The manufacture of pharmaceuticals is a highly regulated and complex process. Patented products are licensed for a finite amount of time before they become generic, equating to a constant sense of ‘the race is on’ to be able to meet demand. 

Now, possibly more than ever before, these businesses globally have an intense focus on reducing supply chain disruption, increasing capacity of batch production and reducing batch losses. Reducing lifecycle maintenance costs and CAPEX remain high on the agenda too.

Notwithstanding compliance and safety, manufacturing equipment availability is therefore a top priority. Without exception, pharma manufacturers tell us that they want to be able to predict asset degradation and failure well in advance of an impending breakdown or disruption to be able to make decisions that can minimize cost and disruption.

Simplicity – high accuracy, fewer false positives

Today’s machine learning solutions allow pharma manufacturers to achieve fast results without needing to write a single line of code. The data science is hidden and allows ‘normal’ workers to manage them. Current staff, already employed, can be easily taught and trained to manage the platform.  The number of “qualified” users is therefore very high, enabling engineers to solve their own engineering problems!

In many industries data is unpredictable. It has anomalies, there is a danger that you can be sent unwillingly down the wrong path.

Pharma, however, has very little, if any ‘crazy data.’ Why?  Because by its very nature, the process of manufacturing drugs is hyper-controlled. The adoption of machine learning can bring rapid results and value to pharma companies within weeks.

Scalability – failure signatures transferable across assets

The pharma industry also has pools of similar equipment, such as the same pumps used in multiple services, or several of the same packaging lines. This is where transfer learning comes in to its own. By sharing the normal and failure behaviours of assets that we find on one machine with the other members of the pool, we can rapidly increase the scale, safety and prevent breakdown of all equipment of the same type and configuration. This ability to rapidly scale an enterprise can create millions of dollars in value.

Speed – faster results as no asset model required

One example that demonstrates such results can be found at a large-scale pharmaceutical plant, where several large chillers and compressors are critical equipment infrastructure. Despite all six sigma efforts, failures were still causing enormous losses. Aging equipment, increasing energy usage, higher maintenance and inadequate equipment health status reporting contributed to the problem. Aspen Mtell®’s industrialized machine-learning reversed the situation. Autonomous agents turned corrective and preventive maintenance into prescriptive maintenance. These agents now advise when equipment should be maintained with early warning of impending failures. This gives sufficient notice for orderly, rapid problem correction at the lowest cost. Overall production has improved dramatically.

In another application, AspenTech’s Aspen Mtell solution was used to determine the early signs of seal failure through learning of similar patterns from live equipment. It was also focused on continuously learning new events (normal and abnormal) together with additional late stage indicators, providing a confidence increase in seal change decisions. The result was a decrease in the frequency of the need to make mechanical seal replacements, leading to a lessening of supply chain disruption; a reduction in lifecycle maintenance costs of 60%, and a reduction in CAPEX and associated lifecycle costs of 50%.

Across the pharmaceuticals industry today, the latest asset performance management (APM) solutions are enabling pharmaceuticals companies to protect their supply chain, increase asset utilization and avoid unplanned downtime by accurately predicting when equipment anomalies will occur, understanding why they do, and prescribing what to do to avoid a potential failure. Given today’s volatile and challenging marketplace, it is exactly the right time for pharmaceutical manufacturers to act to accelerate their approach to digital transformation and machine learning solutions.