Probability Of Failure Occurrence Research Articles

Intelligent optimal preventive replacement maintenance policy for non-repairable systems

This paper developed optimal maintenance policies for industrial machines to ensure uninterrupted production. Analytical and intelligent replacement models were constructed using real-time data from a commercial photocopier operator’s expenditure logbook between 2012 and 2023. A goodness-of-fit test affirmed a log-logistic failure distribution with shape and scale parameters of 1.7233 and 763.9220, respectively. Probability functions from this distribution established replacement models, determining optimal preventive maintenance conditions. Through feature engineering additional features crucial for operational and cost optimization were abstracted, with 4,881 unique data points simulated from boundary conditions defined by the original or raw data. Analytical results showed that the failure distribution-based replacement model outperformed the hazard function-based model in minimum replacement time and unit maintenance cost over time, while key parameters such as availability, reliability, and failure occurrence probability remained constant at 96 %, 94 %, and 0.07 %, respectively. Machine learning models (Extreme Gradient Boosting: XGBoost, Support Vector Machine: SVM, and Multilayer Perceptron: MLP) demonstrated very high accuracy in availability (XGBoost = 99.49 %, SVM = 99.18 %, RF = 100 %, MLP = 99.80 %), reliability (XGBoost = 99.49 %, SVM = 99.18 %, RF = 100 %, MLP = 99.80 %), and maintenance cost (XGBoost = 100 %, SVM = 100 %, RF = 100 %, MLP = 100 %), exceeding results of the analytical models.

Using rare earth element tracers to investigate ridge slope erosion process in contour ridge system under extreme rainfall

Extreme rainfall is becoming more frequent and intense, which no doubt increases the risks and uncertainty of water erosion in widely-used contour ridge systems. Ridge slope erosion can increase the probability of contour failure occurrence and then induce contour ridge losing anti-erosion properties. However, there is little understanding of ridge slope erosive process controlled by contour ridge system responding to extreme rainfall. In this study, five rare earth elements were applied in different ridge segments and soil layers to monitor the successive development of ridge slope erosion process under extreme rainfall in contour ridge system. Simulated rainfall experiments at high rainfall intensity (100 mm h−1) lasting for 60 min were conducted in runoff plots with ridge tillage practices characterized by microtopography indices and ridge geometry indices. Results showed that three sub-processes were observed for ridge slope erosion, including inter-rill erosion, headward erosion, and rill erosion. Headward erosion mainly resulted from the break-point at ridge surface and sidewall collapse at ridge toe, and the pouring of ponding water in furrow greatly promoted rill development. Erosion rate was the largest during headward erosion process, and rill erosion contributed the most to eroded source. The dominated erosion process was transport-limited during inter-rill erosion process, and detachment-limited during headward and rill erosion processes. Sediment transport and deposition processes as well as their transformation occurred simultaneously after the development of headward erosion. Eroded sediment was mainly from the low part of ridge, which was transported by raindrop-impacted sheet flow or overflow. These findings provide better responding of agricultural tillage to global climate change and references for the development of physically-based erosion models.

Using Machine Learning to Evaluate Coal Geochemical Data with Respect to Dynamic Failures.

Dynamic failure events have occurred in the underground coal mining industry since its inception. Recent NIOSH research has identified geochemical markers that correlate with in situ reportable dynamic event occurrence, although the causes behind this correlative relationship remain unclear. In this study, NIOSH researchers conducted machine learning analysis to examine whether a model could be constructed to assess the probability of dynamic failure occurrence based on geochemical and petrographic data. Linear regression, random forest, dimensionality reduction, and cluster analyses were applied to a catalog of dynamic failure and control data from the Pennsylvania Coal Sample Databank, cross-referenced with accident data from the Mine Safety and Health Administration (MSHA). Analyses determined that 7 of the 18 geochemical parameters that were examined had the biggest impact on model performance. Classifications based on logistic regression and random forest models attained precision values of 85.7% and 96.7%, respectively. Dimensionality reduction was used to explore patterns and groupings in the data and to search for relationships between compositional parameters. Cluster analyses were performed to determine if an algorithm could find clusters with given class memberships and to what extent misclassifications of dynamic failure status occurred. Cluster analysis using a hierarchal clustering algorithm after dimensionality reduction resulted in four clusters, with one relatively distinct dynamic failure cluster, and three clusters mostly consisting of control group members but with a small number of dynamic failure members.

