Hybrid procedure to determine optimal workforce without noise hazard exposure
Introduction
It is known that a cumulative effect of repetitive exposure to loud noise is occupational hearing loss. According to the National Institute for Occupational Safety and Health (NIOSH, USA), approximately 30 US million workers are currently exposed to noise hazard on the job and an additional 9 million are at risk of hearing loss. Noise-induced hearing loss is one of the most common occupational diseases and the second most self-reported occupational illness or injury. Some examples of an economic impact of hearing loss are in British Columbia: from 1994 to 1998, the Workers’ Compensation Board paid $18 million in permanent disability awards to 3207 workers suffering hearing loss. An additional $36 million was paid out for hearing aid (NIOSH, 1998).
For industrial facilities having high noise levels, appropriate noise controls must be implemented to prevent workers’ daily noise exposures from exceeding a permissible level. Engineering controls are the most effective noise hazard prevention, which can be done through proper design, maintenance, lubrication, and alignment of machines. The use of barriers or shields can reflect high-frequency noise. Isolating machines from areas where workers are likely to be present also helps to reduce noise levels. The use of hearing protection devices (HPDs) is another common noise control approach. Earplugs and earmuffs are widely used in industry. While the application cost of HPD is attractively inexpensive, its effectiveness is usually the poorest. Noise reduction ratings (NRR) tend to be overly stated by manufacturers (Gasaway, 1984). Wilson, Solanky, and Gage (1981) reported that workers typically adjust their HPDs for comfort rather than to achieve maximum attenuation. In many manufacturing facilities, workers choose not to use HPDs unless strictly enforced and routinely monitored. Comfort seems to be a major factor that influences whether or not HPDs are worn by workers (Casali, Lam, & Epps, 1987).
Other than using engineering controls or HPDs, job rotation is popularly recommended to reduce workers’ daily noise exposures. Job rotation is an administrative noise control that offers a trade-off between safety concern and cost effectiveness (Olishifski & Standard, 1988). Briefly, job rotation requires workers to rotate among workstations within one workday in order to reduce their daily noise exposures. In a few special situations, job rotation not only helps to reduce noise hazard exposure but also increases productivity by sharing very demanding tasks among workers (Royster & Royster, 2003). Despite being an effective yet inexpensive noise control, job rotation has not received much attention from industrial engineers and/or safety practitioners. This is perhaps due to difficulty in implementing job rotation to achieve its optimal level of effectiveness. In practice, the number of workstations where individual workers will attend, work duration at each workstation, and the order of assignment must be defined. Since a major goal is to reduce workers’ daily noise exposures, it is necessary to search for a feasible set of work assignments such that none of the workers receives daily noise exposure exceeding 90 dBA. In a case where there is no feasible set of work assignments with the current number of workers (which is usually equal to the number of workstations), the number of workers must be increased. However, the number of workers who are exposed to high noise levels should be minimized (NIOSH, 1998).
In this paper, we consider an optimization problem of finding the minimum number of workers and their daily work assignments to attend a set of workstations such that none of the workers receives daily noise exposure exceeding 90 dBA. The problem is called the one-dimensional minnum work assignment problem based on noise exposure (1DMAP-N). 1DMAP-N can be viewed as a variant of the one-dimensional bin packing problem (1BPP). The paper is organized as follows. We initially discuss a measure of daily noise exposure and the use of job rotation to reduce it. Next, we review research work on existing 1BPP algorithms. After that, we describe a mathematical model of 1DMAP-N and evaluate the relationship between 1DMAP-N and 1BPP. Then, we explain four solution algorithms and a hybrid procedure for solving 1DMAP-N in detail. Finally, results of a computational experiment with 300 test problems are reported and discussed.
Section snippets
Daily noise exposure and job rotation
Daily noise exposure is measured in terms of an 8 hour time-weighted average (8 hour TWA, dBA) sound level. Letting hj be work duration (in hour) at workstation j, n be number of workstations, and be combined noise level (dBA) measured at workstation j, the 8-hour TWA can be determined using the following formula (adapted from OSHA, 1983).Note that the permissible daily noise exposure is equivalent to the 8-hour TWA of 90 dBA (OSHA, 1983).
The
The 1DMAP-N model and its relation to 1BPP
For readers who are not familiar with 1BPP, its objective is to find a minimum number of equal-sized (i.e., fixed-height) bins that is sufficient for being packed by a set of one-dimensional items with various sizes. When comparing between 1BPP and 1DMAP-N, one can see that an item in 1BPP is analogous to an amount of noise exposure per work period that a worker receives when attending a given workstation in 1DMAP-N, whereas a bin is analogous to a worker. In 1BPP, all bins are identical.
Review on selected algorithms for solving 1BPP
Since 1DMAP-N is a variant of 1BPP, an understanding of existing algorithms for solving 1BPP and their concepts could be useful for the development of solution algorithms for 1DMAP-N. Because 1BPP is also one of the first combinatorial optimization problems in the literature, there are a number of research papers on the bin packing problem and its variants. As reported by Hochbaum (1995), there are over 100 research papers written on the bin packing and related problems since 1960s, and the bin
Solution algorithms to determine optimal workforce
In this section, a method for identifying the lower bound of 1DMAP-N is firstly explained. Then, three approximation algorithms and one exact algorithm are described. Their applications are demonstrated through a series of numerical examples that are based on the same noise data. Work assignment solutions (consisting of number of workers and daily noise exposure) are also compared.
