics evaluation for DHMs 365 Table 3 Definitions of movement characteristics associated with time-related physical load exposure Terms Contributor(s) Definition Measure of time-related MSD risk Dynamic Angular velocity/angular acceleration Monotony/monotonous work Variation/exposure variability Marras (1992) How motion influences force exertion May increase greatly the predicted loading experienced by a joint due to increased muscular co-activation. Implies that body segments are subject to motion velocities, in combination with flexion, torques and lift rate. Motion and exertion Kilbom (1994) Movements around a joint are easily distinguished. Repetitive (joint) motion Grant (1994) Implies that grip force requirements vary during motion, resulting in a peak hand* exertion at some point. Högberg et al. (2007) Dynamic ergonomics evaluation (in a DHM) signifies animated work simulations, where the objective of the DHM software is to compute aggregated loads over entire work sequences. Juul-Kristensen et al. (1997), Marras (1992) and Wells et al. (2007) Measure of angle flexion/extension over time around a joint. The authors point out different caveats regarding differing body segment sensitivity to these components and the difficulty in predicting the components in human-machine interactions. Moore and Wells (1992) High autocorrelation of posture or EMG time history. Mathiassen et al. (2003) Low within work task variance, low between task variance of EMG or posture. Mathiassen (2006) Dispersion within days within subject, between days within subject, between subjects, between tasks, between jobs. (Hand) exertions Cumulative loading Motion and force round joints Muscle use Sameness of motions Möller et al. (2004) A cycle-to-cycle statistical variance of exposure parameters level, frequency and duration. Note: *Description relates to grip force requirements during manual handling of loads.

11 366 C. Berlin and T. Kajaks Table 4 Definitions of repetitiveness* associated with time-related physical load exposures Contributor(s) Definition or comment Physiological Moore and Wells (1992) Should be defined in terms of amount of tissue movement, cycle time and estimate of sameness as they relate to posture. Should be identified as repeated or sustained applications of force. Lack of movement ( postural fixity ) is a special case of repetitiveness. Kilbom (1994) Involves frequent repetition of physically similar work cycles. Escorpizio and Moore (2007) A task involving near-identical muscular movement similarity under the same external load. Gilad (1995) Emphasises the contribution of jobs which are frequent in [...] appearance, short in duration and cyclic in performance to acquiring cumulative trauma disorders. Characterised early on by lack of recovery time, muscle tenderness/overuse and nerve entrapment. Occhipinti (1998) Linked to pause distribution and duration, since high-frequency actions (exceeding 40 actions per minute) necessarily shorten the time available for contraction and decontraction of muscles. Task-based Gilad (1995) Closely coupled with the term cycle and cyclic; in a hierarchical breakdown of jobs into tasks and elements, Gilad distinguishes between non-cyclic and cyclic work elements. Colombini (1998) and Occhipinti (1998) Both authors define the term cycle as being a repetitive occurrence per se. Quantifications** Fallentin et al. (2001) Number of movements of joints in a unit time. Kilbom (1994) Compiles numerous criteria definitions of repetitiveness, expressed in cycle time lengths. The majority of these propose the threshold value <30 seconds. Many of the criteria in Kilbom s (1994) table and/or article are used by other contributors (Gilad, 1995; Colombini, 1998; Occhipinti, 1998; Möller et al., 2004). Colombini (1998) and Occhipinti (1998) Quantified in terms of technical actions, stating that more than 30 technical actions per minute constitutes repetitive work. Bao et al. (2006) Two-part conditional definition of repetitive work (based on Silverstein et al., 1987 and Keyserling et al., 1993): basic cycle times are <30 seconds and/or more than 50% of the job consists of similar upper extremity motion patterns. Suggest a multi-factorial quantification of repetition in terms of frequency and duty cycle of hand exertion, repetitive muscle activity (classified by time study methods), hand activity level, duration of exertion, number of efforts per min and speed of work. It should be noted that this definition is specific to forceful hand activity. Notes: *Also referred to as repetition, repetitivity, repetitive work etc. in literature. **For the purpose of establishing acceptability criteria.

