4 ways generative AI addresses manufacturing challenges – IBM Blog

The manufacturing industry is in an unenviable position. Facing a constant onslaught of cost pressures, supply chain volatility and disruptive technologies like 3D printing and IoT. The industry must continually optimize process, improve efficiency, and improve overall equipment effectiveness.

At the same time, there is this huge sustainability and energy transition wave. Manufacturers are being called to reduce their carbon footprint, adopt circular economy practices and become more eco-friendly in general.

And manufacturers face pressure to constantly innovate while ensuring stability and safety. An inaccurate AI prediction in a marketing campaign is a minor nuisance, but an inaccurate AI prediction on a manufacturing shopfloor can be fatal.

Technology and disruption are not new to manufacturers, but the primary problem is that what works well in theory often fails in practice. For example, as manufacturers, we create a knowledge base, but no one can find anything without spending hours searching and browsing through the contents. Or we create a data lake, which quickly degenerates to a data swamp. Or we keep adding applications, so our technical debt continues to increase. But we are unable to modernize our applications, as logic that is developed over the years is hidden there.

The solution lies in generative AI  

Let’s explore some of the capabilities or use cases where we see the most traction:

1. Summarization

Summarization remains the top use case for generative AI (gen AI) technology. Coupled with search and multi-modal interaction, gen AI makes a great assistant.  Manufacturers use summarization in different ways.

They may use it to design a better way for operators to retrieve the correct information quickly and effectively from the vast repository of operating manuals, SOPs, logbooks, past incidents and more. This allows employees to focus more on their tasks and make progress without unnecessary delays.

IBM® has gen AI accelerators focused on manufacturing to do this. Additionally, these accelerators are pre-integrated with various cloud AI services and recommend the best LLM (large language model) for their domain.

Summarization also helps in n harsh operating environments. If the machine or equipment fails, the maintenance engineers can use gen AI to quickly diagnose problems based on the maintenance manual and an analysis of the process parameters.

2. Contextual data understanding

Data systems often cause major problems in manufacturing firms. They are often disparate, siloed, and multi-modal. Various initiatives to create a knowledge graph of these systems have been only partially successful due to the depth of legacy knowledge, incomplete documentation and technical debt incurred over decades.

IBM developed an AI-powered Knowledge Discovery system that use generative AI to unlock new insights and accelerate data-driven decisions with contextualized industrial data. IBM also developed an accelerator for context-aware feature engineering in the industrial domain. This enables real-time visibility into process states (normal/abnormal), alleviates frequent process obstructions, and detects and predicts golden batch.

IBM built a workforce advisor that uses summarization and contextual data understanding with intent detection and multi-modal interaction. Operators and plant engineers can use this to quickly zero in on a problem area. Users can ask questions by speech, text, and pointing, and the gen AI advisor will process it and provide a response, while having awareness of the context. This reduces the cognitive burden on the users by helping them do a root cause analysis faster, thus reducing their time and effort.

3. Coding Assistance

Gen AI also helps with coding, including code documentation, code modernization, and code development. As an example of how gen AI helps with IT modernization, consider the Water Corporation use case. Water Corporation adopted Watson Code Assistant, which is powered by IBM’s gen AI capabilities, to help their transition into a cloud-based SAP infrastructure.

This tool accelerated code development by using AI-generated recommendations based on natural language inputs, significantly reducing deployment times and manual labor. With Watson Code Assistant, Water Corporation achieved a 30% reduction in development efforts and associated costs while maintaining code quality and transparency.

4. Asset Management

Gen AI has the power to transform asset management.

Generative AI can create foundation models for assets. When we must predict multiple KPIs on the same process or there is a fleet of similar assets. It is better to develop one foundation model of the asset and fine-tune it multiple times.

Gen AI can also train for predictive maintenance. Foundation models are very handy if failure data is scarce. Traditional AI models need lots of labels to provide reasonable accuracy. However, in foundation models, we can pretrain models without any labels and fine-tune with the limited labels.

Also, generative AI can provide technician support and training. Manufacturers can use gen AI technologies to create a training simulator for the operators and the technicians. Further, during the repair process, gen AI technologies can provide guidance and generate the best repair procedure.

Build new digital capabilities with generative AI

IBM believes that the agility, flexibility, and scalability that is afforded by generative AI technologies will significantly accelerate digitalization initiatives in the manufacturing industry.

Generative AI empowers enterprises at the strategic core of their business. Within two years, foundation models will power about a third of AI within enterprise environments.

In IBM’s early work applying foundation models, time to value is up to 70% faster than a traditional AI approach. Generative AI makes other AI and analytics technologies more consumable, which helps manufacturing enterprises realize the value of their investments.

Build new digital capabilities with generative AI