Before companies begin adding a new production facility, modifying the plant layout to accommodate a new piece of equipment, or start other capital expenditure projects, many maintenance, engineering and operations departments use simulation software to better understand the alternatives. Simulation software allows users to model a process or an entire plant on a computer screen before committing to the expense of full implementation.
Simulation software also can be used to train operators and maintenance personnel before entering the live environment, thereby reducing risk. In concept, it's like using flight simulators to train pilots and technicians.
Basic simulation is already a competitive advantage for some CMMS vendors. For example, several packages have a "simulation" mode for scheduling. This feature, albeit somewhat manual, allows a user to build a schedule that best balances resource availability and work backlog by playing with the start and end dates for work orders, task duration, resource assignments and other variables. When satisfied with the schedule, the planner exits simulation mode by saving the selected options.
More powerful CMMS scheduling engines have an automated mode that uses optimization algorithms to generate schedules with optimal workload balancing. The planner can accept, reject or edit this computer-generated schedule within simulation mode.
During the past decade, simulation software has undergone a major facelift. Until the early 1980s, Ph.D. candidates in operations research were required to develop and interpret results of esoteric simulation models. Only a handful of large corporations could afford to buy the complex software to run on mainframe computers and keep the experts on their payrolls. Furthermore, management was asked to act on results backed by reams and reams of tabular data.
The big breakthrough came in the mid-1980s, when simulation software became "visual interactive," and could run effectively on a desktop computer. The term "visual interactive" means a user can produce an animated representation of the model and run it in simulated real time.
With modern simulation software, a user can easily depict a process, including operators, equipment and material, by selecting icons from an on-screen palette. Arrival rates, service times, flow rates, asset uptime, availability and other variables are selected using pull-down menus and dialogue boxes. The software then uses 3-D animation to observe bottlenecks, use of equipment and people. Reports show confidence intervals, graphical scenario comparisons, queuing data and other performance measures.
Simulation saves time and money because it avoids physical prototyping of changes to production equipment, layouts, schedules, material flow, manufacturing techniques, maintenance schedules and techniques, staffing levels and market demand. Simulation allows management to experiment with a computer-based representation of the change, thereby avoiding costly errors or missed opportunities. Many industries have profited from simulation, including defense, automotive, primary resource, telecommunications, electronic, health care and transportation.
One of the more popular applications of simulation is computer-based training. It is far safer and cheaper to train an operator using a simulator than to use the real equipment. Various situations can be simulated, including those for training operators about when to take corrective action as opposed to contacting maintenance. Emergencies can be simulated to better prepare maintenance and operations personnel.
Fairly sophisticated simulation software packages can be purchased for less than $1,000, but most of the popular packages are in the $5,000 to $20,000 range, depending on the number of users, the functionality required and so on.
As described above for the scheduling module, more CMMS vendors are embedding simulation algorithms to enhance their applications. This trend is expected to continue, given that CMMS vendors have expanded their use of workflow engines.
Workflow allows users to model the various maintenance processes, such as work management, from generating work requests through planning and scheduling (including appropriate approvals) to booking labor and parts consumed. Many stand-alone workflow and process-mapping software packages already have embedded simulation algorithms to allow users to test different workflow models. For example, by modifying the number and duration of steps in a workflow or the business rules about the way approvals are obtained, users can run a simulation to calculate which workflow is more efficient and effective.
Over time, CMMS vendors are bound to enhance their simulation capabilities so that maintenance can optimize its workflows. This is especially critical as the lines between condition monitoring, predictive maintenance, shop-floor data collection systems, HMI and ERP systems continue to blur.
Building a model
A certain degree of rigor is required to get the most out of simulation, whether the software is being used on a stand-alone basis or is embedded in a CMMS, workflow or other application. The greater the complexity and a model's cost-to-benefit ratio, the more formal the procedures required.
Develop conceptual design: This stage requires an understanding of which elements are to be modeled and to what level of detail. In turn, this drives what should be displayed on the screen, and the numerical data that is required for analysis. Make sure to state clearly the model's objectives, such as to determine the cost of achieving a 95% service level for "A" class spares inventory, or to determine the number of cranes, forklift trucks or other material-handling equipment required.
Collect and develop data: The quality of data available has a potentially large effect on the model's utility. Knowing, for example, the expected demand curve for spare parts in the warehouse is critical when designing an automated storage and retrieval system. This affects the size of the required facility, type and number of material handling units, feet of conveyor needed, and other dependencies. There is clearly a cost and benefit associated with determining the quantity of data and level of detail required to build a credible model.
Develop the model: Before starting, a thorough understanding of the system to be modeled is required. This encompasses which events need to be modeled, which modeling elements should be used and what period of time the model must simulate. Understanding who will use the model affects how it will be developed. For example, trainees for a new system and senior managers charged with a go/no go decision have very different modeling needs.
Test and validate: This includes a visual check, stepping through the simulation to check model logic and events sequencing, running actual data through the model to check results, performing sensitivity analysis on the data, and reviewing the model operation with individuals who understand the system being modeled. Testing is an important step; users must have confidence that the model is valid before expensive decisions are made.
Use the model: This step consists of multiple iterations of running the model, analyzing results and adjusting variables. Results are then presented for discussion and approval with respect to the stated objectives, for example, to gain support, prove a design or train users.
Contact Contributing Editor David Berger at [email protected].