Vulnerability Assessment of Equipment Excited by Transient Electromagnetic Disturbances Based on Distribution-Free QMU Method

Based on the framework of quantification of margins and uncertainties (QMU) method, a distribution-free QMU method is proposed for the vulnerability evaluation of electrical and electronic equipment excited by transient electromagnetic disturbances. The proposed method can provide a comprehensive evaluation of the equipment considering the stress and the strength with uncertainties. The stress is represented by the induced responses on the equipment conductors, the strength is quantified by the occurrence probability of failure obtained from the experiment. The quantile-based QMU metric can take account of the inherent probability characteristics of stress and strength, as well as the aleatory uncertainties caused by sampling variability. The quantiles and their tolerance bounds are estimated by a nonparametric estimation method based on order statistics, which is a data-driven method without being constrained by the type of probability distribution. A case study of an electronic equipment in the power plant is carried out to illustrate the evaluation procedures of the proposed method.

A New Hybrid Fermatean Fuzzy Set and Entropy Method for Risk Assessment

Risk evaluation is an important part of the product design and product manufacturing process; it entails the pursuit of the highest product quality and preventing failure under the constraints of limited resources. The failure mode and effects analysis approach is one of the most widely applied risk evaluation tools that uses the product of the three risk elements of product failure items, severity, occurrence probability, and detection probability, to calculate the risk priority number, the priority of failure risk. However, the typical failure mode and effects analysis method ignores the consideration of objective weights, which may lead to incorrect evaluation results. Moreover, the method of expressing information about product failure provided by experts also directly affects the results of risk assessment. To comprehensively assess the risk of product failure, in this study, the hybrid of the Fermatean fuzzy set and entropy method was used to prioritize product failure items risk. This study used a service failure mode and effects analysis numerical example of self-service electric vehicles to illustrate and test the correctness of the proposed new hybrid Fermatean fuzzy set and entropy method. The mathematical operation results were also compared with the listing of different calculation methods. The test results prove that the proposed new hybrid Fermatean fuzzy set and entropy method can fully consider the cognitive information provided by experts to provide more accurate risk ranking results of failure items.

Artificial intelligence for predictive maintenance

Maintenance constitutes an important share of modern industrial activities. Reliable operations rely on the adequate application on maintenance. However, in the present competitive environment, maintenance processes must be optimized so that they will be performed only when needed, otherwise resources will be needlessly wasted. This is in contrast to the conventional approach where maintenance is scheduled according to a time plan regardless of it is needed or not. This paper presents the application of artificial intelligence to create a model that can successfully predict the condition of a machine in terms of the probability of failure occurrence. This work uses a synthetic dataset that reflects a realistic scenario where sensors are connected to a machine to monitor its health condition and record failure incidents. The dataset consists of 10,000 records. Each record consists of five numerical measurements: air and process temperatures, machine rotational speed and torque and finally a measurement of machine wear. This in addition to the type of product the machine is producing makes six input variables. The output response was considered as the state of machine failure that was represented as true or false. An artificial neural network was trained and it successfully predicted the state of the machine. The capabilities and limitations of applying artificial intelligence were discussed. In addition, a brief overview of other predictive techniques was given.

Failure Mode and Effects Analysis (FMEA) at the preanalytical phase for POCT blood gas analysis: proposal for a shared proactive risk analysis model.

Proposal of a risk analysis model to diminish negative impact on patient care by preanalytical errors in blood gas analysis (BGA). Here we designed a Failure Mode and Effects Analysis (FMEA) risk assessment template for BGA, based on literature references and expertise of an international team of laboratory and clinical health care professionals. The FMEA identifies pre-analytical process steps, errors that may occur whilst performing BGA (potential failure mode), possible consequences (potential failure effect) and preventive/corrective actions (current controls). Probability of failure occurrence (OCC), severity of failure (SEV) and probability of failure detection (DET) are scored per potential failure mode. OCC and DET depend on test setting and patient population e.g., they differ in primary community health centres as compared to secondary community hospitals and third line university or specialized hospitals. OCC and DET also differ between stand-alone and networked instruments, manual and automated patient identification, and whether results are automatically transmitted to the patient's electronic health record. The risk priority number (RPN=SEV×OCC×DET) can be applied to determine the sequence in which risks are addressed. RPN can be recalculated after implementing changes to decrease OCC and/or increase DET. Key performance indicators are also proposed to evaluate changes. This FMEA model will help health care professionals manage and minimize the risk of preanalytical errors in BGA.

