Is the news of GE’s two months “time out” to fix problems at Predix an early indication that Industry 4.0 is overhyped?
Although Presenso is not officially part of the GE ecosystem, our solution was designed to integrate with the infrastructure of any PaaS vendor. In other words: we follow the IIoT industry closely without a vendor bias. Furthermore, engaging with leaders in digitalization and Industry 4.0 at some of the largest European industrial plants gives us a broad market perspective.
In this article, we address the following questions from the viewpoint of an Industrial Analytics for Predictive Asset Maintenance solution provider:
- Is Industry 4.0 over-hyped?
- Should industrial plants reconsider investments in IIoT infrastructure?
- What is the likely impact on Industrial Analytics for Predictive Asset Maintenance?
Admittedly, we have a commercial interest in these topics, but this article is not an advertorial for our solution.
Is Industry 4.0 over-hyped?
The short answer is “yes”, the long answer is a definite “no”.
It’s not hard to prove we are experiencing a bubble of some nature. An entire cottage industry of consultants, thought leaders and trade show exhibitors have emerged in a short time period. There is also reason to be skeptical of some of the more aggressive analyst forecasts. For example, here are some astonishing forecasts about the German economy from the Boston Consulting Group:
Productivity: Gains of €90 billion to €150 billion over the next 5 to 10 years.
Revenue: Additional revenue of €30 billion or 1% of GDP.
Employment: Gains of 6% over a 10-year period.
In preparing this article, our marketing team reviewed research surveys of C-level executives. Here’s our assessment of these surveys: Executives are asked about their 3-to-5 year plans for investing in IoT and/or Predictive Asset Maintenance. This back-of-the-envelope data is used as input into revenue and investment forecasts that gain credibility in the marketplace. Unfortunately, the attractive charts and credible looking reports are built on weak data sources.
Is our analysis over-simplistic? Perhaps. At the same time, it’s hard to believe economic forecasts of changes in GDP within relatively short-term timeframes.
Now for the “no.”
Industry 4.0 and Industrial Analytics for Predictive Asset Maintenance are unstoppable. The application of Machine Learning to the industrial domain is the confluence of the significant cost reductions in data transfer, computational power, and cloud storage as well as the application of Machine Learning to the industrial domain.
Industry 4.0 is based on operationalizing Big Data. Whereas before the science of Big Data was once an academic discipline, it has now been commercialized.
Returning to the topic of whether IIoT is over-hyped, the underlying trends are real. Although timing and economic impact may not match some of the optimistic forecasts, they will for sure have a tremendous impact.
Should industrial plants reconsider investments in IIoT infrastructure?
Investment decisions relating to an industrial cloud-based platform are complex. First, there is no consensus for common standards for PaaS (Platform as a Service) and there are inherent risks of making long-term commitments given the dynamic nature of the underlying technology.
More importantly, many industrial plants cannot rationalize an investment in IIoT infrastructure. Why? At a strategic level, for both executive and mid-management, the number of unknowns still exceed the number of knows.
McKinsey has analyzed the various initiatives that comprise Industry 4.0 using a ‘digital compass.’ They identify 8 basic value drivers and 26 IIoT levers. The value drivers include Time to Market, Resources, Quality and Asset Utilization. The IIoT levers include 3D-printing, Predictive Maintenance, Digital Control Management and Concurrent Engineering.
Industrial plants need to balance short-term revenue and yield goals and long-term strategies. Each of the McKinsey-defined value drivers and levers is compelling independently. However, it is neither realistic for a plant to implement multiple concurrent changes to production processes without sacrificing present day operations.
In terms of the question of reconsidering investments in IIoT infrastructure, infrastructure investment should be aligned to business processes. Over-extending infrastructure based on undefined strategic plans is financially risky. I am not advocating a larger or smaller investment in infrastructure. Instead, I recommend a recalibration of infrastructure investments based on realistic Industry 4.0 deployment scenarios.
In the case of Industrial Analytics for Predictive Asset Maintenance, there are several options in the market. On the one extreme, solutions such as the GE Digital Twin are based on the Predix architecture. However, there are different options (including but not limited to those provided by Presenso), where a core requirement for implementing a Predictive Asset Maintenance solution is real time access to sensor data.
As can be seen in the Architecture Diagram below, Automated Machine Learning does not require new IIoT Infrastructure.
For Unsupervised Machine Learning solutions such as Presenso (and others), an IIoT infrastructure platform is a nice-to-have. That’s an important consideration because Predictive Asset Maintenance is often the driving force for IIoT infrastructure deployment. For many industrial plants, using Machine Learning for rapid experimentation or simulation may not be a priority and the ROI may not be justifiable. Using Machine Learning to identify anomalous factory asset behavior and to remediate problems before they can occur is far more compelling in the short to medium term.
In summary, the question of investment in IoT infrastructure is a function of the industrial plant’s business requirements. In some cases, lack of infrastructure will prevent deployment of applications. However, there are still scenarios where IIoT infrastructure is merely a nice-to-have.
What is the likely impact on Industrial Analytics for Predictive Asset Maintenance?
In the last few months, we have written several articles explaining the limitations of the Digital Twin. As GE rationalizes its business and focuses on fewer industry verticals, it will become more challenging for them to convince undecided customers. This may not be a positive for GE’s 2017 financial results, but the overall category is unlikely to be affected.
One cannot disagree with GE’s premise that its Digital Twin offering is a robust solution based on advanced technology. The concept is powerful: the virtual clone of the underlying machine asset can detect and predict machine failure before it occurs. There are other applications such as conducting simulations of different manufacturing scenarios on the virtual clone.
Based on our conversation with customers, there are two major challenges that GE is facing:
First, deploying the Digital Twin is both expensive and labor intensive. The Digital Twin is a virtual clone of the physical machine asset. For the clone to be accurate, technicians use the original machine blueprints to create the clone. Any discrepancy between the blueprint and the machine (e.g., modifications due to repair work) need to be reflected in the Digital Twin. Over the lifetime of an asset, machine blueprints are not always accurately maintained.
Second, the Digital Twin uses Supervised Machine Learning algorithms that require human input from the reliability and maintenance facility staff. Deployment is an iterative process as the model learns the underlying asset behavior.
These limitations provide insights into GE’s challenges. To a large extent, GE is at an advantage selling to its existing customer base in energy, aviation and oil and gas. The problem for GE is that many of its industrial clients own assets that are close to end-of-life. Perhaps tomorrow’s oil rig or wind turbine will be sold in a bundle with its virtual clone, but it’s much harder to justify the resources to build a Digital Twin based on 30-year-old blueprints.
In summary, the impact of GE’s move to focus on a limited number of industries is unlikely to have a significant impact on the IIoT Predictive Asset Maintenance category. At this point, and in our opinion, we have not seen the underlying operational drivers for GE’s flagship technology – Predix and the Digital Twin – to generate the growth that it had previously anticipated.
Summary and Conclusion
The fall in GE’s stock price and the reduction in 2020 revenue forecasts are indicative of the disruptive change facing a company that the New York Times calls a “124-year-old start-up”. Regardless of the bumps along the way, Industry 4.0 and Industrial Analytics for Predictive Asset Maintenance are not passing trends. Beneath the hype, the quiet industrial revolution continues.