Demonstrating the feasibility of applying machine learning models to railway condition data: Engine condition monitoring and failure prediction

铁路行业正在经历一场数字革命，这场革命创造了机会，改变了我们与更老的中年列车互动的方式。从历史上看，利用旧资产的数据在技术上和经济上都是一项挑战--这对于了解资产状况、优化维护和预测新出现的故障是非常宝贵的。尽管近年来的数字增强迅速加快了从这些列车中提取有价值数据的能力，但这对理解从许多不同系统产生的海量数据提出了新的挑战。该项目将结合资产知识、运营专业知识、数据科学能力和航空航天领域机器学习工具的应用，以测试使用人工智能和机器学习工具从火车发动机的大量近实时数据中提取见解的可行性。Chrome Angel Solutions和Amygda实验室正在与Angel Trains和Grand Central Trains合作，探索应用Amygda的创新机器学习工具从发动机数据中获取见解的可行性。与已有的方法相比，Amygda实验室使用无监督学习技术构建ML模型的独特方法能够更快地提供见解，通常将模型构建时间从几个月减少到几天，而不需要依赖领域知识，这可能是昂贵和耗时的。这项可行性研究将测试人工智能和机器学习工具是否可以通过检测数据中的关系并实现更快和更主动的决策，从而获得比当前数据科学方法更快、更深入的见解，从而实现更好的规划和更好的资产可用性。

The rail industry is undergoing a digital revolution, which has created opportunities to transform the way we are able to interact with older "mid-life" trains. Historically it has been technically and economically challenging to harness data from older assets - which is invaluable for understanding asset condition, optimising maintenance and predicting emerging failures.Whilst digital enhancements in recent years have rapidly accelerated the ability to extract valuable data from these trains, this presents a new challenge to make sense of the mass of data produced from many disparate systems.This project will combine asset knowledge, operational expertise, Data Science capability and the application of Machine Learning tools borne from the Aerospace sector to test the feasibility of using AI and Machine learning tools to extract insights from large amounts of near real-time data from train engines.Chrome Angel Solutions and Amygda Labs are working in collaboration with Angel Trains and Grand Central Trains to explore the feasibility of applying Amygda's innovative machine learning tools to derive insights from engine data. Amygda Labs' unique approach to building ML models using unsupervised learning techniques enables faster of delivery of insights when compared to established approaches, typically reducing the model build time from months to days without relying on domain knowledge, which can be costly and time-consuming.This feasibility study will test whether AI and Machine Learning tools can derive faster and deeper insights compared to current Data Science methods, by detecting relationships in the data and enabling faster and more proactive decision making, leading to better planning and improved asset availability.