Hybrid procedure
As seen in the previous section, individual algorithms can be separately utilized to solve 1DMAP-N. While the first three approximation algorithms may not guarantee the optimality of the work assignment solution (when UB > LB), the exact algorithm (DA-BnB) will yield the work assignment solution that is optimal. Depending on the problem size, DA-BnB may take a few seconds to yield the optimal solution or may need hours of computation time. To improve the solution approach (to find the optimal
Test problems
Three sets (A, B, and C) of test problems (1DMAP-N) were randomly generated. Each set consisted of 100 problems, which were grouped into five levels of the number of workstations (n), i.e., 10, 20, 30, 40, and 50 workstations, respectively. For each n, there were 20 test problems. The number of work periods per day p was assumed to be four equal periods. In each problem set, the values of noise weight were randomly generated using a uniform distribution between these following ranges:
- Set A:
wj ∈
Discussion
The hybrid procedure clearly outperforms all four algorithms when considering the number of optimal solutions (hits) and the maximum computation time. In terms of the number of optimal solutions or hits, the hybrid procedure achieves the highest hits, which is 265 out of 300 problems, followed by M-LPT swap (236), BnB-H (224), DA-BnB (212), and mFFD (119). In each set of test problems, the hybrid procedure also achieves the highest hits. These results show that the hybrid procedure successfully
Conclusion
The hybrid procedure for solving the one-dimensional minnum work assignment problem based on noise exposure (1DMAP-N) is proposed. The procedure consists of four algorithms. Firstly, the lower bound (LB) for 1DMAP-N is determined using the same method as that in 1BPP (Martello & Toth, 1990b). Next, the first fit decreasing (FFD) heuristic is modified to identify the upper bound (UB) for 1DMAP-N. If UB > LB, then the M-LPT swap heuristic tries to decrease UB using the following algorithms: M-LPT
Acknowledgements
The authors wish to express their gratitude to the Thailand Research Fund for financially supporting this research study through the RGJ-PHD grant (Grant No. PHD/0138/2544). The authors are also indebted to the anonymous referee for the valuable comments.
References (31)
- et al.
An improved lower bound for the bin packing problem
Discrete Applied Mathematics
(1996) - et al.
New heuristics for one-dimensional bin-packing
Computers and Operations Research
(2002) - et al.
Lower bounds and reduction procedure for the bin packing problem
Discrete Applied Mathematic
(1990) - et al.
BISON: a fast hybrid procedure for exactly solving the one-dimensional bin packing problem
Computer and Operations Research
(1997) - Alvim, A. C. F., Aloise, D. J., Glover, F., & Ribeiro, C. C. (1999). Local search for the bin packing problem. Extended...
- et al.
A hybrid improvement heuristic for the one-dimensional bin packing problem
Journal of Heuristics
(2004) - et al.
Rating and ranking methods for hearing protector wearability
Sound and Vibration
(1987) - et al.
Approximate solutions to bin packing problems
- et al.
An application of bin-packing to multiprocessor scheduling
SIAM Journal of Computing
(1978) A hybrid grouping genetic algorithm for bin packing
Journal of Heuristics
(1996)
New classes of fast lower bounds for bin packing problems
Mathematical Programming
Computers and intractability: A guide to the theory of NP-completeness
NIOSH compendium hearing protector attenuation
National Safety News
A new heuristic algorithm for the one-dimensional bin-packing problem
Production Planning and Control
Cited by (15)
Reducing physical ergonomic risks at assembly lines by line balancing and job rotation: A survey
2017, Computers and Industrial EngineeringCitation Excerpt :Asawarungsaengkul and Nanthavanij (2008a) propose a genetic algorithm for the first job rotation problem with the objective of minimizing workplace changeovers. Yaoyuenyong and Nanthavanij (2008) develop a construction heuristic with an integrated local search and Yaoyuenyong and Nanthavanij (2006) design several further algorithms, including a greedy multi-start heuristic, local search procedure, enumeration-based heuristic and a branch-and-bound procedure, for the second job rotation scheduling problem. Thereby, Yaoyuenyong and Nanthavanij (2008) show that the formulated model is suitable for ergonomic measurements with EnerExp.
Production line workforce scheduling optimization considering noise exposure risk
2022, Journal of Physics: Conference SeriesA job rotation-scheduling model for blue-collar employees' hand–arm vibration levels in manufacturing firms
2021, Human Factors and Ergonomics In ManufacturingDesigning cyclic job rotations to reduce the exposure to ergonomics risk factors
2020, International Journal of Environmental Research and Public HealthA Multi-objective Model for Ergonomic Workforce Scheduling with Bi Hazard Considerations
2019, RI2C 2019 - 2019 Research, Invention, and Innovation Congress