Forecasting Asset Failures with Auto-regressive models: A Statistical Approach

A logistics company in South Africa is facing a challenge where approximately 25% of its business assets have been out of service for more than 55 to $\mathbf{1 0 0 0}$ days. To address this, a study was conducted to predict the number of assets that will parked in the future and identify the primary factors associated with the overall number of parked assets. The study used autoregressive models to forecast asset failures and collected data by developing a template that included the asset number, manufacturer, date of stoppage, reasons for stoppage, and reasons for not repairing or returning the asset to service. The results showed that the most significant number of assets were out of service due to unscheduled maintenance, followed by those affected by vandalism and collisions. The results show that the company is more likely to experience an average of eight monthly failures, with the upper limit falling between 34 and 68. The study concluded that the use of autoregressive models can effectively forecast asset failures and facilitate proactive maintenance and management approaches.