12 Time-related ergonomics evaluation for DHMs 367 The most widely discussed movement characteristic term is repetitiveness, which has been summarised in a separate table (Table 4). There are chiefly three definition types present in the studied literature. Physiological definitions of repetitiveness focus predominantly on use of muscular or skeletal structures. Task-related definitions of repetitiveness also occur; they assume that work can be divided into identifiable sub-units. Quantifications of repetitive work have been attempted numerous times in order to establish acceptability limits (assuming that repetitive work is a risk factor for MSDs) Consequences that influence performance The terms in this section deal with time-related consequences of physical loading that may influence work performance, such as fatigue, endurance, discomfort, and pain (Table 5). Each of these terms describes either a consequence or a coping mechanism for time-related ergonomic issues in the workplace. At the end of this section, a discussion of some strategies to avoid these risk factors (as found in the literature) has been included. Fatigue, defined as the inability to maintain the required force or work output level (Ma et al., 2008a), can have a significant effect on work performance and injury risks. Muscular fatigue has been shown to lead to both chronic MSDs, such as repetitive strain and overuse injuries, as well as acute injuries (Kilbom, 1994). Attempts have been made to understand the time-dependent process of fatigue by predicting muscular fatigue through various sophisticated theoretical models (e.g., Freund and Takala, 2001; Liu et al., 2002; Xia and Frey Law, 2008) and EMG processing techniques (e.g., Cifrek et al., 2009; Dimitrova and Dimitrov, 2003). However, the most common method of quantifying fatigue for ergonomic applications is through the use of decaying power or exponential equations, such as Rohmert s equation (Rohmert, 1960), which typically model endurance time as a function of exertion intensity. Despite the simplicity of using such equations to predict both endurance time and recovery time, a recent meta-analysis of fatigue models by Frey Law and Avin (2010) has shown how complex it is to accurately predict muscle fatigue by demonstrating that currently no unique model exists to predict endurance time across all joints. This finding supports that scientists have faced a challenge in developing and implementing fatigue models for DHMs; however, attempts are being made to extend theoretical fatigue models to DHMs (e.g., Ma et al., 2008b, 2009). Related to fatigue is the concept of endurance, or the ability to sustain a physical effort. Evidence suggests that work experience influences endurance as well as resumption time after fatigue has set in; experienced workers have longer endurance times and shorter resumption time than inexperienced workers (Rose et al., 2001). This training effect is also supported by Grant et al. (1996). However, when loading was resumed directly after rest, Rose et al. (2001) report that the second endurance time was shorter and the ensuing (voluntary) time to recovery was longer. Interestingly, they conclude that endurance time is not influenced by loading occurring at the extremes of the worker s range of motion.

13 368 C. Berlin and T. Kajaks Table 5 Definitions of consequences that influence performance associated with time-related physical load exposures Terms Contributor(s) Definition Fatigue Ma et al. (2008a) The point at which the muscle is no longer able to sustain the required force or work output level. Konz (1998a) Appears as cardiovascular, musculoskeletal or mental fatigue. Reflected as increased performance errors rather than decreased performance per unit time. Relates to: lack of sleep, insufficient rest, too many daily work hours and/or work hours at an inappropriate time of day. Endurance time Rose et al. (2001) The time until work is interrupted due to fatigue. Pause/rest/breaks/muscular relaxation Konz (1998a) Defines the following categories of resting time : 1 off-work resting time 2 formal breaks, such as coffee breaks 3 informal breaks (work interruptions, training) 4 microbreaks (short breaks of a minute or less) 5 working rest (performing another task using a different part of the body). Exposure porosity Wells et al. (2007) The occurrence of restorative work breaks and pauses. Recovery Colombini (1998) A period within a working shift or cycle where no repetitive mechanical actions take place, allowing for metabolic and mechanical recovery of the muscle. Gilad (1995) Recovery value Konz (1998b) Recommends that recovery time must be at least twice as long as the duration of a forceful motion and occur between two such forceful motions. Insufficient recovery time between forceful acts, especially repetitions, is a risk factor leading to cumulative trauma disorder. A measure of the effectiveness of the rest. A function of how fatigued the muscle is when rest begins, the length of the rest, and what happens during the rest. Regarded as a dose/response relationship. Recovery time Rose et al. (2001) Has varying meanings depending on which aspect of recovery is in focus: can be related to critical pulse frequency levels or mean power frequencies of EMG signals. Resumption time Rose et al. (2001) The time until a task (that has been interrupted due to fatigue) is resumed again after resting. Cumulative approaches Kumar (1990) Integration of cumulative biomechanical load and exposure time over entire working lives (spanning several years). Assessed by interviews and by calculation of compression and shear at spinal discs using a biomechanical model. Seidler et al. (2001) Cumulative forces on the lumbar spine are calculated over an entire working life. Assumes a long-term exposure profile, focusing on the risk of contracting lumbar spine disease. Norman et al. (1998) Gilad (1995) Cumulative reduction of capacity Buckle and Devereux (2002) Uses a modified Mainz-Dortmund dose model a retrospective estimate based on over-proportional weighting of the lumbar disc compression force relative to the respective duration of the lifting process (Jager et al., 1999). Cumulative loads are calculated based on daily work shifts. Peak loads in a task are multiplied by the number of times the task occurs over the shift and by the duration of exposure for each task and then added to the spinal loading between work tasks (when not working). Cumulative load is calculated as the consequence of a particular task, and the total shift load is the sum of all task load integrals and the pause loads. The result of a repetitive job with limited recovery time between cyclic jobs. The frequency of hazardous movements per cycle and the pause/recovery mechanism is an important factor in avoiding cumulative trauma disorder. A result of overexertion of the muscles or frequent high muscle load, with either mechanism leading to muscular fatigue, which in turn contributes to MSDs.

