Elevators operate all the time under varying people flow. For example in office buildings, the morning traffic is mainly from entrance floor to upper floors, while in the evening the traffic is mainly downwards when workers leave the office. Depending on the building type the similar daily traffic pattern may occur from one day to another, or it may also vary from day to day.
Once in a while some parameter or part need to be changed, and the impact of the change on elevator performance needs to be shown. Elevator performance is measured in terms of average call time. Call time of a landing call (call given by pressing an up or down button at an elevator lobby) is the time when the call is given until the serving elevator starts to decelerating to serve the call.
Call time is at least dependent on, in addition to elevator parameters, the number of calls and elevator availability. For example, an elevator can be under maintenance, or it is taken out-of-service for cleaning or for other purposes.
Your task is to develop a concrete machine learning application / concept that can be used to evaluate the impact of a parameter or part change on elevator performance taking into account the varying traffic conditions and other uncontrollable independent variables.
- Data: Measured data from KONE