Alberta Pacific Forest (Al-Pac) is one of Alberta’s largest forestry companies and produces pulp and wood products sold around the world. Their operation is supported by over 400 team members and has 3 main business areas:
Al-Pac is a natural adopter of innovative technologies and approached AltaML in 2019 to help
understand how an Applied AI Strategy can help maintain their competitive edge.
Working together, we started by identifying data strengths and challenges as well as team capabilities at Al-Pac. After an executive ML 101 and ideation session, the Woodlands operations were selected as their complexity and high costs would likely give the best early results.
Fuel for logging trucks is a significant expense in forestry operations with long distances
traveled from remote logging areas to the mill for processing. Optimizing fuel consumption
stood out as an ideal first use case.
Although truck onboard computers provide raw data about each truck’s driving patterns, the many variables affecting fuel usage (e.g. speed, distances, weather conditions, grade, load) make it difficult to assess driver behavior and its impact on fuel usage. The raw data generated by the onboard computer is simply too much for a human to consider, but this complexity is ideal for machine learning.
AltaML started by building and training an ML model using historical trip data found from the
truck’s onboard computers. The first task was to work with the logistics managers to identify
where data may be present to tell us more about good and bad driver trip behaviour. Then, we
applied unsupervised learning techniques to the dataset to uncover signals that implied trucks
were being driven less efficiently than necessary. Unsupervised learning is used when we don’t
have human-labelled data to train the model on, making it the right technique for this job.
The model worked to uncover anomalies in the driving data and identify the root cause of excessive fuel consumption and is actively being deployed at Al-Pac.