• Subscribe

Pros and cons of various scheduling solutions for semiconductor factories

Choose scheduling that will be the right fit for your factory needs

Schedules can be created for the production floor using a variety of methodologies and technologies. These include: manual/Excel-based methods using simple FIFO or DDO rules which consider current work in process (WIP) and/or future WIP with simple/basic cycle time assumptions; area specific rules-based heuristic scheduling; simulation-based scheduling, and optimization-based scheduling. Methodologies also can be hybrid for both predicting future WIP arrival and for sequencing the lots to generate schedules. The choice and application of a methodology has an impact on the quality of the scheduling, as well as the productivity benefits the solutions can yield. At the same time, requirements for the quality and accuracy of the factory input data also increase across different types of scheduling systems.

Typically, a scheduling solution requires factory data to be extracted and converted or transformed to solution input data. Then the scheduling engine (regardless of methodology) applies the logic based on different considerations and objectives to generate scheduling output data. This then gets processed and published as schedules for end users to consume in a user friendly visual and analytical format. This is represented in figure 1, below.

Figure 1: Scheduling solution process flow
Figure 1: Scheduling solution process flow
The following looks at some of the benefits and limitations of the various methodologies and technologies.

Heuristics based scheduling solutions

This is a rule-based approach to lot assignment and sequencing. Manual factories lose productivity by deploying a simple FIFO, DDO, or lot priority approach to scheduling. Using more sophisticated heuristics based on the area, product mix, tool configuration and factory objectives will quicky improve tool utilization, throughput and cycle time of the tool sets.

Simulation based scheduling solutions

In simulation scheduling, models use simulation to predict future lot or WIP arrival based on their current status and position for all the tools in the factory. Simulation scheduling is considered a fab scheduler; it also can work with area schedulers by using their output as an input for downstream areas.

Optimization based scheduling solutions

In optimization scheduling, Mixed Integer Programming (MIP) or Constrained Programming (CP) models do lot allocation and assignment based on area-specific weighted objective functions, creating optimal equipment schedules. Optimization schedulers are area schedulers that can integrate with simulation-based factory schedulers.

Table 2, below, allows you to easily see the advantages of each solution.

Positive attributes of each solution

Heuristics Simulation Optimization
Easy to develop, configure and deploy
More accurate future lot/WIP arrival prediction
Best possible feasible and optimal schedules
Simple to learn, extend and customize
Can schedule for the whole factory and implement dispatch rules in scheduling
Detailed equipment modeling
Early gains for KPI improvements
Validate dispatch rules in a non-production environment
Better bottleneck management
Moderate sensitivity to input data requirements/quality
Scales across sites and companies and achieves better line balancing
Sensitive and responsive to changing factory conditions and objectives
Table 2: Positive attributes of heuristic, simulation, and optimization-based scheduling solutions
There also are limitations to each solution or technology, as shown in Table 3, below.

Limitations of each solution

Heuristics Simulation Optimization
May require constant fine tuning
Highly sensitive to factory data quality, latency, and granularity
Highly sensitive to factory data quality, latency, and granularity
Doesn't realize all potential gains for production
Uses a simplified version of dispatch rules implementation
Difficult to explain scheduler decisions
Needs to be in production environment to assess effectiveness
Not a mathematically optimal solution
Need longer deployment time
Often provides a good but not optimal solution
Higher resource requirements for maintenance
Higher resource requirements for maintenance
Not sensitive and responsive to changing factory conditions and objectives
Model performance time nonlinear with problem size

Table 3: Limitations of heuristic, simulation, and optimization-based scheduling solutions


Factories should select a scheduling software solution based on assessment of their factory requirements. Ideally, the solution should be a good fit between those requirements, your current and future maturity levels to absorb new technology and business practices, and capabilities of the software solutions available to you. The following questions can help in your decision-making process:

  • Does the factory need a factory wide scheduler or an area scheduler?
  • What is your bottleneck situation and the nature of your bottleneck?
  • Current baseline of your factory automation
  • Comprehensive account of your factory data generation, availability, gathering, storing, and processing capabilities
  • Skilled resource availability to help implement and support your scheduling automation vision

Heuristics, simulation, and optimization scheduling all help improve factory KPIs, and each has its pros and cons. Factories should gain a good understanding of the suitability and capability of these varieties of scheduling to ensure successful implementation.

However, it is important to note that companies without prior experience and implementation of a sophisticated and automated scheduling solution should consider deploying heuristics-based scheduling. This will help you achieve early ROI in terms of factory improvement and, at the same time, allow you to gain experience and expertise. Implementation of more sophisticated scheduling solutions based on simulation or optimization in a company that has no existing dispatching or scheduling system could result in a shallow learning curve or an unsuccessful deployment.

About the Author

Picture of Ravi Jaikumar, Global Product Manager, Real Time and Advanced Scheduling
Ravi Jaikumar, Global Product Manager, Real Time and Advanced Scheduling
Ravi is a Global Products Manager for Real Time Dispatching and Scheduling software solutions for semiconductor front end fabs and Assembly, Test and Packaging factories. Prior to joining Applied Materials Automation Products Group almost two years ago, he was a senior industrial engineer with Qorvo, Inc. He also served as an industrial engineer for ON Semiconductor and was a supply chain consultant with Hyster-Yale Group. He earned a bachelor’s degree in mechanical engineering from Anna University Chennai, and a master’s degree in industrial engineering from the North Carolina State University.