Unknown Failure Research Articles

Slow feature-based feature fusion methodology for machinery similarity-based prognostics

Similarity-based prediction methods utilize degradation trend analysis based on degradation indicators (DIs). These methods are gaining prominence in industrial predictive maintenance because they effectively address prognostics for machines with unknown failure mechanisms. However, current studies often neglect the discrepancies in degradation trends when constructing DIs from multi-sensor data and lack automatic normalization of operating regimes during feature fusion. In this study, a feature fusion methodology based on a signal-to-noise ratio metric that leverages slow feature analysis (SFA) is proposed. This customized metric utilizes SFA to quantify degradation trend discrepancies of constructed DIs, while automatically filtering out the effects of multiple operating regimes during feature fusion. The effectiveness and superiority of the proposed method are demonstrated using publicly available aero-engine and rolling bearing datasets.

Joint Learning of Failure Mode Recognition and Prognostics for Degradation Processes.

The paper aims to develop a joint learning method for failure mode recognition and RUL prediction of operating units. Specifically, the developed method addresses a challenging issue in practice, i.e., how to effectively conduct failure mode recognition and RUL prediction as a joint task based on interpretable extracted degradation features from multiple sensor signals. To implement this method in practice, four steps are included as follows: First, collect multiple sensor signals, failure time, and failure modes of historical units. Second, construct the joint learning model based on features extracted from sensor signals by considering the degradation mechanism. Third, estimate model parameters using the data of historical units. Fourth, recognize the failure mode and predict the RUL of an in-service unit. Since the proposed method is a data-driven neural network with flexible model structure that considers complex data relationships, it is expected to be applicable to many practical situations and use cases, especially for manufacturing systems with complex structures and unknown failure thresholds.

Experimental investigation of machine-learning-based soft-failure management using the optical spectrum

The demand for high-speed data is exponentially growing. To conquer this, optical networks have undergone significant changes, getting more complex and versatile. The increasing complexity necessitates that the fault management be more adaptive to enhance network assurance. In this paper, we experimentally compare the performance of soft-failure management of different machine learning algorithms. We further introduce a machine-learning-based soft-failure management framework. It utilizes a variational autoencoder-based generative adversarial network (VAE-GAN) running on optical spectral data obtained by optical spectrum analyzers. The framework is able to reliably run on a fraction of available training data as well as identify unknown failure types. The investigations show that the VAE-GAN outperforms the other machine learning algorithms when up to 10% of the total training data are available in identification tasks. Furthermore, the advanced training mechanism for the GAN shows a high F1-score for unknown spectrum identification. The failure localization comparison shows the advantage of a low complexity neural network in combination with a VAE over established machine learning algorithms.

A data-driven approach for condition-based maintenance optimization

Developments in sensor techniques enable the continuous monitoring of the health of an operating system. The resulting condition data provides an opportunity for better prediction of failures and thereby for improving maintenance decisions. In this study, we consider condition-based maintenance for a single unit with an unknown, non-decreasing deterioration process and unknown failure behavior. Building on, but different from the existing maintenance optimization literature, we present the first fully data-driven approach, where the condition threshold triggering maintenance is based purely on past condition data and failures. Numerical results for a gamma deterioration process show that the maintenance threshold resulting from our data-driven approach converges to the optimal threshold. The threshold is set higher during the initial runs-to-failure, and this helps to explore the deterioration process. An encouraging result is that the convergence is especially fast during the first few runs-to-failure so that the expected cost rate quickly converges to the minimum cost rate.

Robust event-triggered distributed fusion for multi-sensor systems with unknown failure rates

A robust event-triggered distributed fusion algorithm is investigated in this paper for multi-sensor systems with unknown failure rates. A detection technique based on standard Gaussian distributed filtering innovation is designed and applied to judge whether the measurement is failed. This filtering innovation can also be used to construct the event-triggered condition. Specifically, the event condition is not triggered if the innovation is below the lower event-triggered threshold and the measurement is regarded as the failure measurement if the innovation exceeds the higher threshold. In the above two cases, the sensor measurement data is not transferred to the local estimator; otherwise, it will be transferred. Then, the sequential fast covariance intersection (SFCI) fusion algorithm is used for local estimation fusion. Besides, to analyze the estimation performance, sufficient conditions are given to demonstrate the boundness of the local estimation and fusion estimation covariance. Finally, a simulation example is given to show the usefulness of the presented algorithm.

