In the life sciences industry, the key to success is both accelerated innovation and operational excellence. And workers are a critical part of the equation across the board.
On the plant floor, your operators must make timely decisions to help ensure processes run within defined setpoint ranges and support approved market authorization requirements. But as biologics and other pharmaceutical products become more complex, so does monitoring – and optimizing – the process.
The crux of the problem?
While smart devices and complex systems generate a wealth of data, operators often have limited information to support better decisions.
Thanks to new developments in advanced analytics and machine learning, you can now put data science to work on your plant floor, drive operator action in real time and extend insights across the enterprise.
Let’s take a look at a few examples.
Lineside Next Best Action with Soft Sensors
When it comes down to it, each action an operator takes has the potential to affect the quality of your product or the throughput of your process. A “lineside next best action” strategy delivers real-time decision support using machine learning models that both predict future performance and prescribe actions to mitigate negative impacts.
How?
Here’s a recent use case we developed. Let’s say you must achieve a critical moisture level for a granular drug substance. The substance begins as liquid and is passed through a fluid bed dryer. The drying process is routinely stopped for moisture testing of physical samples, introducing latencies into the drying process.
A model predictive control (MPC) approach was utilized to train a machine learning model to predict moisture content and essentially build a virtual or “soft sensor” that can reliably infer when moisture content is optimal.
While trained in the cloud, the machine learning model runs on the edge to deliver critical information – and prescribed actions – to the operator in real time.
First, a machine learning model is trained with the historical dataset to determine how multiple independent variables – including inlet and exhaust temperature, dryer fan speed and other environmental conditions – impact the dependent variable of moisture level. Once the model is trained, MPC uses real-time sensor data, the current dynamic state of the process, the trained MPC model and the process variable targets and limits to calculate future changes in the dependent variable of moisture levels with a high degree of confidence.
The result is shorter drying process cycle time with fewer interruptions for physical sampling and measurements.
In fact, one of our clients experienced a 28%-30% reduction in dryer cycle time using this solution.
Continued Process Verification
Advanced analytics and machine learning also support decision-making in situations that have some level of latency tolerance. In other words, it can support decision-making in situations that do not require immediate operator response. Continued process verification (CPV) is one example.
Historically, the pharmaceutical industry has taken a retrospective approach to confirm that a batch meets all market authorization-specific quality, safety and efficacy requirements. Workers review achieved critical process parameters after the batch is executed and scrap batches that don’t match market authorization requirements.
Increasingly, the industry is adopting CPV as a proactive alternative to this approach. The system continuously monitors production processes in real time, uses statistical process control methods to plot trends so operators can immediately see how well the process is adhering to defined and approved setpoints – and identify any drifts or emerging trends. Corrective actions can then be taken to address the drifts or trends to avoid violating the approved ranges.
Scaling Up for Enterprise-Level Success
So what’s the big picture? Advanced analytics and machine learning can be applied to multiple use cases to support operator decision-making across the life sciences value chain.
To begin your journey, bring together a coalition of multifunctional stakeholders – and plan for and create short-term wins with use cases aligned with your digital vision.
Each use case can deliver extraordinary results on an individual process line. But value increases when those new approaches are institutionalized and scaled across multiple lines and facilities.
The ultimate vision? For many, it’s a manufacturing control tower (MCT), which will provide an aggregate view of actionable, analytic and predictive information throughout the production landscape and enable better decision-making across the enterprise.
Learn more about how advanced analytics can transform pharma manufacturing.
We can help you achieve your vision – learn how.
