Mining equipment reliability assessment models based on genetic algorithms and theoretical probability distributions

该提案是 NSERC 拨款的延续，该拨款试图开发和测试基于遗传算法 (GA) 的采矿设备可靠性评估方法，并以理论概率分布作为 GA 模型的适应度函数。遗传算法是基于自然选择、遗传和遗传学原理的随机搜索技术。选择 GA 的原因是，采矿设备的可靠性会随着时间的推移而变化，因为它依赖于几个协变量/因素（例如操作环境、维修质量）。这些因素结合起来对设备的可靠性功能产生复杂的影响。随着时间的推移，这种影响在某种程度上概括并继承了因素的个体特征。 总体思路是研究与经典可靠性分析相比，基于遗传算法的新型可靠性评估模型是否能够产生令人满意的解决方案。到目前为止，已经开发出基于指数分布的 GA 模型和基于对数正态分布的模型，并使用以地下矿井故障间隔时间形式收集的数据进行测试。已发现这两种概率分布适合历史故障数据。然而，数据表明收敛标准还不够，必须开发基于其他概率分布的模型。以实现更高的可预测性。该提案将首先关注新模型的开发和测试以及与现有 GA 模型的比较。我们之前的研究将矿山生产模拟研究与基于遗传算法的可靠性评估模型相结合，以分析移动设备故障对矿山生产力的影响，最终开发出了矿山模拟器原型。 该模拟器开发的下一阶段将需要实现更高的功能，以更好地代表现实生活中的矿山案例研究。

This proposal is a continuation of an NSERC grant which attempts to develop and test a mining equipment reliability assessment methodology based on Genetic Algorithms (GAs) with theoretical probability distributions as the fitness functions for the GAs models. GAs are stochastic search techniques based on the principles of natural selection, heredity and genetics. The reason for selecting GAs is the fact that the reliability of mining equipment changes over time due to its dependence upon several covariates/factors (e.g. the operating environment, quality of repair). These factors combine to create a complex impact on a piece of equipment's reliability function. This impact encapsulates and inherits to some degree the individual characteristics of the factors as they evolve over time. The overall idea is to investigate whether novel reliability assessment models based on GAs can produce satisfactory solutions compared to classical reliability analysis. So far, an Exponential distributions based GAs model and a Lognormal distributions based model have been developed and tested using collected data in the form of times between failures from underground mines. These two probability distributions have been found to fit historical failure data. However, the data indicated that the criteria of convergence does not suffice and it is imperative to develop models based on other probability distr. for achieving higher predictability. This proposal will first focus on the development and testing of new models and their comparison with the existing GAs models. Our previous research, on the combination of mine production simulation studies with GAs based reliability assessment models for analyzing the impact of mobile equipment failures on the productivity of mines, resulted in the development of a prototype mine simulator. The next phase in the development of this simulator will require the implementation of higher functionality to better represent real life mine case studies.