14 Time-related ergonomics evaluation for DHMs 369 Generally, cumulative exposure is associated with a long-term exposure to physical loads resulting in reduced work capacity as a result of overexertion and/or frequent high muscle loads (Buckle and Devereux, 2002). Job repetition, with limited recovery time between cycles, is also widely investigated in cumulative loading studies (Gilad, 1995). Other approaches assume a shorter time span and more of a task division focus, such as loads accumulating over a work shift (Norman et al. 1998). Another important ergonomic risk factor is postural discomfort. Recently, scientists have begun to include discomfort functions into their posture prediction models (Kee and Karwowski, 2001; Marler et al., 2005; Yang et al., 2005; Fritzsche and Bubb, 2007). This push to include a discomfort function in DHMs is driven by the assumption that humans tend to adopt the most comfortable neutral position that avoids the extreme ranges of motion (Marler et al., 2005). Thus, such a discomfort function, particularly if used in conjunction with a muscular fatigue model to account for cumulative exposure, would be useful in identifying and preventing awkward postures that leave workers susceptible to workplace injuries. The antidote to fatigue and its related risk factors is usually termed pause, rest, breaks, muscular relaxation or recovery. A related term, exposure porosity, is defined as the occurrence of restorative work breaks and pauses (Wells et al., 2007). It has been suggested that recovery time must be at least twice as long as the duration of a forceful motion and occur between two such forceful motions (Gilad, 1995). Insufficient recovery time between forceful acts, especially repetitions, is a risk factor leading to cumulative trauma disorder, which is one of the most sought-after concepts in describing physical work exposure. Definitions of cumulative differ conceptually in the literature depending on which time perspective the physical load is related to, among other things. Konz (1998a) lists several strategies for preventing fatigue (general, muscular or mental) in a guideline, as summarised below: 1 revise work-scheduling policy (to avoid too many consecutive hours or hours at the wrong time) 2 optimise stimulation at work (to avoid mental over- or under-stimulation) 3 minimise the fatigue dose (by lessening work intensity and establishing a work/rest schedule) 4 use work breaks 5 use frequent short breaks 6 maximise recovery rate (by removing environmental stressors and muscle stressors) 7 increase the recovery/work ratio (by increasing recovery time OR decreasing work time) (adapted from Konz, 1998a). 3.2 Time-related evaluation methods In addition to gathering definitions of time-related terms, the literature review also yielded a number of ergonomics evaluation methods that are, in some way, able to describe load variation over time (Figure 3). The list of time-related ergonomic

15 370 C. Berlin and T. Kajaks evaluation methods shown here is not exhaustive; rather, the highlighted methods, which are described below, serve as examples to demonstrate the breadth and variety of available time-related evaluation methods. These evaluation methods each offer unique contributions to the evaluation of workplace injury risks over a time scale ranging from seconds, or a single cycle (e.g., the Strain Index), to hours, or an entire work shift [e.g., OCRA or the exposure index described by Krajcarski and Wells (2008)]. Some of the presented methods are used for prognostic purposes, such as to calculate the effects of varying certain parameters, while others are used for observation purposes, sometimes requiring subjective input. Figure 3 Examples of how time-related ergonomics evaluation methods apply to different time scales this demonstrates the variety of tools which are available to cover a workday time scale spectrum 1 Beginning at one of the smallest units on the time scale, the single force exertion, and extending to an application of multiple-task jobs (Bao et al., 2009), the Strain Index is a scoring system based on observation of work (Moore and Garg, 1995). It is calculated for exposure over a day, but is based on a task breakdown. With regards to time factors, the Strain Index rates a number of exposure parameters with qualitative and quantitative criteria levels and computes the Strain Index from the associated parameter-specific multipliers. Time-relevant aspects taken into consideration are duration of exertion,