Task Failure Prediction in Cloud Data Centers Using Deep Learning

A large-scale cloud data center needs to provide high service reliability and availability with low failure occurrence probability. However, current large-scale cloud data centers still face high failure rates due to many reasons such as hardware and software failures, which often result in task and job failures. Such failures can severely reduce the reliability of cloud services and also occupy huge amount of resources to recover the service from failures. Therefore, it is important to predict task or job failures before occurrence with high accuracy to avoid unexpected wastage. Many machine learning and deep learning based methods have been proposed for the task or job failure prediction by analyzing past system message logs and identifying the relationship between the data and the failures. In order to further improve the failure prediction accuracy of the previous machine learning and deep learning based methods, in this article, we propose a failure prediction algorithm based on multi-layer Bidirectional Long Short Term Memory (Bi-LSTM) to identify task and job failures in the cloud. The goal of Bi-LSTM failure prediction algorithm is to predict whether the tasks and jobs are failed or completed. The trace-driven experiments show that our algorithm outperforms other state-of-art prediction methods with 93 percent accuracy and 87 percent for task failure and job failures respectively.

A two-stage stochastic programming model for selective maintenance optimization

The problem of selective maintenance exists in many multicomponent systems carrying out an alternating sequence of missions, with scheduled breaks where only a limited number of components can be maintained due to time limit. In this paper, a stochastic programming approach is proposed for determining an optimum maintenance plan to minimize maintenance costs and expected failure costs, while maximizing the probability of successful accomplishment of the next mission under uncertainties in future operating conditions. Traditionally, future operating conditions that affect failure time distribution when calculating reliability in selective maintenance models were assumed as deterministic. In this study, future operating conditions are assumed to be uncertain. The system is subject to several uncertain condition scenarios of exposure, conditional, usage, stress, etc. Each scenario is modeled with its associated occurrence probability. The presented model is a two-stage stochastic mixed-integer nonlinear programming model with fixed recourse, where the first stage is associated with maintenance decisions made before uncertainties are revealed, and the second stage is modeled as a recourse function which is related to the occurrence probability of system failure. A numerical example of a series-parallel system is used to demonstrate the effectiveness of the suggested model.

Performance and failure analysis of a retrofitted Cessna aircraft with a Fuel Cell Power System fuelled with liquid hydrogen

Proton-Exchange Membrane-Fuel Cells (PEM-FC) are regarded as one of the prime candidates to provide emissions-free electricity for propulsion systems of aircraft. Here, a turbocharged Fuel Cell Power System (FCPS) powered with liquid H2 (LH2) is designed and modelled to provide a primary power source in retrofitted Cessna 208 Caravan aircraft. The proposed FCPS comprises multiple PEM-FCs assembled in stacks, two single-stage turbochargers to mitigate the variation of the ambient pressure with altitude, two preheaters, two humidifiers, and two combustors. Interlinked component sub-models are constructed in MATLAB and referenced to commercially available equipment. The FCPS model is used to simulate steady-state responses in a proposed 1.5 h (∼350 km) mission flight, determining the overall efficiency of the FCPS at 43% and hydrogen consumption of ∼28 kg/h. The multi-stack FCPS is modelled applying parallel fluidic and electrical architectures, analysing two power-sharing methods: equally distributed and daisy-chaining. The designed LH2-FCPS is then proposed as a power system to a retrofitted Cessna 208 Caravan, and with this example analysed for the probability of failure occurrence. The results demonstrate that the proposed “dual redundant” FCPS can reach failure rates comparable to commercial jet engines with a rate below 1.6 failures per million hours.

Towards accurate failure prediction in manual assembly by establishing a System Dynamics model

The objective of this paper is the optimization of failure management in manual assembly. An advanced System Dynamics model is applied to indicate the interactions between production and quality processes. To quantify the model, the determination of the probability of failure occurrence is considered a necessary prerequisite. In manual assembly, the failure occurrence is directly related to workers' failure. Therefore, this paper analyses existing methods for calculating the human failure probability and evaluates them based on their suitability for the application in manual assembly. Subsequently, the best-evaluated method is further developed to fulfil industrial requirements. A mathematical model is developed according to the specific requirements of manual assembly, which enables the determination of the probability of failure occurrence.