The Method Based on Clustering for Unknown Failure Diagnosis of Rolling Bearings

The effective fault diagnosis is an important guarantee for the safe and stable operation of mechanical system. Nowadays, the most of fault diagnosis methods are for known classes. However, unknown fault usually occurs in the process of mechanical operation, and may be wrongly divided into defined classes, which increases the risk of mechanical shutdown and the difficulty of maintenance. Therefore, in order to achieve high-accuracy diagnosis with unknown fault, a method based on transfer learning and deep transfer clustering is proposed, named Transfer Clustering Calculation Center Points (TCCP). TCCP completes the task of fault diagnosis with unknown class by calculating the feature space distribution between sample to be tested and the known samples, and further improves the accuracy by matching the target distribution. In two open-source bearing vibration datasets and an acoustic emission dataset of the self-made bearing, five bearing diagnostic models are used to compare the unknown failure diagnosis. The average accuracy with unknown failure by TCCP is 97.08%, 89.78%, and 95.24%, respectively. The accuracy is 24.45%, 23.16% and 10.08% higher than the average accuracy without TCCP. TCCP can get better diagnostic results under both vibration signals and audio signals. TCCP has predominant universality and generalization capabilities, and provides an effective approach for the identification of unknown failure.

Failure patterns for thymic carcinoma with completed resection and postoperative radiotherapy

Background and purposeWe aimed to evaluate the pattern and risk factors of disease failure in patients with thymic carcinoma after complete resection and postoperative radiotherapy (PORT). Materials and methodsWe retrospectively analyzed 127 patients with thymic carcinoma who underwent PORT after complete resection between 2003 and 2020 in our center. Data on clinical characteristics and radiation fields were collected. Failure patterns were recorded as locoregional (disease appearing in the tumor bed or regional lymph nodes), pleural, or distant failure (including hematogenous metastasis and nonregional lymph node metastasis). ResultsAll patients underwent tumor bed irradiation. During a median follow-up period of 64 months, disease failure was observed in 51 patients (40.2 %). The 5-year disease-free survival (DFS) and overall survival rates were 58.9 % and 85.0 %, respectively. The sequence of failure patterns was distant (n = 41, 32.3 %), pleural (n = 28, 22.0 %), and locoregional failure (n = 19, 15.0 %). Of the locoregional failure patients, failures occurred in-field in three patients (2.4 %), marginal failure in one patient (0.8 %), out-of-field failure in nine patients (7.1 %), synchronous in-field and out-of-field failures in two patients (1.6 %), synchronous marginal and out-of-field failures in two patients (1.6 %), and unknown failure fields in two patients (1.6 %). Multivariate analysis showed that Masaoka stage (hazard ratio [HR], 3.88; p = 0.000) and adjuvant chemotherapy (HR, 0.47; p = 0.015) were independent predictors of DFS. ConclusionThe most common failure was distant, the Masaoka stage and adjuvant chemotherapy were independent predictors of DFS, and low locoregional failure-supported tumor bed irradiation was sufficient for patients with thymic carcinoma after complete resection.

DeepRepair: Style-Guided Repairing for Deep Neural Networks in the Real-World Operational Environment

Deep neural networks (DNNs) are continuously expanding their application to various domains due to their high performance. Nevertheless, a well-trained DNN after deployment could oftentimes raise errors during practical use in the operational environment due to the mismatching between distributions of the training dataset and the potential unknown noise factors in the operational environment, e.g., weather, blur, noise, etc. Hence, it poses a rather important problem for the DNNs’ real-world applications: how to repair the deployed DNNs for correcting the failure samples under the deployed operational environment while not harming their capability of handling normal or clean data with limited failure samples we can collect. In this article, we propose a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">style-guided data augmentation for repairing DNN in the operational environment</i> , which learns and introduces the unknown failure patterns within the failure samples into the training data via the style transfer. Moreover, we further propose the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">clustering-based failure data generation</i> for much more effective style-guided data augmentation. We conduct a large-scale evaluation with 15 degradation factors that may happen in the real world and compare with four state-of-the-art data augmentation methods and two DNN repairing methods. Our technique successfully repairs three convolutional neural networks and two recurrent neural networks with averaging 62.88% and 39.02% accuracy enhancements on the 15 failure patterns, respectively, achieving higher repairing performance than state-of-the-art repairing methods on the most failure patterns with even better accuracy on clean datasets.

