Achieving Trusted AI in Manufacturing

In the dynamic landscape of modern manufacturing, AI has emerged as a transformative differentiator, reshaping the industry for those seeking the competitive advantages of gained efficiency and innovation. As we navigate the fourth and fifth industrial revolution, AI technologies are catalyzing a paradigm shift in how products are designed, produced, and optimized. 

With the ability of manufacturers to store a huge volume of historical data, AI can be applied in general business areas of any industry, like developing recommendations for marketing, supply chain optimization, and new product development. But with this dataalong with some context about the business and processmanufacturers can leverage AI as a key building block to develop and enhance operations. 

There are many functional areas within manufacturing where manufacturers will see AI’s massive benefits. Here are some of the key use cases: 

  1. Predictive maintenance: With time series data (sensor data) coming from the equipment, historical maintenance logs, and other contextual data, you can predict how the equipment will behave and when the equipment or a component will fail. With AI, it can even prescribe the appropriate action that needs to be taken and when.
  2. Quality: Use cases like visual inspection, yield optimization, fault detection, and classification are enhanced with AI technologies. While outcomes within industry segments will vary, the potential is huge. For example, improving yield in the semiconductor industry even by a small fraction of a percentage point could save millions of dollars. 
  3. Demand forecasting: AI can be used to forecast demand for products based on historical data, trends, and external factors such as weather, holidays, seasonality, and market conditions.

While AI stands to drive smart intelligent factories, optimize production processes, enable predictive maintenance and pattern analysis, personalization, sentiment analysis, knowledge management, as well as detect abnormalities, and many other use cases, without a robust data management strategy, the road to effective AI is an uphill battle.

The universal industrial data challenge

Dataas the foundation of trusted AIcan lead the way to transform business processes and help manufacturers innovate, define new business models, and establish new revenue streams. Yet many manufacturing executives say they are challenged in adopting new technologies, including AI for new use cases. According to Gartner, 80 percent of manufacturing CEOs are increasing investments in digital technologies—led by artificial intelligence (AI), Internet of Things (IoT), data, and analytics. Yet Gartner reports that only eight percent of industrial organizations say their digital transformation initiatives are successful. That is a very low number. 

The lack of universal industrial data has been one of the major obstacles slowing the adoption of AI among mainstream manufacturers. Advanced technologies are only part of the digital transformation story. Manufacturers who want to get ahead must understand data’s role and value. With the very low cost of sensors: new equipment is being standardized with sensors and old manufacturing equipment is being retrofitted with sensors. Manufacturers now have unprecedented capacity to collect, utilize, and manage massive amounts of data.  

In this age of industrial IoT, it’s possible to rapidly introduce tools to produce actionable results with huge data sets. But without the highest level of trust in these data, AI/ML solutions render questionable analysis and below-optimal results. It is not uncommon for organizations to construct solutions with faulty assumptions about datathe data contains every scenario of interest and the algorithm will figure it out. Without a thorough grounding with trusted data and a robust data platform, AI/ML approaches will be biased and untrusted, and more likely to fail. Simply put, many organizations fail to realize the value of AI because they rely on AI tools and data science that is being applied to data which is faulty to begin with.  

Trusted AI begins with trusted data

What resolves the data challenge and fuels data-driven AI in manufacturing? Develop a data strategy built on a robust data platform.

Manufacturing operations and IT have to work hand-in-hand to develop a data-centric culture, with IT responsible for end-to-end data life cycle management focused on reliability and security. 

There are several best practices specifically when it comes to the data:

  • You don’t need to boil the ocean. Start with a pilot problem on the manufacturing floor that needs to be solved. 
  • Identify the use cases that help manufacturing operations add value. Let that dictate the data you want to collect.
  • Build out capabilities to collect and ingest data with IT/OT convergence, and collect and ingest the shop floor and equipment data onto a centralized platform on the cloud.
  • Add appropriate contextual data (IT/business data), which is critical in AI analysis of manufacturing data.
  • Eliminate data silos. Data from multiple sources must be centralized and stored on a common data lake so that you will have one source of truth across the value chain.
  • Apply AI tools and data science to the data that you trust and provide insights to the appropriate people or the system to make the best, most informed decisions.

The value of a hybrid data platform

AI can help manufacturers improve operations and achieve the next level of operations excellence. But the key is to focus on data first, not complex AI systems. Manufacturing organizations still use legacy infrastructure and data sources on varied types of platforms (on-prem, provide cloud, public cloud etc.). To resolve these challenges, it’s essential to leverage a hybrid data platform where data can be collected and ingested from any system and in turn delivered to any system or platform.

Cloudera provides end-to-end data life cycle management on a hybrid data platform, which includes all the building blocks needed to build a data strategy for trusted data in manufacturing. The key capabilities include ingesting data, preparing data, storing data, and publishing data, along with common security and governance capabilities across the data life cycle. Cloudera enables data transfer from anywhere to anywhere (private cloud, public cloud, on-prem, and platform agnostic), giving manufacturing the ability to use next-gen AI tools and applications on “trusted” data. Find out more about Cloudera Data Platform (CDP), the only hybrid data platform for modern data architectures supporting AI in manufacturing with data anywhere at Manufacturing at Cloudera.

Ganesh Hegde
Global Managing Director Industry Solutions - Manufacturing and Industrial
More by this author

Leave a comment

Your email address will not be published. Links are not permitted in comments.