06.11.2017Blogg

The Hiab hackathon – A recap: Making use of big data

In September, we took part in HIAB’s CargoHack3 coding competition, or Hackathon, in Huddiksvall, Sweden – and our team, Hii-op! took home the victory! We learned a lot about how to approach a company’s Big Data reserves and how to make best use of it, which is what this post is going to be about. These hints are the sum of our experience. They really do work – and the proof of that is the victory of our team’s solution as the crowd favourite as well as the winner of the entire competition!

Members of the Hii-op! Team: Tomi Lahtinen (left), Sami Köykkä (centre) and Mikael Ahonen (right). Source: https://www.hiab.com/en/HIAB/highlights/cargohack/

Time for Elevation was the theme for the Hackathon, and its goal was to explore new opportunities in digitalisation and connectivity to ensure easier and safer intelligent load handling. Our team focused on maximising the uptime of equipment. This is a look at what we learned during the competition.

Its starting point was a very common problem that organisations face: they have vast reserves of data that they don’t know how to use.

We were tasked with presenting a solution to maximise the uptime of loader cranes with the data collected from them.

source: https://www.hiab.com/en/HIAB/highlights/cargohack/

In the beginning, there were no complete ideas for the Hiab solution

As is usual when sifting through data, at the start it’s often hard to figure out what kind of useful information you can glean from it. Our team only knew that there is a wealth of data that loader crane sensors gather, and we wanted to use that for our solution. We had to proceed step by step.

I’ve listed the three main principles that you can use to find a solution for using massive amounts of data in a real way. They were the starting point for our solution, as well:

1. Data cannot give you an answer, if you’re not asking the right questions

Data can confirm assumptions and answer the questions you ask. It is no more than a tool to solve the real problem. Even Google won’t give you the right answer if you don’t use the right keywords.

Even though the idea was to make use of the loader crane data, that alone never became (and never should have become) the sole point. During the competition, many people asked us if it was possible to just feed all the data into a deep-learning neural network that would search for regularities in it. Unfortunately, even the best algorithms only start working when the problem is given to them in the right format.

2. Discuss and learn

Before you can solve the problem, you have to understand it very well. What is the problem? Is it worth solving? What solutions have been tried, and why have they been unsuccessful?

Instead of rushing to code straight away, we interviewed as many Hiab employees as we could. We asked what types of challenges they and their clients face in their daily work. Gradually we learned that incorrect use of the loader cranes was in fact the primary reason for breakdowns.
We listened and reflected on new information and finally, we had understood which of the crane’s many sensors showed the best picture of this incorrect usage.

3. Quick trials and gathering feedback

It’s better to have a non-feature-complete solution that works rather than an incomplete version of the perfect solution. The simple solution will at least work.

Based on the information we received, we had to try the devices for ourselves and to gather data on the built-in sensors. We performed various tests: good, average and poor lifts. Based on our tests, we were able to create a profile of the users and to encourage them in more careful use.

Since hackathons are always time-limited, our goal was to get something ready quickly, and then improve that if we had the time. We saw the benefits of this approach right away. We also presented our idea and the application’s appearance to our mentors and sought their feedback at all times. This process helped refine our solution.

Our final suggestion was to gamify use of the device. Devices send data when the crane is in use. We analyse the data and users can see, on their mobile device, how carefully they have used the machine in the form of different prizes and points.

There is a huge amount of data – how do we make use of it

To simplify the answer, let’s take the perspective of a house builder. You can rephrase the question as “There are a lot of bricks, how do we use them?” If all you want is to build a brick house, you can get to it and Bob’s your uncle. But if the builder wants a log cabin instead, bricks aren’t going to help. Data makes things possible, but it cannot be a goal and a starting point in and of itself.

Before you can even start discussing the data, you need to define the goals. Our second step was to find out background information, and the third was to draft a plan to reach our goals.

It’s only at this point that data enters the field. It might be that to reach your goal, you first have to change the construction material – and when data is involved, you may have to modify and restructure it.

Data makes it possible to set and track the key values related to your goal. Data can confirm an assumption. It can help optimise an operation when the goals are clear.

For example, you could make it a little bit more fun to control a loader crane and in doing that, save the owner a couple of maintenance visits per year.

Solita’s Hii-op! team consisted of the author of this post, Mikael, as well as Sami Köykkä and Tomi Lahtinen. Congratulations to every competition award winner! Read more about it here. Also, check out our new Machine Learning Engineer training programme for software developers. The aim is to find experienced software developers and architects who want to learn alongside their work how to use techniques and tools related to machine learning and artificial intelligent that will be needed in the future.