Sensor-Driven Predictive Vehicle Maintenance and Routing Problem with Time Windows

Iman Kazemian,Bahar Cavdar,Murat Yildirim

2024-12-06

Abstract:Advancements in sensor technology offer significant insights into vehicle conditions, unlocking new venues to enhance fleet operations. While current vehicle health management models provide accurate predictions of vehicle failures, they often fail to integrate these forecasts into operational decision-making, limiting their practical impact. This paper addresses this gap by incorporating sensor-driven failure predictions into a single-vehicle routing problem with time windows. A maintenance cost function is introduced to balance two critical trade-offs: premature maintenance, which leads to underutilization of remaining useful life, and delayed maintenance, which increases the likelihood of breakdowns. Routing problems with time windows are inherently challenging, and integrating maintenance considerations adds significantly to its computational complexity. To address this, we develop a new solution method, called the Iterative Alignment Method (IAM), building on the structural properties of the problem. IAM generates high-quality solutions even in large-size instances where Gurobi cannot find any solutions. Moreover, compared to the traditional periodic maintenance strategy, our sensor-driven approach to maintenance decisions shows improvements in operational and maintenance costs as well as in overall vehicle reliability.

Optimization and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to combine sensor - driven vehicle health prediction with the Traveling Salesman Problem with Time Windows (TSPTW) for a single vehicle, in order to optimize fleet operation and maintenance decisions. Specifically, the author points out that although current vehicle health management models can accurately predict vehicle failures, they fail to effectively integrate these predictions into actual operational decisions, limiting their application effectiveness. To solve this problem, the paper proposes a new framework, aiming to improve in the following ways:

1. **Combining sensor data with route planning**: Incorporate sensor - driven vehicle failure prediction into the Traveling Salesman Problem with Time Windows (TSPTW), thereby achieving more efficient fleet management and maintenance decisions.

2. **Introducing a dynamic maintenance cost function**: Define a dynamic maintenance cost function \( f_{to}(t) \), which is used to balance the trade - off between early maintenance (using maintenance resources too early) and delayed maintenance (increasing the risk of failure). This function is updated based on sensor data and takes into account the remaining life distribution prediction:
\[
f_{to}(t) = P(R_{to} > t)C_p + \frac{P(R_{to} < t)C_f}{\int_0^t P(R_{to} > z)dz + t_o}
\]
where \( C_p \) and \( C_f \) represent the costs of preventive maintenance and failure repair respectively.

3. **Proposing a new algorithm**: Develop a new algorithm called the Iterative Alignment Method (IAM), which utilizes the structural characteristics of the problem to solve the computational complexity problem. IAM can generate high - quality solutions in large - scale instances and can work effectively even when Gurobi cannot find any solutions.

4. **Performance evaluation**: Evaluate the performance of the proposed method in different operating environments through computational experiments, and compare it with traditional fleet management methods to verify its advantages in reducing operation and maintenance costs and improving overall vehicle reliability.

In summary, the goal of this paper is to provide a more intelligent and efficient fleet operation management method by integrating sensor - driven prediction and route planning, thereby reducing vehicle failure risks and operation interruptions while optimizing maintenance costs.