Dealing with prognostics uncertainties: Combination of direct and recursive remaining useful life estimations

Data-driven prognostics and health management is an important part of the future industry. It allows the detection of system faults and estimation of its remaining useful life (RUL) to anticipate failures and schedule appropriate prescriptive maintenance actions. Moreover, with the development of data acquisition tools in such industries, it is possible to collect large amount of data from similar systems, which prompts the use of machine learning algorithms for efficient estimation of the system RUL. However, due to different type of faults and degradation rates that can occur in real processes, there exists high variability of the end-of-life (EoL) time of each system degradation trajectory, making more difficult to fix a failure threshold and consequently generate several uncertainties in the estimation of RUL. To address this situation, this paper proposes a new data-driven approach for estimating the system RUL when dealing with the variability of degradation trends and unknown failure thresholds. Particularly, the proposed approach combines two RUL techniques, recursive and direct RUL estimation. First, the historical collected raw data are fed into a processing algorithm to construct prognostic health indicators (HIs) and choose the one that characterizes well the system’s degradation trajectory. This latter indicator allows identifying the high and low EoL amplitude values from the historical data and is used to build a recursive prediction model that estimates in long-term the degradation trend evolution. After that, the trained model forecasts every degradation trend which EoL amplitude is less than the predefined high value. Thus, a set of possible RULs of each trajectory can be predicted. Finally, the ensemble of the derived RULs and their HI trajectories are fused to directly estimate the final RUL. The proposed approach is applied to a subway door system with multiple degradation scenarios while taking into account different operating conditions.

Impedance Characterization of OECT Behavior in Enzyme-Embedded Conductive Polymer Matrix

Latest generation glucose sensors use small-form wearables which provide continuous data. Thesedevices represent the commercial forefront aiming towards personalized and real-time point-of-carebiomedical applications. Currently, device lifetimes are limited due to unknown failure mechanisms andex-situ calibration strategies. Creating a method for in-situ sensor diagnostics will allow a bottom-upapproach for designing enzyme-supporting matrixes and calibration strategies inherent to biosensors. Tothis end, this work seeks to enhance the understanding of performance, stability, and failure mechanismsof biologically based electrochemical transistors using impedimetric measurements.Here, we reveal the fundamental interplay between glucose oxidase loading and the electronic propertiesof a commonly used PEDOT:PSS matrix as an encapsulating medium. Fundamental to these studies wasthe discovery of conductivity enhancing strategies compatible with preserving enzyme activity. Theenzyme-embedded conductive polymer matrix was used in an organic electrochemical transistor probedvia electrochemical impedance (EI) to determine the causes of performance loss over sensor lifetimes.Supporting spectroscopies aided in detailing film morphology and electrical property changes associatedwith sensor behavior.

Planar Array Diagnosis Based on Bayesian Learning With a Bernoulli–Gaussian Prior Model

A new approach is proposed for the planar array failure diagnosis from a set of far-field measurements. Under the framework of Bayesian learning, a Bernoulli–Gaussian (BG) model is constructed to exploit the binary prior information of the unknown failure elements. Subsequently, the variational expectation maximization (EM) strategy is used to perform the Bayesian inference. Moreover, the damped generalized approximate message passing (GAMP) is further adopted to circumvent the matrix inversion in the variational EM procedure, thus making the overall process computationally very efficient. A set of representative numerical experiments are provided to validate the accuracy, efficiency, and flexibility of the proposed method.

Efficacy and safety of atezolizumab concurrent with radiotherapy in patients with muscle-invasive bladder cancer: An interim analysis of the ATEZOBLADDERPRESERVE phase II trial (SOGUG-2017-A-IEC(VEJ)-4).

