Motor cars and gas turbines: Two ‘predictive downtime’ use cases from Hitachi
As we have seen already, through separate discussion of Hitachi’s approach to ‘predictive quality’ and ‘dynamic scheduling’, the digital transformation of industry is multi-faceted. There is a third way, apart from managing defects and bottlenecks, for industrial operatives to set about this change, the company reckons.
Greg Kinsey, vice president of its Hitachi Vantara digital transformation division, offers up a version of ‘predictive maintenance’ as a third basis for operational excellence. He does not care for the term, however, and Hitachi does not use it. “It’s over-used, and mis-used,” he says. Instead, Hitachi deals in ‘predictive downtime’, which it says describes the challenge, rather than the solution.
“It is about the availability of a factory. If a machine is broken, or people don’t show up to work, or the trucks don’t deliver the right materials, on time, then you have downtime,” explains Kinsey. Hitachi is working with manufacturing companies and other industrial operations to eliminate downtime, completely.
The process, familiar through the previous cases, is always the same: it starts as an exploratory pursuit in close collaboration with the client, before it finds it mark, and shifts hidden levers in the industrial systems and processes.
“It is all done in co-creation. We run failure and affects analysis. We draw a heat-map of all the data. We seek to understand where the key data sources are, and how they crossover. We understand the process, and ask people why they do things. Eventually, we start to build a hypothesis,” explains Kinsey.
Ultimately, the Hitachi sets a 3:1 threshold for its digital transformation projects; Kinsey’s team won’t start work, proper, until its hypothesis achieves such a return on investment. A minimum viable data set is born from the hypothesis, and a minimum viable algorithm is born from the data set, presenting a logic to predict downtime and, therefore, to adjust processes and schedule the maintenance of equipment.
“It gets complex quite quickly,” he says. Hitachi tends to run two-week ‘sprints’, where the theory is developed and tested, and repeated again, until a fine-tuned algorithmic logic is achieved. “It is a learning curve for the client,” says Kinsey. Everyone is wiser in the end; the systems are smarter, and the culture is brighter about the process of digital change.
Below, Kinsey describes Hitachi’s work with an unnamed European car maker. As the project progressed, and solutions were developed with the client, the penny dropped, finally. “‘This project is important because we knew we had to change,’ they told us. ‘We knew we were creative at designing cars, but not at making them,’ they said.”
Cultural change is crucial, of course, and invariably difficult, especially for well established firms in well-established markets. “In fact, their culture was the very opposite of innovative – it was about compliance, and not taking risks,” says Kinsey. “It’s the same for all these big car makers, across Europe – in Germany, Italy, Lithuania, France.”
He adds: “Injecting innovation into the factory floor is a very interesting challenge for them.” And so, the saga continues with Kinsey’s description of two ‘predictive downtime’ use cases.
USE CASE 3.1 | PREDICTIVE DOWNTIME (CARS)
“European car makers are quite advanced, as a rule. They have loads of data, but a lot of it is not being used, or at least not to solve problems. With one car maker, we set up on the shop floor, right across from the assembly line; each day a handful of our people and a handful of their people are together in a team, working on this project.
“There are three core processes on the production line: the first is the body shop, where you mould and weld metal into a car; the second is the paint shop, where you paint the vehicle; the third is the assembly, where you stick all the bits on, the seats and mirrors. We are involved in the assembly process, at the end of the line, and devised three drivers to hit that 3:1 ROI.
“The first is about ‘lost cars’, where a car falls out of takt on the production line. The cadence is interrupted and bottlenecks are created, and the car has to be put back on the production line later to be re-worked. The second thing is a machine breaks, and needs adjustment, or where a vat is out of glue and has to refilled. These create small maintenance incidents, which add up – and make the pit crew run around.
“If we can predict some of these things, we can stop them from happening, and stop all the running around and fire-fighting, and the cost that goes with that.
“The third drive is to eliminate downtime completely, and get that to zero. The challenge is to know enough about your factory that you will never be surprised by unplanned events – that you always know in advance. The big thing for the car industry is absolute certainty about uptime – just like with the predictive quality use cases, where companies want certainty about ingredients.”
USE CASE 3.2 | PREDICTIVE DOWNTIME (TURBINES)
“This conceot of predictive downtime is something we are exploring ourselves, too; we have experienced the same challenges in our in our power generation business in Japan. We make these huge gas turbines, loaded with IoT devices to gather data from them; we try to monitor their condition and predict their failure – and run ‘prognostics’ to recommend certain interventions to stop downtime before it happens.
“We have done the same with windmills. So we are taking our experience down, and applying it to factory lines in every sector.”
This is an excerpt from a wide-ranging interview with Hitachi Vantara vice president Greg Kinsey. For more from the interview, check out the links below.