Packaging, process and progress: Unlocking new possibilities with your data

In day two of our Packaging, process and progress series, Alex McClung argues that a combination of data-infrastructure, robotic-infrastructure, and existing plant-infrastructure should be unified to enhance the capabilities of Australian manufacturing.

In industrial businesses, there are swathes of untapped data potential on the production floor. Data tells a story about the operations, and can be refined into useful information through data mining.

The confluence of novel sensing technologies, computational power availability, novel advancements in robotics, and the advent of deep learning is unlocking the true potential of industrial automation and lifting the focus towards an era of industrial cognition.

Learning algorithms can encode real-world operational phenomena into machine interpretable representations; these learned patterns can be visualised as a tool to aid operational staff and executives with better informed decision making – uncovering actionable insights, increasing capability, and reduced operating expenses. It is paramount that Australian manufacturers appropriately leverage their data to remain globally competitive.

The applicability of deep learning to manufacturing is immense, especially when paired with physical plant-infrastructure and robotic-infrastructure. Deep learning is causing a fundamental shift in the way that we draw inferences from data. Tasks such as “complex reasoning” are chiefly being driven by increased compute resources and model scale: “As the scale of the model increases, the performance improves across tasks while also unlocking new capabilities” (Chowdhery et al., 2022).

Unlike most physical commodities, data can be continuously refined. A new lens, algorithm, or analytical approach can enable more value to be extracted from the exact same unit of raw data input.

A modern plant is a highly sophisticated network of systems, with each component constantly measured and monitored. Often, there are many disparate non-interoperable processes, resulting in data islands. Aggregating and interpreting your data at a plant-wide scale increases the visibility of your operations. This enhanced visibility enables bottlenecks to be identified more clearly, and issues to be inferred and caught before they manifest as costly damages.

The pairing of robotic infrastructure with data infrastructure enables greater manufacturing flexibility and increases the downstream capacity for workers to engage. This allows workers to engage in more valuable and less-repetitive assignments resulting in higher job satisfaction.

Adoption of machine learning is increasing rapidly. Often, businesses start with the low-hanging fruit, then re-apply their newly-gained tacit knowledge to achieve operational improvement in other areas. As global competition increases and labour costs rise, companies that fail to automate will find themselves at a severe disadvantage.

Adoption of Industry 4.0 technologies and mindsets is necessary for companies to remain globally competitive, as the Australian economic complexity rankings dwindle and Australia experiences an “Innovation Deficit” (See this 2019 Australian Financial Review article, titled “Australia is rich, dumb and getting dumber”.)

The confluence of machine learning, sensing technologies, and robotic technologies is shaping up to be a transformative combination which unlocks the next generation of industrial capabilities. This is an exciting time for businesses and those that embrace machine learning will be well positioned for success in the years to come.

References

Chowdhery, A., Narang, S., Devlin, J., Bosma, M., Mishra, G., Roberts, A., Barham, P., Chung, H.W., Sutton, C., Gehrmann, S. and Schuh, P., 2022. Palm: Scaling language modeling with pathways. arXiv preprint arXiv:2204.02311.

Alex McClung is the founder of Traversal Robotics; he helps industrial companies set and reach their automation goals, facilitating the implementation of advanced robotics and machine learning technologies. He has been building autonomous robots for ten years, and developing tailored deep learning solutions for six years. His speciality is Machine Perception; enabling machines to see.

@AuManufacturing’s editorial series, Packaging, process and progress, is brought to you with the support of SMC Corporation.