4588 Background: Combined-modality treatments are bladder-preserving alternatives for patients (pts) who are not candidates for radical cystectomy by medical reasons, refusal, or patient´s choice. Immune therapies seem to potentiate tumor-specific immune response induced by radiotherapy (RT). Combination of RT with anti-PD-1/PD-L1 therapy appears safe and there are signs of promising activity. The aim of this study is to assess the efficacy and safety of atezolizumab (ATZ) concurrent with external beam radiotherapy (EBRT) for the treatment of muscle-invasive bladder cancer (MIBC) with bladder preservation intent. Here we present an interim analysis. Methods: This is an open, multicenter, and phase II trial, sponsored by SOGUG, in pts with confirmed diagnosis of MIBC in clinical stages cT2-T4a N0 M0 who are not candidates for radical cystectomy. Treatment consists of 6 doses of ATZ 1200 mg IV every 3 weeks, starting on day 1 of EBRT, and 60 Gy of RT in 30 fractions over 6 weeks at 2 Gy/day. The primary endpoint of the study is pathological complete response (pCR) defined as a response of grade 5 according to Miller and Payne criteria, 1 to 2 months after the last dose of ATZ. A planned interim analysis has been performed (data cut-off date: November 2021) on the primary endpoint to avoid exposure to ineffective treatment according to the minimax two-stages Simon’s design (stopping rule: 9 out of the first 13 evaluable pts should achieve pCR). Incidence of adverse events (AE) and serious AE (SAE) has been also secondarily assessed. Results: From September 2019 to November 2021, 39 pts were screened, of whom 13 were excluded due to non-compliance with eligibility criteria. Thus, the evaluable population consisted of 26 pts. The safety analysis was performed in 22 pts who had received at least one dose of ATZ. 14 pts were assessed pathologically and, thus, included in interim efficacy analysis (median age: 78.6 years; clinical stage: 71.4% T2a, 14.3% T2b, 7.1% T3a, 7.1% T3b). All 14 pts had achieved pCR at the cut-off date. 20/22 (91%) pts experienced at least one AE, with asthenia (11 pts), diarrhea (9 pts), and urinary tract infection (4 pts) being the most common. 9 SAEs were reported in 7 (32%) pts (bacteriemia, COVID-19 infection, depressed LVEF, unknown origin fever, hepatic toxicity, kidney failure, rectorrhagia, respiratory infection, and urinary sepsis). 6 (27%) pts suffered AEs leading to treatment discontinuation. No AEs leading to death occurred. 17 pts with available data on survival were alive at the cut-off date. Conclusions: Interim results suggest that ATZ combined with EBRT is a feasible and effective treatment in terms of pCR, with a manageable safety profile. The final results from this trial will provide information about its effects on clinical outcome, including survival and updated safety findings. Clinical trial information: NCT04186013.

Fracture toughness and fractal analysis of ceramic benchmark materials.

Previous studies have reported various methods of measuring the fracture toughness of brittle ceramics. The purpose of the present research was to use a new method of fractal dimension measurement on benchmark materials (silica glass, Viosil SX, Shin-Etsu, n = 13, and silicon nitride standard reference material, SRM2100, NIST, n = 10), to compare the fracture toughness calculated using different methods, and to study the effect of noise filtering on the fractal dimension and fracture surface roughness. Fracture toughness was determined using surface crack in flexure method according to ASTM C1421 and fractal analysis method. Fractal dimension was determined using the Minkowski Cover algorithm on atomic force microscope scans of epoxy replicas of fracture surfaces. The mean ± standard deviation of fracture toughness using surface crack in flexure method and fractals method were 0.97 ± 0.18 MPa·m1/2 and 1.03 ± 0.07 MPa·m1/2 for silica glass and 4.62 ± 0.14 MPa·m1/2 and 2.54 ± 0.07 MPa·m1/2 for silicon nitride, respectively. The mean ± standard deviation of fractal dimension was 2.17 ± 0.03 for silica glass and 2.13 ± 0.01 for silicon nitride. The mean ten-point roughness (Rz) before and after noise filtering was 89 ± 102 nm and 87 ± 101 nm for silica glass and 355 ± 132 nm and 357 ± 134 nm for silicon nitride, respectively. Noise filtering had no significance on the fracture surface roughness of the two materials. The newly developed fractal analysis method can be used to predict the baseline fracture toughness of specimens with unknown failure stress.

