Industrial Analytics – Optimizing Operational Efficiency

Searching for AI in Google returns over four billion results (climate change returns only 200 million) and buzz words like analytics, condition monitoring, predictive maintenance and industry 4.0 are commonplace at conferences. How much is hype and how much is delivering measurable results?

Collecting and analyzing data in real time are the core of predictive analytics, a branch of AI. Managing operational costs, improving efficiency, extending equipment lifecycles’, and working to getting more out of connected equipment are key themes in industrial analytics built around AI platforms.

According to Zion Market Research,  the industrial analytics market is expected to jump to $37 billion by 2024, up from $12.8 billion in 2017, growing at a continued annual growth rate of 16.4%.

Analytics innovation

Artificial Intelligence or AI explained by function and outcome
Figure 1 Shows the branches of AI, courtesy of G2 Crowd

Over the last decade the industrial analytics market has transitioned from reactive to proactive and now to predictive and/or prescriptive as shown in the table.

Industrial analytics by type including company
Figure 2 shows the various types of analytics

Condition-Based Analytics (CbA)

The shift from proactive analytics scheduling to CbA has been adopted more slowly, due to the installation costs connected to machine learning requirements and required specialist skills. Despite the slow start in the sector, larger operations with a high cost of downtime have seen the benefit of investment. CbA is effective in industries where safety and reliability are key concerns. Advances in the technology now allow for mass data collection, processing and analysis, making CbA more accessible to the wider industry. For many users, CdA acts as enabler for wider predictive maintenance capabilities, helping to connect and define sensory control and asset visibility.

Founded in 2016, Onyx Insight provides condition monitoring software and services for wind turbines using hardware/software platform. The company analyses data from vibration CMS, SCADA and through lubrication analysis to produce inspection and maintenance reports. Core customers include Siemens, Vestas and General Electric.

Predictive Analytics (PdA)

The General Electric Predictive Maintenance 2018 report found that 83% of the companies interviewed have already invested in predictive analytics initiatives or plan to further invest in the next two years. While many of these are in pilot-phase, 2018 saw a broader transition towards system wide implementation. There is a strong case for all industrial players to implement solutions as obvious appellation specific solutions continue to emerge, leading to large revenue savings. Niche applications including food spoilage prediction and supply chain risk analysis, are examples. The Predikto CEO explained that ports experience around 1000 hours a year in downtime due to crane malfunctions. This is extremely costly for port operators, and predictive analytics is reducing downtime by 50%.

Founded in 2014, the predictive analytics SaaS platform, Uptake, has raised $263 million in equity to date. Uptake’s machine learning algorithms analyze data from equipment sensors, learning the operation patterns and identifying upcoming failure. A strong portfolio of corporate customers include Caterpillar, Berkshire Hathaway Energy subsidiaries and Codelco, where Uptake is deploying AI to monitor the health of mining equipment.

Prescriptive Analytics (PsA)

In prescriptive analytics the system takes charge and optimizes a system dynamically, driving a steady state. Entirely automated control can be a dicey path, as systems may be at risk of unexpected action. Speaking to Intelecy, an early-stage developer of machine learning predictive analytics software, we learned that this is an issue for critical assets, for which there is a lack of real-case test data on extreme condition specific operating states.

closed loop automation to improve performance
Figure 3 shows Vigilent’s closed loop automation system, courtesy of Vigilent

We spoke to Vigilent, a company that brings closed-loop optimization into specific applications, achieving net-profit in 2018. Their prescriptive closed-loop automation system controls airflow by dynamically predicting cooling requirements in data centers, before automating control to drive a steady-state. Vigilent’s advanced analytical capabilities develop predictive models for areas such as system reliability and energy capacity, so that operators can achieve optimized cooling management, reducing energy usage by up to 40%. We learned that through prescriptive analytics the company has provided a total of $50 million in annual savings across 625 customers’ data centers.

Large-scale success stories can already be seen across industry. Ameren Illinois, a US-utility, partnered with ABB to integrate the Ability™ Ellipse® Asset Performance Management solution horizontally into its network. This allowed the company to transition away from vertically integrated system silos and carry out preventative analytics with regular checkups to maintain its grid infrastructure. The solution offered a prescriptive approach to maintenance, which integrated advanced analytics into more than 1,200 substations, 4,500 miles of transmissions lines and 46,000 miles of distribution lines, and ultimately enabled Ameren to prioritize short- and long-term maintenance needs, increase productivity and see a return on investment from critical grid assets.

Keep an eye on analytics processing to the edge

IoT software platforms have been, and still are, the popular method for integrating analytics at scale. Element Analytics and Augury both have raised equity in the last 12 months, providing IoT platform software for digital industrial solutions to help original equipment manufacturers (OEMs) and corporates capture operational efficiencies.

Centralized cloud platforms are at risk of serious bandwidth problems as more assets are connected. They may also have increasing security concerns with data transmission over wireless networks. An emerging innovation involves pushing the processing of analytics to the edge, enabling higher fidelity analysis, lower latency, enhanced security and cost savings. Gartner predicts that by 2025, 75% of generate data will be processed to the edge. For CdA, PdA and PsA, the ability to accelerate processing is far more attractive. Foghorn, Halio Tech, Xnor AI and AlwaysAI have all closed Series A rounds in the past eight months.