Predictive Maintenance on Tire Failure

Client information / Background

Spatialedge designed and implemented a predictive maintenance solution for a client in the mining industry. The solution used existing fleet data to determine when tires are likely to need replacing. It was necessary to use a sophisticated machine learning approach in order to solve this problem due to the complex lead times of these products as well as the dependencies between machines.

The Challenge

Operations at various sites relied heavily on machinery that utilized specialized, expensive tires. The unpredictability of tire failure posed a significant challenge, as a damaged tire will halt operations,  from a few weeks to up to three months due to the lead time required for procuring replacements. This downtime resulted in substantial income loss for the company.

The Solution

The solution involved developing a predictive model to forecast tire failures up to six months in advance, allowing for proactive management of tire rotations, maintenance and replacements. This model analyzed various factors, including the sites where tires were used, machine types, weather conditions, and other external factors, to predict when a tire was likely to fail. By understanding these predictions, the company could order tires well before potential failures, minimizing operational downtime. In addition, the model outputs could be used to determine the reliability of various tire vendors to assist in procurement.

Data and Approach

The predictive model utilizes a comprehensive dataset that includes information on tire usage across different sites, machine types, and operating conditions. This data encompasses  historical tire performance, including failure rates under various conditions, the types of terrain encountered, weather conditions, machine weights, and operational intensity.

Predictive Analytics

Using machine learning algorithms, involving regression analysis and more sophisticated techniques like neural networks, the model analyzes patterns in the historical data to identify key predictors of tire failure. These predictors could include factors like the average distance traveled, loads carried, terrain ruggedness, and weather exposure.

Forecasting Tire Life

The model generates predictions on when each tire is likely to fail, with forecasts extending up to six months into the future. This advanced warning system allows for strategic planning in tire procurement and replacement, ensuring that replacements can be ordered and received well before a potential failure might occur.

Implementation and Integration

The model's predictions are integrated into the operational planning processes, with dashboards or reports highlighting which tires are at risk of failing in the coming months. 

The predictive model also incorporates a sophisticated "range approach" to quantify potential savings under varying scenarios of decision-making. This approach calculates the savings spectrum based on the premise that if a user (in this context, the operational manager or procurement team) makes all the right decisions regarding when to replace tires, the savings will reach an optimum level. Conversely, if all the wrong decisions are made—such as delaying replacements too long or replacing tires prematurely—the savings will drop to a minimum. The model also accounts for any combination of decisions in between these extremes, offering a nuanced understanding of potential financial outcomes of different decisions.

This range approach enables a dynamic evaluation of decision-making impacts on operational costs and efficiency. By forecasting tire life with a degree of uncertainty and mapping out the financial implications of various decision-making paths, the model empowers users to understand not just the most likely outcomes, but also the best and worst-case scenarios. This insight facilitates more informed strategic planning and risk management, allowing for adjustments in procurement and maintenance schedules to optimize savings and minimize downtime.

The Result

The primary benefit was the drastic reduction in downtime associated with tire failure, significantly mitigating income loss due to operational halts. The predictive model enabled better planning and inventory management for tires, ensuring that replacements were available when needed without unnecessary overstocking.