The remaining useful life (RUL) of transformer insulation paper is largely determined by the winding hot-spot temperature (HST). Frequently the HST is not directly monitored and it is inferred from other measurements. However, measurement errors affect prediction models and if uncertain variables are not taken into account this can lead to incorrect maintenance decisions. Additionally, existing analytic models for HST calculation are not always accurate because they cannot generalize the properties of transformers operating in different contexts. In this context, this paper presents a novel transformer condition assessment approach integrating uncertainty modeling, data-driven forecasting models and model-based experimental models to increase the prediction accuracy and handle uncertainty. The proposed approach quantifies the effect of measurement errors on transformer RUL predictions and confirms that temperature and load measurement errors affect the RUL estimation. Forecasting results show that the extreme gradient boosting (XGB) algorithm best captures the non-linearities of the thermal model and improves the prediction accuracy amongst a number of forecasting approaches. Accordingly, the XGB is integrated with experimental models in a Particle Filtering framework to improve thermal modelling and RUL predictions. Models are tested and validated using a real dataset from a power transformer operating in a nuclear power plant.

Critical technological systems exhibit complex dynamic characteristics such as time-dependent behaviour, functional dependencies among events, sequencing and priority of causes that may alter the effects of failure. Dynamic fault trees (DFTs) have been used in the past to model the failure logic of such systems, but the quantitative analysis of DFTs has assumed the existence of precise failure data and statistical independence among events, which are unrealistic assumptions. In this paper, we propose an improved approach to reliability analysis of dynamic systems, allowing for uncertain failure data and statistical and stochastic dependencies among events. In the proposed framework, DFTs are used for dynamic failure modelling. Quantitative evaluation of DFTs is performed by converting them into generalised stochastic Petri nets. When failure data are unavailable, expert judgment and fuzzy set theory are used to obtain reasonable estimates. The approach is demonstrated on a simplified model of a Cardiac Assist System.

Accurate diagnosis of power transformers is critical for the reliable and cost-effective operation of the power grid. Presently there are a range of methods and analytical models for transformer fault diagnosis based on dissolved gas analysis. However, these methods give conflicting results and they are not able to generate uncertainty information associated with the diagnostics outcome. In this situation it is not always clear which model is the most accurate. This paper presents a novel multiclass probabilistic diagnosis framework for dissolved gas analysis based on Bayesian networks and hypothesis testing. Bayesian network models embed expert knowledge, learn patterns from data and infer the uncertainty associated with the diagnostics outcome, and hypothesis testing aids in the data selection process. The effectiveness of the proposed framework is validated using the IEC TC 10 dataset and is shown to have a maximum diagnosis accuracy of 88.9%.

Condition-based maintenance strategies adapt maintenance planning through the integration of online condition monitoring of assets. The accuracy and cost-effectiveness of these strategies can be improved by integrating prognostics predictions and grouping maintenance actions respectively. In complex industrial systems, however, effective condition-based maintenance is intricate. Such systems are comprised of repairable assets which can fail in different ways, with various effects, and typically governed by dynamics which include time-dependent and conditional events. In this context, system reliability prediction is complex and effective maintenance planning is virtually impossible prior to system deployment and hard even in the case of condition-based maintenance. Addressing these issues, this paper presents an online system maintenance method that takes into account the system dynamics. The method employs an online predictive diagnosis algorithm to distinguish between critical and non-critical assets. A prognostics-updated method for predicting the system health is then employed to yield well-informed, more accurate, condition-based suggestions for the maintenance of critical assets and for the group-based reactive repair of non-critical assets. The cost-effectiveness of the approach is discussed in a case study from the power industry