Incipient bearing damages on heavy haul vehicles can lead to detrimental disruptions in heavy haul operation and even to derailment of trains. The consequences are damage of the railway infrastructure, loss of freight and equity, and an interruption of traffic. This paper presents a novel method to detect and predict the onset of bearing damages using a combination of multiple way-side detectors. The method is based on a statistical normalization of detector information and subsequent generation of a bearing damage score time series reflecting the abnormal condition of a specific bearing on a rail car. The method is implemented in a cloud-based service solution which reflects each bearing as a digital twin and tracks the condition throughout the operation of a railcar. The solution is applied to a heavy haul operation in Scandinavia to quantify performance of the analytics in terms of true and false positives is currently ongoing.