Heap of medical pills. Credit: Volodymyr Hryshchenko on Unsplash
The US Food and Drug Administration (FDA) set out guidance in 2011 for “Process Validation: General Principles and Practices.” Not only should a pharmaceutical maker consider manufacturing controls as part of the product design phase and create a process to comply with Current Good Manufacturing Practice (CGMP) regulations, but also they also need to perform continued process verification (CPV) to maintain qualification status. Specifically, they recommend:
Continued monitoring and sampling of process parameters and quality attributes at the level established during the process qualification stage until sufficient data are available to generate significant variability estimates. These estimates can provide the basis for establishing levels and frequency of routine sampling and monitoring for the particular product and process. Monitoring can then be adjusted to a statistically appropriate and representative level. Process variability should be periodically assessed and monitoring adjusted accordingly.
Variation can also be detected by the timely assessment of defect complaints, out-of-specification findings, process deviation reports, process yield variations, batch records, incoming raw material records, and adverse event reports. Production line operators and quality unit staff should be encouraged to provide feedback on process performance. We recommend that the quality unit meet periodically with production staff to evaluate data, discuss possible trends or undesirable process variation, and coordinate any correction or follow-up actions by production.
This guidance is rigorous to ensure only high-quality pharmaceuticals reach the US market. However, some in the industry argue that continuous process verification has been slow to adopt. In the years since the guidance was released, we have undergone a transformation in applied statistics with a renaissance in the fields of machine learning and neural networks (AI). Some startups have latched onto the opportunity to use machine learning to analyze variation faster, more precise and at scale.
- Falkonry is “Assisting Continued Process Verification with AI”
- Mareana writes about “How To Create A Consistent CPV Process For All Your Plants”
- Quartic.ai’s products “create context from heterogenous discrete and analog data sources to detect hidden CPV variability signals with machine learning to create a real-time dynamic product control view for the enterprise.”
- Aizon discusses meeting strict compliance requirements in Pharma & Biotech
The Genius Build Process of Wind Turbines
Artificial intelligence for throughput bottleneck analysis
Identifying, and eventually eliminating throughput bottlenecks, is a key means to increase throughput and productivity in production systems. In the real world, however, eliminating throughput bottlenecks is a challenge. This is due to the landscape of complex factory dynamics, with several hundred machines operating at any given time. Academic researchers have tried to develop tools to help identify and eliminate throughput bottlenecks. Historically, research efforts have focused on developing analytical and discrete event simulation modelling approaches to identify throughput bottlenecks in production systems. However, with the rise of industrial digitalisation and artificial intelligence (AI), academic researchers explored different ways in which AI might be used to eliminate throughput bottlenecks, based on the vast amounts of digital shop floor data.
The history of Amazon’s forecasting algorithm
Historical patterns can be leveraged to make decisions on inventory levels for products with predictable consumption patterns — think household staples like laundry detergent or trash bags. However, most products exhibit a variability in demand due to factors that are beyond Amazon’s control.
Today, Amazon’s forecasting team has drawn on advances in fields like deep learning, image recognition and natural language processing to develop a forecasting model that makes accurate decisions across diverse product categories. Arriving at this unified forecasting model hasn’t been the result of one “eureka” moment. Rather, it has been a decade-plus long journey.
PLCs improve predictive maintenance
There is no doubt PLC technology is already strongly established on the plant floor. However, by embedding IT protocols, Cloud connectivity, and security features into today’s PLCs, it is possible to gather data that may have existed idly and use it to provide a much stronger idea as to what condition devices and machines are in to prevent unplanned downtime.
The digital transformation of the industrial enclosure
Before the proliferation of Industry 4.0, the design and engineering process for industrial enclosures and/or switchgear cabinets was a laborious, manual process with a variety of roadblocks in terms of collaboration, communication and visibility into the process. In addition, the ability to accurately project the durability of building materials and the enclosures as a whole was the equivalent of a dice roll, which provided even more uncertainty.
However, with the introduction of Industry 4.0, computer-aided design (eCAD) programs as well as software like EPLAN have provided engineers with real-time visualizations and insight into everything from material sourcing to the creation and alteration of panel and enclosure schematics. With EPLAN, enclosure-configuration data can easily be stored, retrieved, and shared to mitigate issues with downstream tasks.
Panel engineers and builders can gain powerful predictive insights into the environmental conditions where these enclosures will be deployed, enabling them to combat ingress of contaminants and increase the overall lifespan of the enclosure.
The Potential Impact of Automation on Manufacturing Profitability
Even at early stages of automation, the correlation between robot density and employee wages as a percent of revenue, or labor share, is clear: as a company becomes more capital intensive, labor’s share of revenue declines, as shown below, but automation increases productivity, potentially increasing both labor income and corporate profits, a win-win.
Manure Spreading goes High-Tech with IIoT
Manure spreaders have a tandem hydraulic pump. One pump drives the beater system at the backend that spreads, or applies, the product onto the field. A hydraulically driven end gate, or tailgate, opens up to allow the product out the backend, and the system also has a hydraulically driven variable speed floor.
An essential function of the control system is to monitor the torque load on the beater. With the beater requiring the highest horsepower load, it is crucial to use a pressure control, essentially a torque control, to keep the entire operation under maximum load the drive line can handle. For example, if the operator is driving the floor too fast, which increases the pressure, the control system will stop the floor or slow it down accordingly based on the load that you would see on that beater.
Printing process holds promise for bendable displays
A new process for creating flexible large area electronics could lead to breakthroughs in technologies including prosthetics, high-end electronics and fully bendable digital displays.
Until now, the most advanced flexible electronics have been mainly manufactured via a three-stage stamping process called transfer printing. Processes have been developed to make the stamping transfer more effective, but they often require additional equipment like lasers and magnets, which adds extra manufacturing cost.
The Glasgow team said they have eliminated the second stage of the conventional transfer printing process and replaced it with ‘direct roll transfer’ to print silicon straight onto a flexible surface.
Deep Learning Boosts Robotic Picking Flexibility
Gripping and manipulating items of diverse shapes and sizes has long been one of the biggest challenges facing industrial robotics. The difficulty is perhaps best summed up by the Polanyi Paradox, which states that we “know more than we can tell.” In essence, while it may be easy to teach machines to exhibit a high level of performance on tasks that require abstract reasoning such as running computations, it is substantially harder to grant them the sensory-motor skills of even a small child in all but the most standardized and predictable environments.
Assisting Continued Process Verification with AI
Patterns of behavior reflected in the data from equipment sensors can give insight into these performance affecting factors. In many cases, these patterns develop before product quality is significantly affected. Putting in place analytics that can detect these patterns gives the plant operations team actionable warning before CPV limits indicate a problem. This warning can be used to limit costly production impacts. Importantly, because the CPV process itself is untouched, these kinds of pattern detection analytics can be implemented without additional filings or regulatory delay. Assisting CPV does not mean replacing or even changing CPV.