Can Artificial Intelligence Boost PLM Productivity and Collaboration?

Product Lifecycle Management (PLM) covers various aspects of product conceptualization, design, production, maintenance, and upgrades. Creation, storage, and processing of data related to all these lifecycle phases is a core component of PLM.

Your organization’s success in using various PLM methods and software tools depends on a complex combination of the selected software tools, adoption by various teams, employee skills and collaboration. The general push for achieving success has been around defining and redefining the processes and improving the software tools.

However, a new viewpoint is emerging around access to information at the right time and collaboration across functions working simultaneously on different parts of this information creation process. This is a part of Industry 4.0 that differentiates itself from the prior focus of optimizations only around manufacturing and processes.

Modern product development teams are more likely to be a diverse group – multilingual and geographically dispersed. The questions around demography and diversity in product teams is not new. But the questions around productivity and collaboration in such teams are now much more relevant.

Why is Artificial Intelligence Relevant in This Context?

To answer this question, we’ll take a look at current, state-of-the art approaches, and what AI means for product design, enterprise knowledge and collaboration. But, before diving deeper into enterprise applications, a quick overview of a broader set of applications can help in understanding the common traits of such applications.

Applications like text translation, voice interaction, product recommendations and autonomous driving are familiar to most people who would not consider themselves tech-savvy.

And, regardless of whether we’re aware of it, applications like internet search, financial fraud detection, targeted marketing, scheduling of public transportation and customer churn prediction impact a vast majority of people.

What Does, “AI Embedded Inside an Application” Mean?

Consider a translation webpage or application as an example. You can input sentences or paragraphs, choose a target language, and get the translated text.

Now consider a scenario in which you use your mobile phone to automatically get translated versions of entire webpages.

This makes it possible to read local newspapers in your non-fluent languages. Or, to look at offers from restaurants during a vacation or business trip. This hassle-free experience becomes a go-to solution when accessing these types of information.

How Can AI Improve Information Access in the Product Design Workflow?

Your organization might have a large set of 3D models, related comments and reviews, manufacturing instructions and product feedback created over many years. The design and manufacturing may span multiple countries and often part or product adaptations have to be done.
When a new designer starts the product creation workflow, can they receive feedback for similar, prior work based on prior comments or reviews?

This would be a perfect case for reuse of knowledge preserved in the PLM system -initiating collaboration without rigid team definitions. Fortunately, language models and representations used in translation algorithms can be tweaked and used for this process.

Can we combine this information with public knowledge? For those familiar with stackoverflow, a popular platform for software engineering topics, the enterprise version allows combination of internal answers with related, public answers. A solution like this would change the nature of knowledge access for PLM users.

How Can Artificial Intelligence Improve Product Validation Efficiency?

After a product has been designed and a prototype is available, it goes through multiple test cycles, generating a lot of data. There can be multiple variables measured with different granularity at different sampling rates. For many organizations, several product variants exist, and such tests are carried across all variants.

Comparing successful and failed tests then linking them back to design is often a multi-step process involving data pre-processing, visualization, and inference.

Sometimes this is a repetitive task with high value for the product quality, but low value in terms of new knowledge. In other cases, the knowledge update is of high value each time. For both scenarios, machine learning algorithms can extract features and summarize results so that engineers only need to supervise.

Now, algorithms for anomaly detection and handling signals scaled in time are very mature and can be adapted to domain-specific needs. This can be used for automating repetitive tasks to improve efficiency, and for extracting specific patterns and sharing them in a simple visual form to improve knowledge sharing. Additionally, it can open new ways of collaboration between product design and manufacturing teams beyond predetermined team structures.

Can We Reuse Existing Structured Knowledge?

The examples presented above use one form of AI which has seen very high level of success in recent years – Machine Learning. However, AI is not one technique or even a group of only very closely related techniques. There can be many other components for an AI system and knowledge graphs is one of them.

For domains where a lot of structured knowledge exists in the form of entities and their relationships, it makes perfect sense to reuse this knowledge. We can combine these with learning to refine these structures or to cover information for which defining a structure is difficult. This has become a domain on its own, with the representations popularly referred to as knowledge graphs.

When you search for a term on Google, you often find a result with an infobox containing summary text in the top right corner of the page. Here, behind the scenes, Google’s knowledge graph is in action.

Are all the possibilities for productivity improvements known? Most probably not. But Artificial Intelligence, with all its methods, can open previously unexplored opportunities for making product creation more efficient.

In Our Experience…

We’re now open to consider different forms of collaboration that go beyond rigid constraints of defined teams, roles, and workflows. This is likely the future for PLM as a platform for creativity and information exchange.

Collaboration can lead to higher productivity and eventual success when skills for research, design, manufacturing, and distribution are spread around the world. The role of AI in this was evident for some companies that emerged winners during the pandemic.

As a peek into the future, the success of DeepMinds’s AlphaFold demonstrates that Artificial Intelligence will play an important role in achieving faster convergence towards solutions for difficult problems.

This is a very high-level view of possibilities and the role of AI in the future of PLM. To look at concrete solutions, algorithmic details, and deployment scenarios, check out our recent blog post, AI for enterprises – beyond prototypes and benchmark datasets.