16 Time-related ergonomics evaluation for DHMs 371 efforts per minute, speed of work and duration per day. The calculated Strain Index places the work in one of three categories; safe, action or hazard, indicating whether alteration of the job is necessary. One observation-based method that deserves special attention in this context is Yen and Radwin s (2000) use of spectral analysis on time series electrogoniometer data. The main rationale for this method is that, compared to any observation method, it offers considerably greater sampling resolution and precision of posture classification. The reason why it is brought up here is that its calculation algorithm could be very easily implemented in a DHM. Its main output is a quantification of posture magnitude, repetition rates, posture amplitude differences and repetition frequency. If used in a DHM, the method would describe joint motion and characterise repetitive behaviours, joint angle deviations and sustained postures. Quantification of variation in observed physical work can be calculated using exposure variation analysis (EVA), which was developed by Mathiassen and Winkel (1991). It applies to work movements up to entire job durations, where a job is a subset of a shift. EVA is based on the assumption that physiological response to the working load changes as a function of exposure time and can be administered to any continuously recorded signal. Instead of showing a loading sequence, EVA thus shows how loads are distributed over the working period as a percentage of maximum voluntary contractions, allowing for comparison of distribution profiles between different jobs. However, as it is a data-reduction method, EVA does not provide a real-time dispersion of exposure levels and leaves out exposure duration. To account for time-related factors such as repetitiveness and duration, fatigue models may be more beneficial. Ma et al. s (2009) dynamic muscle fatigue model operates on the premise that the influences of external load, workload history and individual differences are of major importance to fatigue development. The model s parameters include maximum voluntary contraction, current exertable maximum force and the external load of the muscle (i.e., how much force it needs to generate). The model also assumes that there is no recovery during the duration of work, and that fatigue is a growth function with time that is reciprocal to muscle force capacity (the greater the fatigue, the less capacity for force generation, and the longer a load is applied, the greater the fatigue). Calculation of the fatigue index is a data reduction method excluding the influence of time. Ma et al. have expressed an ambition to apply the model to a virtual reality framework involving DHMs (Ma et al., 2008a). Other sophisticated fatigue models exist (e.g., Freund and Takala, 2005; Liu et al., 2002; Xia and Frey Law, 2008), although their implementation into DHM software are hindered both by the purely theoretical nature of their models and the inherent challenges in creating a unique model that can be generalised across all human joints (Frey Law and Avin, 2010). Assessments of postural upper-body loading can be performed using LUBA (Kee and Karwowski, 2001), which is based on discomfort caused by joint motion and maximum holding time, and includes a posture classification scheme based on joint angle data. It is, in effect, a subjective method based on perceived discomfort, expressed as a ratio or postural load index for joint motions by the hand, arm, neck and back and the corresponding maximum holding times in static postures. The different index values obtained for varying postures allow for comparison of stresses and perceived discomfort levels across postures. However, one major limitation to this tool is that it is an evaluation of static postures, and does not consider postures held for longer than 60 seconds. By failing to account for

17 372 C. Berlin and T. Kajaks dynamic and time-related factors, important criteria in accurately assessing discomfort may be lacking. Lifting, pushing, or pulling tasks can be assessed using Ergo-Index (Glimskär et al., 1987; Rose, 2001). The original model requires input data in the form of population percentile, load as percentage of the maximum exertion force for the specified population, load distance, type, and magnitude, and operation time. Outputs that can be calculated using this model include recovery time, production time (i.e., work durations + recovery time) and the risk of back injury according to NIOSH criteria levels. The model has evolved to include the use of experimentally determined resumption times in the calculation of recovery, plus a general validation and extension of the original model s scope (Rose, 2001). Rose also suggests taking the influence of worker experience into account, that endurance-limiting pain arises in muscular rather than joint structures, and that the difference between sexes in endurance can be disregarded for the same relative load (% of MVC). Rose cautions that the model is still not entirely reliable for all types of EMG readouts, since the readouts tend to vary greatly when loads are low. Ergo-index also exists as a software program (BELAB, 1992). Specific to risks for low back pain, 4D WATBAK (Neumann et al., 1999) is assessment software based on an epidemiological database of risk factors. This tool calculates shift-long cumulative spine load (force-time integrals) and peak forces in the hands and on the spine using a biomechanical link segment model. After entering different work actions and specific work durations, the user can then compare generated values of peak force and cumulative load with threshold limit values to assess the ergonomic acceptability for the shift. Although 4D WATBAK as a software may be considered technologically outdated today, its principles can easily be implemented onto a more modern manikin. Another advantage is that it operates from a task-breakdown stance, which makes it compatible with many other analysis tools. Repetitive motions of the upper limbs