Using Simulation to Optimize Device Production

PRECISION TECHNOLOGY

The diversity and complexity of medical devices are likely to increase with time, as will the associated design risks and manufacturing challenges. Manufacturing systems must not only be compliant, but also be built on a solid understanding of the process in relation to its predefined design state. Manufacturers can often achieve such an understanding through modeling and simulation. This article details how the systematic and integrated use of simulation tools can be valuable throughout the device life cycle—from part and mold design, through injection molding cell performance, to factory-floor optimization.

What Are Simulations?

Simulations are computer-generated models that mathematically describe the operation of a manufacturing system. Computer-generated dynamic simulations describe, over a specific period of time, the working relationships between resources in a system.

Usually presented in graphic format, such simulations can be either stochastic (involving statistical inputs and outputs) or deterministic (requiring fixed input variables leading to fixed output variables). They can also be discrete-event simulations (models involving random events, each causing a change of state) or continuous-event simulations (models that are essentially built using differential equations). Material-flow modeling at the molded-part level is deterministic, continuous-event simulation. Dynamic simulations of material flows at the mold-cell and the plant-floor levels are stochastic, discrete-event models.

Simulation tools act as substitutes for physical experimentation with prototypes or full production systems. When compared with part prototyping or production trials, simulation has the advantage of time compression, physical scaling, experimental control, and, most importantly, risk and cost mitigation.

The Regulatory Environment

Medical manufacturers have been using more modeling and simulation over the past several years, encouraged by FDA and other regulatory bodies that are increasingly stressing science-based compliance. Simulation allows what-if design and operation scenarios to be developed and quantified, and such scenarios can then be used to evaluate and lower design risk. Manufacturers can use them to satisfy regulators' demands for hard science in both design and operations.

Simulation can facilitate a more-effective premarket notification [510(k)] or premarket approval (PMA) process. For example, PMAs require a device master file with details on the manufacturing process and physical characteristics of the device, as well as the proposed science-based controls. Both 21 CFR 820 and ISO 13485 require quality systems not only to detect and prevent problems, but also to analyze root causes of failure. Simulations can demonstrate such details.

Process simulations allow the desired state (or design space) of a process—and the probability of deviating from it—to be identified, assessed, and measured for risk. This reflects the what-if capabilities of simulators. Simulations can also contribute to process emulation, when real and idealized operations are compared in real time. Simulation tools are often economical methods to achieve more-consistent quality, improve efficiency, and lower the cost of regulatory compliance.

Effects on Process Efficiency

Operational efficiency is commonly measured in terms of overall equipment efficiency (OEE). This measure is a best-practice method for determining true plant capability. OEE is measured by combining availability, performance, and quality to produce consistent and repeatable metrics of business performance.

Calculating these three elements enables manufacturers to have the best chance at achieving maximum
quality for the lowest possible cost consistently over the life of the product. Manufacturing efficiency in the medical device industry is already lower than the manufacturing average. For example, medical device plant OEE levels below 50% are not uncommon,¹ while auto part and semiconductor manufacturers often have OEE levels of 85% or higher.

This discrepancy may explain why most device operations have recoverable hidden plant capacities that are factored into unnecessary downtime; process-related losses, scrap, or rework; and inefficient scheduling. Using temporary solutions such as overtime for employees or extended production hours is more difficult than in other industries because of the need to build and train for regulatory compliance.

Dynamic simulation is an inexpensive and powerful method to uncover and minimize hidden capacity. Simulations can directly influence better use of assets through the following:

Simulation software packages such as Arena, Promodel, Simul8, and Witness are used at the factory level. Aggregate stochastic input variables such as machine output, machine reliability, operator absenteeism, yield and scrap levels, and mold and color change time and frequency are used to simulate different outcome scenarios. Improvements or changes in these input parameters affect overall factory performance. With a simulation tool, this effect can be quantified and the results can be used to determine the potential for operational cost improvements.

Simulation at the Part and Tool Levels

Simulation has dramatically changed the way that plastic medical parts are designed. New designs can be modeled and their related molding processes can be improved well before the prototype is built or production tool steel is cut. This allows manufacturers to predict, and possibly prevent, potential part defects, which enhances part quality and reduces scrap.

Computer-aided engineering tools such as SigmaSoft, MoldFlow, and MoldEx allow mold process options to be selected, their sensitivities to various parameters analyzed, and the entire activity to be improved. These design tools vary in their ability to differentiate the processing characteristics of many new engineering polymer materials used in the medical device field. Therefore, part-level simulation must be tempered with an understanding of the material flow assumptions being made and, if necessary, in-mold resin trials may be performed.

There are three levels of related simulations that can be performed at the part level. The first is the mold cavity, that being the molded part itself. Analysis at this level allows manufacturers to simulate cavity filling, cooling, and packing. Relationships are modeled between pressure and temperature distributions on one hand, and wall thickness, gate locations, part geometry, and material-flow properties on the other. Part-quality risks associated with air traps, weld lines, sink marks, and structural weakness due to nonoptimized part filling can be predicted and corrected.

Gate and runner systems can also be simulated and should be added to the part-filling analysis. When designing runner systems, it is important to calculate the optimum melt-channel sizing and evaluate such parameters as pressure loss through the runner system. Multicavity molds require an evaluation of the filling balance.

