Open challenges in manufacturing data-driven approaches

In the last decade, in several surveys applied to the most senior corporate executives with oversight and responsibility for data within their firms have noted two significant trends. First, mainstream companies have steadily invested in Big Data and AI initiatives in efforts to become more data-driven: 91.9% of firms report that the pace of investment in these projects is accelerating, and 62.0% of firms reporting data and AI investments of greater than $50 million. Second, firms are continuing to struggle to derive value from their Big Data and AI investments and to become data-driven organizations. Often saddled with legacy data environments, business processes, skill sets, and traditional cultures that can be reluctant to change, mainstream companies appear to be confronting greater challenges as demands increase, data volumes grow, and companies seek to mature their data capabilities [1].

In 2021, Fortune 1000 companies have reported a decline in the leading metrics which are used for measuring the success of their data and AI investments. Companies are struggling to make progress and, in many cases, even losing ground on managing data as a business asset, forging a data culture, competing on data and analytics, and using data to drive innovation. Only 29.2% report achieving transformational business outcomes, and just 30% report having developed a well-articulated data strategy.

Without doubt on of the sector that is pushing hard the data-driven approaches is the manufacturing industry. Data traceability, trust, quality, sources type (structure, non-structure, etc.), frequency (real time, almost real time, batch, etc.), administration, taxonomy and many other definitions are in the roadmap of the Industry 4.0 strategies. Nevertheless, there are still critical challenges in processing manufacturing data as useful knowledge, resulting from the unique characteristics of manufacturing data from the following aspects:

Multiple sources: data comes from multiple sources, such as manufacturing management systems, shop-floor equipment, imagens & video records, and human (many times free text) inputs.

Multi-dimensional: asset features can be associated with process data, such as temperature, velocity and strain. These data can be scalars, vectors, and tensors. The multi-dimensional nature of the data reflects the knowledge graph’s semantic heterogeneity.

Structured and unstructured: Manufacturing processes include not only structured data, such as data stored in relational databases but also a large amount of unstructured and semi-structured data, such as images, text and videos. The booming of IoT technology has led to an explosion of these unstructured data. These data’s size is much larger and more challenging to handle than structured data.

Unbalanced data: Engineering data often contains a small amount of anomalous data and a large amount of useless data. Direct application of data analysis models on a biased dataset likely would get unreliable results.

Siloed data: manufacturing data tend to aggregate around a particular assembly process, production line or a particular asset, losing many times the end-to-end overview of the entire value chain.

Explainability: One of the biggest challenges that machine learning and artificial intelligence application are facing it is the poor explainability of the correlation between data inputs and model output, as well as, the poor understanding by the user (workers) of the impact on the decision making if some of the inputs are modified during the manufacturing process.

Reliability: finally, knowledge and data gathered from shopfloors tend to be context-specific and implicit. It is hard to develop an accurate semantic model based on unreliable data sources.

Smart manufacturing or Industry 4.0 solutions are a good example, where the data-driven business transformation is a long-term process that requires patience and fortitude. Investments in data governance, data literacy, programs that build awareness of the value and impact of data within an organization, may represent an eventual step in the right direction, but organizations must show that they are in it for the long haul and stick with these investments and not lose patience or abandon efforts when results are not immediately forthcoming.

In our opinion, nowadays, one of the most important short-term actions to accelerate the data-driven adoption, it is to build digital-oriented culture where customers, digital providers, and governments co-create the financial mechanism and technical roadmaps to smoothly migrate to a connected and digitalized industry.

In future posts, we will be continue writing about technology and business trends for enterprises. Furthermore, we recommend consulting the following literature to continue your digital transformation journey:

• Designed for Digital: How to Architect Your Business for Sustained Success, MIT review
• The Future Is Faster Than You Think: How Converging Technologies Are Transforming Business, Industries, and Our Lives, by Simon & Schuster
• Artificial Intelligence: The Insights You Need, by Harvard Business Review
• The Year in Tech, 2021: The Insights You Need, by Harvard Business Review
• The Deep Learning Revolution, by MIT Press
• Competing in the Age of AI, by Harvard Review Press

The objective of this blog is to provide a personal vision of how digital transformation trends will be impacting in our daily activities, businesses and lifestyle.

———————————————

[1] https://hbr.org/2021/02/why-is-it-so-hard-to-become-a-data-driven-company

———————————————

By Gerardo Beruvides

Industry 4.0 and Smart-mobility expert, his research interest includes Industry 4.0, Smart-Maintenance, Process Optimization, Machine Learning, AI engineering and Cloud-based solutions.