20% fewer complaints thanks to data-driven maintenance reports

Otis elevators are equipped with a large number of sensors that measure, among other things, whether the doors close properly and which alarm codes are sent. They also collect statistics such as the number of kilometers an elevator travels per month and how often the doors reopen. Otis asked us to develop a data model that generates a 'health value' for each elevator, indicating how well the elevator is functioning.

Approach

We initiated the development of the data model based on Otis's existing on-premise infrastructure.

• We combined sensor, alarm, and statistical data, ensuring data usability by cleaning up everything. The various types of data had varying definitions, values, and units. Using Python code, we transformed everything into monthly data and aligned the definitions.
• Subsequently, we programmed the data model based on time series. The model compares the current values of an elevator with its past and with the values of other elevators.
• Detected deviations are automatically converted into a numerical value we call the 'health value.' These health values are calculated for both individual components and the elevator as a whole.
• Based on the health values, elevators are marked as 'green' or 'yellow' using a threshold. 'Green' elevators are functioning properly, while a 'yellow' score may indicate the need for maintenance.
• The health values are checked by a Remote Engineer, who then instructs the maintenance technician with specific actions. The qualitative feedback from the Remote Engineer and the maintenance technicians, along with our analyses, is used to improve the model.

Although our model functioned as Otis desired, the next step in the process was to automate the data model and migrate it to the cloud. We restructured the data model within Azure using Databricks and PySpark. Data input and export were facilitated through Azure Data Factory pipelines, and stored within Storage Accounts managed by Otis. The model's output was returned via this pathway to the central database of Otis Netherlands, where it is accessible to the Remote Engineer.

Result

Twice a month, Otis receives an overview of the health values per elevator. This allows Otis to track which elevators may require maintenance. Customers receive an automatic maintenance report monthly, and Otis technicians address this during the next scheduled service. Due to this new approach, customer complaints have decreased by 20%.

The data model is operational, and the surrounding process is fully automated. Otis can now independently work with it and is not reliant on our expertise for the model's functioning.

Future

Otis elevators incorporate even more sensors and alarms than previously utilised. With this data, we can further enhance the model. Additionally, we aim to explore which components have more or less influence on the elevator's status. We will achieve this by incorporating scores, influenced by technician feedback, into the various types of data within the model.

Receive data insights, use cases and behind-the-scenes peeks once a month?

Sign up for our email list and stay 'up to data':