If You Build It (Using Machine Learning) Will They Come?

By August 15, 2016Design Reuse, PLM
If You Build It They Will come

Autodesk Claims Machine Learning Technology Will Transform 3D Engineering

Autodesk announced recently the availability of a shape-based search capability in A360. A blog article titled How Machine Learning Will Transform 3D Engineering describes the new capability, called Design Graph, as a “Google search-like functionality for the world of 3D models.”

Google search functionality is probably the wrong metaphor for 3D search. Web search is fundamentally text based, whereas searching for a part or a design requires a combination of textual and geometric terms and attributes, and sufficiently deep domain semantics. In fact, the blog article makes the very same argument later, describing Design Graph’s purpose to “identify and understand designs based on their inherent characteristics—their shape and structure—rather than by any labeling (tags) or metadata” (i.e. not Google-like).

Design Graph is not the first attempt to offer the engineering community a geometry-based search tool.  Geometry-based search technologies such as Siemens PLM’ Geolus and Dassault Systemes’s Exalead have been available in the market for 10 years. Analytic software companies aPriori Technologies and Akoya use shape recognition engines to calculate manufacturing costs and optimize inventories based on a part’s shape and features. Standalone search engines are also available from Enfinio, and 3DSemantix. And there is no shortage of academic work in this field.

Still, the engineering community remains unimpressed. The use of shape-based search in day to day engineering activities is tepid at best.

The benefits of finding and reusing parts and validated designs are obvious. Over the years, I have been writing about this topic extensively and you may want to follow some of the links at the bottom of this article. To summarize the benefits of design reuse quickly:  save valuable time during design and manufacturing volume ramp-up; eliminate redundant work and reduce costly errors and engineering changes; and reduce inventory.

Furthermore, as I repeatedly point out to manufacturing companies, design reuse is much more than using an existing part instead of designing a new one. Reusing a design means that you take advantage of existing processes and knowledge that are associated with this part, such as manufacturing work instructions, tooling, initial quality metrics, supplier performance data, and warranty claims forecast.

So why are engineering organizations not standing in line to exploit shape-based search as a key design practice?

I believe that 3D-search technology is yet another victim of business practices and culture standing in the way of technology innovation.

We educate design engineers and reward them based on their ability to invent and innovate, not reuse. In many product organizations, design reuse is not a stated goal that is recognized and rewarded engineering activity—although it should be. This is not to say that organizations are not looking for ways to lean their design and manufacturing processes and reduce costs; they do. But a crucial point that many software vendors and even management consultants often miss, is that while the investment and process changes take place withing the engineering organization, the benefits and savings occur mostly downstream, outside the sphere of influence, and, indeed, of interest, of engineering. This makes for a hard sell of any new technology!

If you want to sell this type of tools and process changes to engineers, don’t try to impress them with irrelevant (and often inaccurate) fanciful descriptions of machine learning and artificial intelligence the way the Autodesk blog article does. Tell them how about the new capabilities improve activities and outcomes they care about. And involve the other stakeholders, such as manufacturing, operations, procurement and service, as they understand and gauge the value of process changes from a complete product lifecycle and total lifetime cost point of view.

Additional Reading:


Image: Field of Dreams (1989)

  • James Gehan

    Completely agree with points in this article.

    Shape-based search “seems” like it should be appealing to many manufacturers to save time, increase re-use and lower product costs. It “seems” appealing to software vendors providing solutions as well. I have no vested interest in shape-based search.

    However, if we look at two business use cases, it becomes clearer to me where the potential business value may be for shape-based search:

    1. The general case for part re-use – If we step back and look at the “general” business needs for a manufacturer at the highest level – a part – is an implementation of a set of requirements (function, performance, cost, maintainability, availability, compliance, MOQ, etc).
    – Reqts – Finding an existing part that has the same/similar shape only confirms that it meets a significant subset of those requirements.
    – Verification – tasks to ensure the existing part meets the other design requirements are difficult/impossible to complete – if the original reqts of the existing part are not readily available.
    – Schedule risk – If there is more than 1 candidate part for re-use, how much energy should be dedicated to verifying these candidates before it may affect the design delivery schedule?
    – Stimulating for engineers – Let’s face it, creating a part design is often more stimulating than re-using someone else’s work – unless the engineer is appropriately rewarded for re-use.

    – Summary: Shape-based search not a clear fit – for the general case, driving higher levels of part re-use is a top-down organizational/product line goal which is not “uncovered” by shape-based search – rather it is a set of strategic decisions and focus on commonality goals.

    2. Engineer-to-Order/Application Engineering case for part re-use.
    – If we narrow the manufacturers’ use case to engineer-to-order businesses which often dictate significant project-specific parts with application-engineering teams, the business case for shape-based search may offer more clear benefit to the engineering team – reduced order-specific engineering costs.
    – In these ETO use cases, often times the categorization / classification of previous work is lower priority to working on the next project. Instead, the engineering teams often rely on the memory of their team to bid re-use from previous projects.
    – Compared with the general case above, shape-based search may enable these design teams to more effectively tap into previous similar order-specific part designs vs. rely on an engineer’s memory of previous projects/products.
    – Reqts – In many use cases of ETO products (such as elevators, trains, transformers, industrial applications, etc.), several categories of business requirements driving implementation designs of parts are similar project-to-project. Typically some parts must be adapted to fit some personalization or project-specific needs.
    – Verification – these tasks are still needed, but for some companies who do capture the order-specific requirements, the context of the previous order a part was designed for may offer a sufficient starting point for verification.
    – Schedule risk – While verification tasks are still needed, the ability to re-use or adapt an existing order-specific part design may directly reduce the application engineering time/risk on a project.
    – Stimulating for engineers – for application engineering teams, they create many project-specific parts. If they are able to re-use previous order parts, they may be rewarded with shortened design times for an order.

    – Summary: In some ETO/application engineering businesses, shape-based searching may provide a relatively robust approach to find parts from previous projects which can translate into shorter design time, quicker order delivery and offer the benefits in the rest of the organizations of the company.