Active learning-assisted semi-supervised learning for fault detection and diagnostics with imbalanced dataset

Data-driven Fault Detection and Diagnostics (FDD) methods often assume that sufficient labeled samples are class-balanced and faulty classes in testing are precedent or seen previously during model training. When monitoring a large fleet of assets at scale, these assumptions may be violated: (I) only a limited number of samples can be manually labeled due to constraints of time and/or cost; (II) most of the samples collected in the engineering systems are under normal conditions, leading to a highly imbalanced class distribution and a biased prediction model. This work presents a robust and cost-effective FDD framework that integrates active learning and semi-supervised learning methods to detect both known and unknown failure modes iteratively. This framework allows to strategically select the samples to be annotated from a fully unlabeled dataset, while labeling cost is minimal. Specifically, a novel graph-based semi-supervised classifier with adaptive graph construction is developed to predict labels with imbalanced data and detect novel classes. We designed a multi-criteria active learning sampling strategy to select the most informative samples from unlabeled data in order to query minimal number of labels for classification. We tested the framework and algorithms in three synthetic datasets and one real-world dataset of vehicle air intake systems, and demonstrated the superior performance compared to the state-of-the-art methods for fleet-level FDD.

A One-Shot Learning Approach for Similarity Retrieval of Wafer Bin Maps With Unknown Failure Pattern

Due to the increasing complexity of semiconductor manufacturing, new process failures occur and cause unknown types of defective patterns on wafer bin maps (WBMs). Similar patterns have a high probability of being induced by the identical process failure. The commonality analysis of unknown patterns can help to track down the new faults. To explore the similarity between defective WBMs, a retrieval method based on deep learning is adopted to detect WBMs with unknown failure patterns. Since there are few available WBMs with unknown patterns at the initial troubleshooting stage, more WBMs with unknown patterns are generated by morphology transformation. The deep retrieval model is trained by real and generated WBMs, which has a similar structure to the generative adversarial networks (GANs) and can create more WBMs to provide more information. Experiments on a real-world dataset have demonstrated that the proposed method has a high precision for measuring WBM similarity and retrieving similar unknown pattern wafers.

Modelling and condition-based control of a flexible and hybrid disassembly system with manual and autonomous workstations using reinforcement learning

Remanufacturing includes disassembly and reassembly of used products to save natural resources and reduce emissions. While assembly is widely understood in the field of operations management, disassembly is a rather new problem in production planning and control. The latter faces the challenge of high uncertainty of type, quantity and quality conditions of returned products, leading to high volatility in remanufacturing production systems. Traditionally, disassembly is a manual labor-intensive production step that, thanks to advances in robotics and artificial intelligence, starts to be automated with autonomous workstations. Due to the diverging material flow, the application of production systems with loosely linked stations is particularly suitable and, owing to the risk of condition induced operational failures, the rise of hybrid disassembly systems that combine manual and autonomous workstations can be expected. In contrast to traditional workstations, autonomous workstations can expand their capabilities but suffer from unknown failure rates. For such adverse conditions a condition-based control for hybrid disassembly systems, based on reinforcement learning, alongside a comprehensive modeling approach is presented in this work. The method is applied to a real-world production system. By comparison with a heuristic control approach, the potential of the RL approach can be proven simulatively using two different test cases.

Parameter Estimation Based on Censored Data under Partially Accelerated Life Testing for Hybrid Systems due to Unknown Failure Causes

In general, simple subsystems like series or parallel are integrated to produce a complex hybrid system. The reliability of a system is determined by the reliability of its constituent components. It is often extremely difficult or impossible to get specific information about the component that caused the system to fail. Unknown failure causes are instances in which the actual cause of system failure is unknown. On the other side, thanks to current advanced technology based on computers, automation, and simulation, products have become incredibly dependable and trustworthy, and as a result, obtaining failure data for testing such exceptionally reliable items have become a very costly and time-consuming procedure. Therefore, because of its capacity to produce rapid and adequate failure data in a short period of time, accelerated life testing (ALT) is the most utilized approach in the field of product reliability and life testing. Based on progressively hybrid censored (PrHC) data from a three-component parallel series hybrid system that failed to owe to unknown causes, this paper investigates a challenging problem of parameter estimation and reliability assessment under a step stress partially accelerated life-test (SSPALT). Failures of components are considered to follow a power linear hazard rate (PLHR), which can be used when the failure rate displays linear, decreasing, increasing or bathtub failure patterns. The Tempered random variable (TRV) model is considered to reflect the effect of the high stress level used to induce early failure data. The maximum likelihood estimation (MLE) approach is used to estimate the parameters of the PLHR distribution and the acceleration factor. A variance covariance matrix (VCM) is then obtained to construct the approximate confidence intervals (ACIs). In addition, studentized bootstrap confidence intervals (ST-B CIs) are also constructed and compared with ACIs in terms of their respective interval lengths (ILs). Moreover, a simulation study is conducted to demonstrate the performance of the estimation procedures and the methodology discussed in this paper. Finally, real failure data from the air conditioning systems of an airplane is used to illustrate further the performance of the suggested estimation technique.