can be assessed for MSD risks using OCRA (Occhipinti, 1998). This tool s main function is to sort repetitive tasks into problematic and non-problematic in order to steer priority of interventions. In effect an observation tool, OCRA calculates an index of exposure, based closely on the procedure for the NIOSH lifting index (Waters et al., 1993). The OCRA Index is based on a relationship between the actual number of actions performed in one day (or shift) and a recommended (maximum) number of actions. Under optimal conditions, the recommended number is 30 actions per minute, and the constant is diminished gradually as a function of the presence and characteristics of a number of risk factors (force, posture, recovery periods etc.). The OCRA index is fairly easy to calculate and requires the number of repetitive tasks over the shift and the duration of each repetitive task (in minutes). It also takes into account the effects of insufficient recovery time and perceived effort. It therefore appears fairly easy to integrate as a module in a DHM application. However, it is to be noted that differentiations must be made between left and right limbs. Also, Occhipinti emphasises that at the time the article was written (1998), the OCRA index had not yet been validated. Obtaining time-varying exposure information can be difficult, particularly in comparison to mechanical exposure and the amplitude of the exposure (Wells et al., 2004). However, Krajcarski and Wells (2008) describe an index-based approach, based on an epidemiological study of reported lower back pain among production workers. They argue that time-variation of loads (at the tissue level) and their historical exposure sequence is relevant to the body s biological response to work-related stressors.

18 Time-related ergonomics evaluation for DHMs 373 Subsequently, they describe an approach where the reporting of low back pain was correlated to individual worker exposure profiles (converted to exposure indices). Their findings show that their exposure index, which was sensitive to the order of tasks and the load variation, was more informative towards identifying back pain risk than a previously established set of risk factor indices that only acknowledged peak and cumulative loading (Norman et al., 1998). Krajcarski and Wells (2008, p.67) conclude that their approach could be valuable as a time-history sensitive exposure assessment tool and appeared to be sensitive to the duration, magnitude and timing of exposures in a manner leading to the enhanced identification of low back pain cases. 4 Discussion In this review, time-related ergonomics terminology has been studied, mostly with the conclusion that differing implicit time scale perspectives lead to a variety of definitions and near-synonymous uses. It appears that while the potential of incorporating time-related aspects into DHMs has been applauded, there is no strict consensus on what many of the concepts mean. The implicit definitions vary between different lines of research, leading to ambiguity when comparisons are made of different results. Thus, it appears that DHM developers are at liberty to define the input, throughput and output terms based on literature of their choice. As long as this is stated clearly and the terms are used consistently, the authors see no immediate problems. To appreciate the width of the contributions, it is important to recognise that there are some basic paradigmatic differences between the viewpoints of the different authors. Li and Buckle (1999, p.687) note that in epidemiological studies, different exposure variables are rarely considered simultaneously since little is known about the interaction between risk factors. In a similar manner, many authors reviewed here have acknowledged that isolated findings always have the caveat that a real work situation implies numerous other influential factors. For example, Kilbom (1994) notes that tolerance for repetitiveness has been known to be reduced by psychosocial work factors such as work control, time pressure and training. This multifactorial problem has been increasingly addressed in later contributions, such as in the study of the effects of operator experience on endurance (Rose et al., 2001) and the desire to extend the Strain Index to dynamic, multi-task jobs (Ma et al., 2009). Thus, DHM developers are more obligated to use research findings that take heed of multifactorial inputs. In addition to the multifactorial nature of injury risks in the workplace, many sources indicate that injury mechanisms causing MSDs vary by body part (e.g., Juul-Kristensen et al., 1997; Ma et al., 2009; Marras, 1992; Yen and Radwin, 2000). As mentioned before, it has been found that different body segments may be sensitive to different motion/load components. These segment-specific injury mechanism discrepancies and the associated injury prevention acceptability criteria should be considered in existing body-specific models. The implementation of these models into DHM software should also include temporal considerations. In concurrence with the scientific community s guidelines for how tasks should be described with regard to posture, forces, durations, and cycle times, DHM tools must also account for dynamic loading and muscular efforts over time (Chaffin, 2005).