Quality assurance laboratory planning system to maximize worker preference subject to certification and preference balance constraints
Introduction
The increased attention on economic considerations and operational efficiency has resulted in an extensive focus on personnel scheduling problems in the past few decades. The assignment of workers to tasks is a challenging problem encountered in both the manufacturing and service industries. For example, in manufacturing operations the problem may be assigning workers to particular machines, while in hospitals the problem may be the assignment of nurses to particular tasks or work shifts. Such problems are challenging, as each task requires specific skills that individual workers may not possess. Also, there are only a limited number of workers and tasks that need to be matched in a manner so as to optimize an objective function.
The classic assignment problem involves matching the elements of two sets on a one-to-one basis (i.e., tasks or jobs to agents or labor). Depending on the objective function, different types of problems are encountered ranging from linear assignment problems to quadratic and higher dimensional assignment problems. There is a large body of literature dealing with such problems and there are numerous applications in diverse areas such as production planning, telecommunications, and semiconductor design. The personnel scheduling problem encountered in today's workplace is different from the one addressed in most earlier studies, as the relative importance of satisfying employee needs in scheduling decisions has increased; many organizations consider employee preferences and offer flexible work schedules.
This research addresses the assignment of technicians to quality control tests (called test tasks) in a pharmaceutical manufacturing environment. In this environment test tasks are assigned to technicians by considering several factors, including worker certification and their task type preferences. Each technician is certified through internal training and examination to perform a certain set of task types. To maintain this certification a technician must perform a task of this type with certain regularity (or lose certification). Thus, at the start of each plan there is a set of task types that should be assigned to technicians in order to maintain their certifications. Furthermore, technicians have different preferences regarding performing each of the test tasks. In general, these preferences relate to two factors: the self-perceived skill of the technicians (typically, technicians prefer to perform tests they are good at), and the complexity of the tests (generally, technicians prefer the simpler tests or tests that have low probability of failure). The primary reason for such technician preferences is that they are incentivized to perform tests correctly. When a test fails it must be repeated, taking additional labor time and delaying the production / release processes. Clearly, these are outcomes management would like to avoid and therefore incentivizing technicians to execute tests correctly.
Each plan has a specified time window, and not all the pending tasks must be scheduled within the time window (test tasks not completed in the current plan are left for the next plan). Tasks are either priority tasks or non-priority tasks. Priority tasks are typically related to the quality tests required of production lots that are due within the time window of the plan, and to stability tests, which are tests performed for production lots already in the market that are being re-evaluated to ensure that they meet all specifications. In general, priority tasks consist of about half of the available workload. The scheduling process should use resources as fully as possible to maintain a high utilization level.
The production planning and scheduling literature has primarily focused on environments with either machine or labor type resources (e.g., Parsa et al., 2010, Wang and Chou, 2010). This research is different from the previous literature as it considers several factors that are particular to labor scheduling: preference to certain types of work and certification related to training in specific test types. It also has the uncommon consideration that not all the tasks must be scheduled within the specified time window. We propose and use technician satisfaction metrics and develop a heuristic to maximize them. We also conduct experiments to gain insights about the relationship among key variables, and compare the performance of the proposed heuristic with the optimal solution.
The rest of the paper is organized as follows. A review of the relevant literature is presented in Section 2. Section 3 provides a detailed problem description. Section 4 presents an example to further describe the problem and discuss the tradeoffs involved. Section 5 provides the proposed heuristic procedure, while Section 6 presents the results of the computational experiments. Finally, Section 7 provides conclusions and presents directions for future work.
Section snippets
Literature review
There is a large body of research dealing with the assignment problem. Votaw and Orden (1952), and Kuhn (1955) are regarded as pioneers in this area. Their work inspired others to study several variations of the classic assignment problem to solve a variety of practical problems in the past several decades. Pentico (2007) provides a comprehensive survey of the different variations of the assignment problem, as well as several examples of the problems.
Bergh et al. (2013) and Ernst et al. (2004)
Problem description and formulation
The problem considers that a fixed set of tasks is available at the start of the planning process. Tasks are divided into priority and non-priority tasks, and priority tasks must be included in the schedule. Each task is of a specific type and has an explicit duration. There is a fixed set of technicians available for the duration of the plan, and each technician can perform one task at a time. Technicians are certified to perform a specific set of task types. To maintain certification, they
Example
We assume three technicians (w = 3), twelve test tasks (n = 12), six test types (q = 6), and the planning horizon consisting of 17 time units (t = 17). Table 1 provides the test task information, including the type, duration, and priority for each task. Table 2 provides the technician to test type preference index, noting that a blank space indicates the technician is not certified to perform tasks of that type. The index ranges from 1 to 10, where 10 indicate a very high preference, and 1
Proposed heuristic procedure
The described problem becomes computationally complex as the number of technicians and test tasks increase. According to two recent surveys of chemical industry professionals, there is a need for decision-making support tools that are quick and easy to implement (Hodgett et al., 2014). Thus, one of our key goals is the ability to generate feasible solutions in a timely manner, given a set of management performance constraints. To accomplish this goal a heuristic is proposed. The heuristic is
Experimental results
We present a set of experiments to gain insights about the relationship among the problem's key variables and analyze the performance of the proposed heuristic. The performance of the heuristic will be compared with the optimal solution found (generated by an implementation of the mathematical program using CPLEX). The experimental setup is based on two factors: the production environment that serves as the basis for this research, and other studies related to parallel machines which analyze
Conclusions
The assignment of workers to tasks is a challenging problem encountered in different manufacturing and service industries, ranging from telecommunications to health care facilities. This research addresses the assignment of technicians to quality control tests in a pharmaceutical manufacturing environment, characterized by intense competition, demand uncertainty, stringent environmental/safety regulations, high cost and low success rate in product discovery, and long product development cycle
References (22)
- et al.
Modeling methods and a branch and cut algorithm for pharmaceutical clinical trial planning using stochastic programming
Eur. J. Oper. Res.
(2010) - et al.
Perspectives for process systems engineering- personal views from academic and industry
Comput. Chem. Eng.
(2009) - et al.
Challenges and opportunities in enterprise-wide optimization in the pharmaceutical industry
Comput. Chem. Eng.
(2012) - et al.
Review: state-of-the-art of optimization methods for short-term scheduling of batch processes
Comput. Chem. Eng.
(2006) Scheduling part-time personnel with availability restrictions and preferences to maximize employee satisfaction
Math. Comput. Modell.
(2008)Assignment problems: a golden anniversary survey
Eur. J. Oper. Res.
(2007)- et al.
Modeling realistic hybrid flexible flowshop scheduling problems
Comput. Oper. Res.
(2008) - et al.
Solving the parallel batch-processing machines with different release times, task sizes, and capacity limits by metaheuristics
Expert Syst. Appl.
(2010) Survey, categorization, and comparison of recent tour scheduling literature
Ann. Oper. Res.
(2004)- et al.
Empowerment Scheduling for a Field Workforce
J. Sched.
(2011)