More on the team

IBM Researcher Nat Mills is applying his expertise in advanced analytics to help improve performance monitoring and maintenance scheduling for complex systems.

As part of the global push toward a “smarter planet,” more industries are taking advantage of advanced analytics capabilities for performance monitoring and systems management. For example, using condition-based monitoring for automotive fleets can help provide advance notice when a vehicle requires maintenance. The result is fewer breakdowns in the field, reducing the time and cost involved in vehicle retrieval and repair.

At IBM’s Watson Research Center, Nathaniel Mills is applying his expertise in reasoning techniques and visualization technology to provide Global Business Services (GBS) clients in the automotive industry with tools for mobile asset monitoring and management. “The goal is to understand both the component model and actual use of complex equipment to tailor vehicle maintenance based on projected need, rather than the passage of time,” says Nathaniel. “Using distributed agents and telematics to dialogue with automotive systems can help investigate transient problems that are hard to reproduce in the service bay. As part of the process, we have developed a process for compressing and encrypting large volumes of operational data, which we can use to compare current operation against historic performance, and for comparative analysis with other systems used in similar contexts. We are also researching how to balance collaborative analysis between on-platform agents in constrained environments, and off-platform analytics using historic data. In related research, we are working on temporal, semantic data management to address changes in configuration and sensor logic.”

Nathaniel notes that the increased use of sensors has proven a boon for remote monitoring, but at the same time has produced an abundance of low-level data that leave users struggling to glean actionable intelligence. Sensors can also occasionally malfunction or deliver faulty results, creating an additional level of complexity. And while autonomic computing can handle some of these issues without the need for human intervention, some of the more complex problems require specialized engineering resources to investigate, diagnose and resolve. “In addition to using automated analytics to search for known patterns in data that identify degrading or failed conditions, companies can also use mining and inductive rule generation to get their data to provide insights for improved engineering and/or service contract agreements,” says Nathaniel. “A ‘smarter planet’ requires tools to refine the level of detail presented to people for them to take action. Our research is intended to facilitate how people teach computers to reason about data details so automated action can be taken where appropriate, and when this is not possible, to present meaningful information to speed accurate, intelligent human interaction.”

Prior to joining IBM in 1997, Nathaniel ran his own company for 10 years. During his time with IBM, he has been instrumental in the development of several patented IBM Research assets, including the award-winning IBM Page Detailer, which has been used by thousands of companies to improve customer experience through improved Web content design and location. In addition to Page Detailer, Nathaniel has spearheaded other projects, including Web transaction performance metrics and the impact of server performance on client experience.

Nathaniel’s expertise has also extended to research in high volume e-commerce Web sites, working with the Agent Building and Learning Environment (ABLE), which focuses on applying pattern matching and other reasoning techniques to complex discounting strategies and sales promotions. This research introduced him to a semantic reasoning framework called IRIS: Information Representation, Inferencing and Sharing. “I combined the analytic and empirical reasoning from ABLE with the semantic reasoning from IRIS, as well as some visualization technology for data exploration to create the Parametric Analysis Center (PAC).”

With Nathaniel’s help, Research and GBS used the PAC framework in a First of a Kind (FoaK) project with the U.S. Army Tank Automotive Research, Development and Engineering Center (TARDEC) and a commercial trucking company. “The positive feedback from the FoaK allowed us to pursue asset commercialization,” he says. “We have licensed this framework for condition-based management of military vehicles and have also engaged many automakers for diagnostic/prognostic studies and embedded systems modeling. PAC is industry-agnostic and can be applied to automating analysis of large volumes of event and time series operational data.”

The PAC solution goes beyond the automotive industry, with applications in almost every industry that uses complex systems to provide customer services. “PAC is poised to leverage the analytic assets developed in Research by making them available for customer applications in a common authoring and execution framework to automate analysis. Condition-based maintenance can help companies provide consistently improving centralized support to distributed field personnel and/or directly to consumers,” says Nathaniel. “The next stage is to use semantics and modeling to adapt the unique characteristics of differently configured systems to more generic models that leverage more general purpose reasoning. This can also lead toward common command and control language and rules for problem diagnosis and problem avoidance or recovery.”

Nathaniel predicts that the uses for collaborative analytics between distributed components will expand with cloud computing and pervasive, high bandwidth communications. “Inexpensive higher powered processing combined with larger, faster, robust storage for embedded environments and higher numbers of network-addressable, intelligent devices will drive increased demands for smart data reduction and correlation of information closer to the source. The pipeline that was once relegated to protective movement of data will become a participant in analysis, transformation and selective distribution of secured information. The net effect is a faster reaction time to address emerging problems and greater flexibility to respond to changing conditions, such as weather, load or bandwidth availability. The smarter planet is an adaptive world driven by need and current status and opportunities. Distributed, well managed analytic frameworks will provide the underlying insight and direction required for empowering a smarter planet.”

Working closely with customers and understanding their needs is a major benefit of working with IBM Research, says Nathaniel. “Applying real world requirements to theoretical topics through applied research helps keep my work grounded and oriented to providing useful solutions.”