Explainable Boosting Machines for Slope Failure Spatial Predictive Modeling

Machine learning (ML) methods, such as artificial neural networks (ANN), k-nearest neighbors (kNN), random forests (RF), support vector machines (SVM), and boosted decision trees (DTs), may offer stronger predictive performance than more traditional, parametric methods, such as linear regression, multiple linear regression, and logistic regression (LR), for specific mapping and modeling tasks. However, this increased performance is often accompanied by increased model complexity and decreased interpretability, resulting in critiques of their “black box” nature, which highlights the need for algorithms that can offer both strong predictive performance and interpretability. This is especially true when the global model and predictions for specific data points need to be explainable in order for the model to be of use. Explainable boosting machines (EBM), an augmentation and refinement of generalize additive models (GAMs), has been proposed as an empirical modeling method that offers both interpretable results and strong predictive performance. The trained model can be graphically summarized as a set of functions relating each predictor variable to the dependent variable along with heat maps representing interactions between selected pairs of predictor variables. In this study, we assess EBMs for predicting the likelihood or probability of slope failure occurrence based on digital terrain characteristics in four separate Major Land Resource Areas (MLRAs) in the state of West Virginia, USA and compare the results to those obtained with LR, kNN, RF, and SVM. EBM provided predictive accuracies comparable to RF and SVM and better than LR and kNN. The generated functions and visualizations for each predictor variable and included interactions between pairs of predictor variables, estimation of variable importance based on average mean absolute scores, and provided scores for each predictor variable for new predictions add interpretability, but additional work is needed to quantify how these outputs may be impacted by variable correlation, inclusion of interaction terms, and large feature spaces. Further exploration of EBM is merited for geohazard mapping and modeling in particular and spatial predictive mapping and modeling in general, especially when the value or use of the resulting predictions would be greatly enhanced by improved interpretability globally and availability of prediction explanations at each cell or aggregating unit within the mapped or modeled extent.

Cost Effective Hybrid Fault Tolerant Scheduling Model for Cloud Computing Environment

Cloud computing provides flexible and cost effective way for end users to access data from multiplatform environment. Despite of support by the features of cloud computing, there are also chances of resource failure. Hence there is a need of a fault tolerant mechanism to achieve undisrupted performance of cloud services. The task reallocation and duplication are the two commonly used fault tolerant mechanisms. But task replication method results in huge storage and computational overhead, when the number of tasks is increasing gradually. If the number of faults are high, it incurs more storage overhead and time complexity based on task criticality. In order to solve these issues, we propose to develop a Cost Effective Hybrid Fault Tolerant Scheduling (CEHFTS) Model for cloud computing. In this model, the Failure Occurrence Probability (FoP) of each VM is estimated by finding the previous failures and successful executions. Then an adaptive fault recovery timer is maintained during a fault, which is adjusted based on the type of faults. Experimental results have shown that CEHFTS model achieves 43% reduced storage cost and 13% reduced response delay for critical tasks, when compared to existing technique.

Do Fully Threaded Transiliac–Transsacral Screws Improve Mechanical Stability of Vertically Unstable Pelvic Fractures? A Cadaveric Biomechanical Analysis

To determine whether fully threaded transiliac-transsacral (TI-TS) fixation is biomechanically superior to partially threaded TI-TS fixation of vertically unstable transforaminal sacral fractures. Vertically unstable zone 2 sacral fractures were created in 20 human cadaveric pelves with a unilateral osteotomy and resection of 1 cm of bone through the foramen of the sacrum to represent comminution. Ten specimens received either 2 7.3-mm fully threaded or 2 7.3-mm partially threaded TI-TS screw fixation at the S1 and S2 body, and every specimen received standard 3.5-mm 8-hole parasymphyseal plating anteriorly. Each pelvis was loaded to 250 N at 3 Hz for 100,000 cycles and then loaded to failure. The primary outcome was fracture displacement at the S1 foramen, which was measured at 25,000, 50,000, 75,000, and 100,000 cycles. Secondary outcomes were simulated clinical failure of ≥1 cm displacement at the S1 foramen to determine occurrence probability of failure, and load at failure was defined as 2.5 cm of the linear loading system displacement. