• Subscribe

Algorithmic scheduling: Solving the pharma scheduling crisis

Embrace new technology to deliver agile, accurate manufacturing schedules

Algorithmic scheduling solution

How much time do schedulers have to spend wrangling data? Gathering updates from operators on the shop floor is part of the ‘daily grind’ of their job, with endless phone calls, emails and messages going back and forth over delays. All this takes an enormous toll on schedulers and leads to frustrations in manufacturing when changes aren’t turned around quickly enough.

A new approach

This technique uses machine learning and automated algorithms to collect and analyze real-time data feeds from the shop floor. The algorithms connect to and constantly monitor manufacturing data – providing automatic schedule updates whenever significant changes need to be made to operations.

This brings speed, accuracy and a more agile approach to commercial manufacturing schedules, all with considerably less effort from operators, supervisors and schedulers.

Improved shop floor scheduling

This technology makes four key improvements to shop-floor scheduling for pharmaceutical operations.

Reduces the burden of data collection. A real-time algorithmic scheduling hub cuts almost all the data chasing, emails and phone calls, cut-and-paste activities, and data cleaning from the schedule workflow. Once the platform is connected to the relevant data sets, it automatically pulls the data into a central location where algorithms review it in real-time. Each stakeholder group – operators, QC specialists, maintenance teams, supply chain – can link their data directly to the hub, eliminating the need for schedulers to chase it down. Once the process is set up, algorithms use the available data to make schedule predictions.

Uses machine learning to improve schedule accuracy. Because an algorithmic scheduling hub can integrate directly with data sources, it learns patterns of behavior over time. The machine learning algorithms within the hub can then be used to predict future performance even if the causal relationship between manufacturing patterns is not fully understood. For example, if the time to do a CIP (Cleaning in Place) on a particular skid is always longer than the scheduled time, a machine learning algorithm can alert the operator and allow extra time between when the CIP finishes and the next operation starts. These types of changes make for better schedules that are more responsive to delays. Additionally, variability in one process step won’t adversely impact others.

Generates more agile schedules that can be updated more quickly. The automation of scheduling data management means schedulers can reschedule, in real-time, publishing updates to the shop floor within a single shift. In the same way that airlines have moved to real-time notification of users when there’s a flight delay, real-time schedule notifications can help operators and technicians in manufacturing react sooner and more optimally. This makes the overall facility more agile to any delays or disruptions associated with single process steps. Algorithmic scheduling transforms the schedule into a responsive tool that enables facilities to respond on the fly as new data becomes available.

Schedulers become more strategic. By eliminating the manual tasks associated with data collection and cleaning, schedulers are free to spend their time on strategic tasks, like scenario analysis and assessing business impact of schedule outcomes. This elevates the role of the scheduling function from a tactical set of tasks to that of a key advisor for the leadership team within the organization. Algorithmic scheduling makes the scheduling role one of the most pivotal analytics roles within a manufacturing network, advising stakeholders on the best course of action across a wide range of planning scenarios.

How can I get started?

Scheduling today is in crisis, but schedulers are often wary of trying new things when even the basics of doing their jobs are so hard. The good news is that it isn’t difficult to implement basic algorithmic scheduling. Schedulers can ease into this new paradigm through incremental automation.

Our Smart Scheduling platform gives schedulers the tools they need to implement a ‘light’ algorithmic scheduling approach. Because it doesn’t require schedulers to learn Python or how to train a machine learning method, it can be easier to implement and ease schedulers and decision-makers down the path to total automation. Many manufacturing facilities begin with such a ‘light’ approach, in which only a subset of manufacturing systems (say, just the ERP and DCS systems) are integrated. In this case algorithmic scheduling can directly get information from operators and supervisors on the shop floor through basic web-based interfaces. The tool can then suggest changes but requires confirmation from the scheduler to move forward. This first phase shortens the time to results and drives significant efficiencies while still leaving the final decisions to the scheduler.

With rising pressure to generate more accurate schedules that allow facilities to be agile and reliable, schedulers can no longer delay embracing new technology. Algorithmic scheduling provides a series of algorithms and data handling technologies that can help best-in-class pharma manufacturers achieve that goal.

To learn more about how algorithmic scheduling can improve accuracy and agility,

About the Author

Picture of Rick Johnston PhD, Head of Manufacturing Data Science
Rick Johnston PhD, Head of Manufacturing Data Science
Rick Johnston PhD is the Head of Manufacturing Data Science at Automation Products Group Pharma, where he leads a team of data scientists and software engineers who develop and deploy cutting-edge solutions for optimizing pharmaceutical manufacturing. His specialties include forecasting, simulation, and AI/machine learning across a wide spectrum of solution areas including scheduling, analytics, process control, digital twin and others. Dr. Johnston has a PhD from Berkeley University and multiple patents in data science and simulation, with over 15 years of experience in building innovative software products for the life sciences industry. Software he designed, built and deployed is used today at more than 80% of the world’s top-30 pharmaceutical manufacturing organizations.