[论文解读] Two Birds with One Network: Unifying Failure Event Prediction and Time-to-failure Modeling
本文提出了一种统一的深度学习框架,采用RNN联合预测剩余使用寿命(RUL)和故障概率,结合生存分析(Deep Weibull-RNN)与多任务学习(MTL-RNN)。该方法在两项任务中均实现了稳定且高性能的预测,故障预测性能优于专用基线模型,同时保持了出色的RUL预测精度。
One of the key challenges in predictive maintenance is to predict the impending downtime of an equipment with a reasonable prediction horizon so that countermeasures can be put in place. Classically, this problem has been posed in two different ways which are typically solved independently: (1) Remaining useful life (RUL) estimation as a long-term prediction task to estimate how much time is left in the useful life of the equipment and (2) Failure prediction (FP) as a short-term prediction task to assess the probability of a failure within a pre-specified time window. As these two tasks are related, performing them separately is sub-optimal and might results in inconsistent predictions for the same equipment. In order to alleviate these issues, we propose two methods: Deep Weibull model (DW-RNN) and multi-task learning (MTL-RNN). DW-RNN is able to learn the underlying failure dynamics by fitting Weibull distribution parameters using a deep neural network, learned with a survival likelihood, without training directly on each task. While DW-RNN makes an explicit assumption on the data distribution, MTL-RNN exploits the implicit relationship between the long-term RUL and short-term FP tasks to learn the underlying distribution. Additionally, both our methods can leverage the non-failed equipment data for RUL estimation. We demonstrate that our methods consistently outperform baseline RUL methods that can be used for FP while producing consistent results for RUL and FP. We also show that our methods perform at par with baselines trained on the objectives optimized for either of the two tasks.
研究动机与目标
- 解决将RUL估计与故障预测作为独立任务、目标不一致且次优的问题。
- 通过利用长期RUL与短期故障概率之间的隐含关系,提升故障预测性能。
- 通过单一模型实现RUL与故障概率的联合预测,以支持一致的维护决策。
- 有效利用未发生故障的设备数据进行RUL估计,无需预训练或单独嵌入。
- 开发一个统一框架,在两项任务上均保持高性能,避免使用RUL预测进行故障预测时出现性能下降的问题。
提出的方法
- 提出Deep Weibull-RNN,一种基于生存分析的RNN,通过威布尔分布建模故障时间,网络输出通过约束变换参数化形状和尺度参数。
- 使用生存似然目标训练Deep Weibull-RNN,直接优化底层故障时间分布。
- 在MTL-RNN中引入一种新颖的约束损失,以利用未发生故障的设备数据进行RUL估计,同时联合训练故障预测。
- 采用多任务学习,使用共享的RNN编码器同时建模RUL与故障概率,共享表征捕捉退化动态。
- 使用ReLU和softplus变换确保训练过程中威布尔参数(如正尺度、形状)的有效性。
- 应用早停法和学习率调度以稳定训练,尤其针对对超参数更敏感的Deep Weibull-RNN。
![Figure 2 : An example from C-MAPSS, a turbofan engine degradation dataset from NASA [ 26 ] showing sensor data and the proposed window scheme used in this work. We first divide our data into windows (of size $\mathsf{w}$ ). The filter window $\tau_{f}$ is removed from the data and windows falling in](https://ar5iv.labs.arxiv.org/html/1812.07142/assets/x2.png)
实验结果
研究问题
- RQ1统一的深度学习模型能否在两项任务中均实现一致的高性能联合预测RUL与故障概率?
- RQ2通过威布尔似然显式建模故障时间分布,是否相比端到端训练能提升泛化能力与预测一致性?
- RQ3采用共享编码器与新颖约束损失的多任务学习,能否有效利用未发生故障的数据进行RUL估计,同时提升故障预测性能?
- RQ4统一模型在RUL估计与故障预测两项任务上的性能,与专用基线模型相比如何?
- RQ5与使用独立模型相比,使用同一模型进行RUL与故障概率预测时,其预测结果的一致性程度如何?
主要发现
- 所提出的MTL-RNN方法在Backblaze数据集上,预测故障概率与RUL值之间的Spearman等级相关系数达到-0.956,表明两者具有极强的一致性。
- DW-RNN在FD01数据集上达到-0.860的相关系数,表明RUL与故障预测输出之间具有很强的一致性。
- 在C-MAPSS数据集上,MTL-RNN在故障预测上的表现优于RUL-RNN(F1分数为0.85 vs. 0.71),同时保持了相近的RUL精度。
- 将RUL预测结果作为输入用于故障预测时性能极差(F1 < 0.5),凸显了独立模型之间的一致性问题。
- MTL-RNN框架在故障预测上达到最先进性能,同时保持了强大的RUL估计能力,优于为任一任务专门训练的基线模型。
- Deep Weibull-RNN模型由于显式建模生存函数,可在无需微调的情况下实现多时间范围的故障预测。

更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。