A more-sophisticated analysis can be carried out with finite-element analysis (FEA), which can be used to refine the thermal and mechanical properties of the entire mold. In-depth computational fluid dynamics (CFD) can also be used to further analyze the entire part-filling process from a flow, temperature, or mechanical perspective. These complex properties affect the performance of the entire tool.

Part-Level Flow Simulation. Flow simulation can be used to eliminate an air trap on a medical part that could not be well vented on one of its functional surfaces. For example, Figure 1a shows the mold flow simulation with a predicted air trap. Figure 1b shows the actual part with the fault appearing as modeled. To eliminate the air trap, a 0.3-mm-thick flow leader (Figure 1c) was simulated as a means of changing the filling pattern.

A flow leader enhances the speed of filling by increasing a part's wall thickness in a selected area without affecting its functionality. Part designers typically use a flow leader as a last resort. In this example, the gate location has to be on a functional area of the part or the wall thickness has to be decreased in another area—neither of which is an attractive solution. Simulation can confirm the feasibility of using the flow leader in a nonfunctional part area and also optimize this fix through iteration (Figure 1d).

Tool-Level CFD Simulation. Most part-level simulations focus on runner layout, as well as part filling, packing, cooling, and warpage. More-sophisticated programs can evaluate multiphysics issues that involve melt-tool interactions such as shear-induced imbalance, resin mixing or blending, and complex part geometries.

A program such as Ansys can evaluate the thermal shear behavior on the thermal balance of a hot-runner manifold designed for a polyethylene terephthalate (PET)–based medical part. PET has a high viscosity index and is prone to shear-induced viscous heating upon injection. For the example medical part, thermal FEA shows a 25°C gradient between heater bands (see Figure 2). Knowing the potential shear-induced thermal effects and the 25° differential, a CFD analysis can confirm the tool design elements suggested by flow modeling.

By integrating conjugate heat transfer (melt and steel), high melt-channel temperatures can be expected. For this part, the heat transfer results in a thermal imbalance of only 14°C. Ansys CFD simulation allows the designer to confirm final channel and temperature layouts for an optimized balance system.

Simulation at the Production Cell and Floor Levels

Discrete-event, stochastic, dynamic simulations predict how process variables can change over time when moving from one steady state to another. Unlike melt-flow simulations, input parameters for baseline models are partly established with statistical inputs. Table I lists some examples of manufacturing inputs.

The system's operating and control logic parameters are modeled as discrete values and are taken from the system's design. However, the system's reliability and procedural parameters must be calculated or observed, and they should define statistical probability distributions rather than discrete values.

Optimization of a Molding Cell. The example in Figure 3 involves the following conditions:

Data for the baseline condition were obtained through several days of time and motion studies. This method may be less expensive than downloading module-by-module operating records from the programmable logic controllers driving the cell.

Figure 3 shows the line operating at an unacceptably low steady-state OEE level of 62%, excluding mold changeover. It also reveals that production output, which was 2830.6 parts per hour, was well below the calculated potential of 3600 parts per hour. Cell module use appeared relatively well balanced, and both the infeed dial and mold machine were characterized by high uptime. Neither showed a significant trend toward waiting or blocked conditions in steady-state operation.

In this example, the problem was with the mold. Thermal imbalance has been built into the tool. One of the resins used was high-viscosity PET. Also, the CFD recommendations described earlier had not been implemented. At that point, manufacturers may consider a what-if situation with a reworked tool. The predicted improvement in OEE levels simulated in Figure 3—up to 85%—seem to justify investment in tool refurbishment.

Optimization of a Molding Floor.Simulations of cell OEE after mold refurbishment can predict significant productivity improvement at the individual cell level. The challenge is to simulate both the effect of this cell-level improvement on the plant and the effect of other suggested improvements.

A baseline simulation can be created from time and motion studies. Simulated equipment utilization improvements can then be applied to this baseline. A simulated baseline floor is shown in Figure 4, along with production graphs quantifying the already high and increasing levels of work in progress (WIP) and the high- and slow-turning levels of final goods inventory.

Figure 4 also reflects the simulated floor future state after all the proposed improvements in Scenario 2. The simulated future state is characterized by WIP levels that are both lower and constant. The future state also has lower levels of final goods and increased final goods inventory turns. Machine OEE over the entire floor is higher (69% versus 57%), and on-time product delivery is predicted to improve dramatically.

Conclusion

Dynamic simulation is a powerful tool for optimizing medical device production. It can evaluate part design and also characterize production system behavior and system interdependencies. Overall, simulation technology is a cost-effective way for medical molders to improve part design and manufacturing efficiency while lowering the cost of regulatory compliance.

Joseph Fox is business manager of medical at Husky Injection Molding Systems (Bolton, ON, Canada). Santiago Archila is manager of factory planning at Husky, and Martin Baumann is responsible for business development for hot runners at the company. Contact the authors at [email protected], [email protected], and [email protected], respectively.

Reference

1. G Nystedt, “Production Logistics Improves Productivity in the Medical Device Industry,” FlexLink White Paper (2007).

Copyright ©2008 Medical Device & Diagnostic Industry