New FMEA Risks Ranking Approach Utilizing Four Fuzzy Logic Systems

Process equipment and plant maintenance problems are complex in the oil refinery business, since effective maintenance needs to ensure the reliability and availability of the plant. Failure Mode and Effects Analysis (FMEA) is a risk assessment tool that aims to determine possible failure modes, and to reduce the ratio of unknown failure modes, by identifying business-critical systems and the risks of their failures. For the identified failure modes, FMEA determines risk mitigation action(s). The goal is to prevent failure and keep assets and plants running at peak performance by providing fully integrated operations, maintenance, turnarounds, modifications, and asset integrity solutions, during all phases of the asset life cycle. This research was based on FMEA use/application in refineries’ units, and proposes the new fuzzy FMEA risk quantification approach method: “four fuzzy logic system”. The model included a pre-assessment, by sets of fuzzy logic systems, that examined the input parameters that affected the variables of severity, occurrence, and detectability. The proposed model prioritized risks better and addressed the drawbacks of the conventional FMEA method.

Toward a Practical Zn Powder Anode: Ti3C2Tx MXene as a Lattice-Match Electrons/Ions Redistributor.

The renaissance of aqueous Zn ion batteries has drawn intense attention to Zn metal anode issues, including dendrites growth, dead Zn, low efficiency, and other parasitic reactions. However, against the widely used 2D Zn foil, in fact, the Zn powder anode is a more practical choice for Zn-based batteries in industrial applications, but the related solutions are rarely investigated. Herein, we focus on the Zn powder anode and disclose its unknown failure mechanism different from Zn foils. By utilization of 2D flexible conductive Ti3C2Tx MXene flakes with hexagonal close-packed lattice as electrons and ions redistributor, a stable and highly reversible Zn powder anode without dendrite growth and low polarization is constructed. Low lattice mismatch (∼10%) enables a coherent heterogeneous interface between the (0002) plane of deposited Zn and (0002) plane of the Ti3C2Tx MXene. Thus, the Zn2+ ions are induced to undergo rapid uniform nucleation and sustained reversible stripping/plating with low energy barriers via the internally bridged shuttle channels. Paired with cyano group iron hexacyanoferrate (FeHCF) cathode, the FeHCF//MXene@Zn full battery delivers superior cycle durability and rate capability, whose service life with a CE of near 100% touches 850% of bare Zn powder counterparts. The proposed Ti3C2Tx MXene redistributor strategy concerning high-speed electrons/ions channel, low-barrier heterogeneous interface, is expected to be widely applied to other alkali metal anodes.

Using Probabilistic Fault Tree Analysis and Monte Carlo Simulation to Examine the Likelihood of Risks Associated with Ballasted Railway Drainage Failure

Inadequate track drainage can lead to a variety of issues, including flooding, accelerated track degradation, and progressive or sudden failure of railway track, slope, or embankment. These can result in unplanned track maintenance, additional passenger travel costs, and damage to third party property. However, railway drainage asset management is challenging because it involves the consideration of large interconnected assets, limited maintenance budgets, and unknown failure probabilities. To address this issue, this paper introduces a risk-informed approach for railway drainage asset management that uses fault tree analysis to identify the factors that contribute to railway drainage flood risk and quantifies the likelihood of the occurrence of these factors using Monte Carlo simulation. This rational approach enables drainage asset managers to evaluate easily the factors that affect the likelihood of railway track drainage failure, thereby facilitating the prioritization of appropriate mitigation measures and in so doing improve the allocation of scarce maintenance resources. The analysis identified 46 basic and 49 intermediate contributing factors associated with drainage failure of ballasted railway track (undesired event). The usefulness of the approach is demonstrated for three sites on the UK railway network, namely, Ardsley Tunnel, Clay Cross Tunnel, and Draycott. The analysis shows that the Clay Cross Tunnel had the highest probability of drainage failure and should be prioritized for maintenance over the other two sites. The maintenance required should focus on blockages because of vegetation overgrowth or debris accumulation.