Predictive maintenance is no single tool, but a set of tools and procedures aimed at a goal according to Boeing. “For us, it represents an umbrella of activities to help operators turn unscheduled maintenance into scheduled activities,” summarizes Dawen Nozdryn-Plotnicki, director of advanced analytics for digital aviation and analytics at the aircraft manufacturer. The analytics chief says the approach is already well along, “yet with new advances in bigger data, more powerful analytical methodologies and newer airplane designs, we continue to have more to do.”
Boeing’s predictive umbrella covers a range of actions: maintenance strategy; maintenance planning, day-of-operations monitoring, execution, reliability analysis, maintenance and post-operations monitoring for feedback and improvement. Predictive maintenance alerts can influence both modification of aircraft design and scheduled maintenance, Nozdryn-Plotnicki notes.
And there are many ways predictive maintenance can be achieved at individual airlines. For example, Boeing offers airlines self-service analytics, consulting services to address specific needs, digital solutions that include both analytics and expertise and of course its turn-key maintenance, engineering, and supply chain program, Global Fleet Care.
However deployed, Boeing’s predictive tools are working. “When customers tell us our predictive algorithms helped reduce 80% of maintenance burdens on a problem, that makes our day,” Nozdryn-Plotnicki says.
The OEM has invested in further gains in designing the 737MAX and developing the 777X. Another major investment is developing more algorithms and technology platforms to exploit ever bigger and better aircraft data.
Boeing’s predictive services are already widely used. For example, Airplane Health Management conducts over two million calculations each hour for over 100 airlines flying 4,700 aircraft. And AHM is just one of the OEM’s predictive services.
Some of the biggest future challenges in predicting problems stems from the near perfection of aircraft performance, with technical reliability usually above 99%. That means most generated data is not useful for finding signals of failure. “The skewed nature of the data makes modelling really difficult, an interesting challenge to overcome,” Nozdryn-Plotnicki says. Impact also matters. She contrasts aviation with e-commerce industries that have high-volume, low-consequence environments: “where . . . getting results wrong isn’t so significant . . . you just get the wrong website or a bad movie recommendation.”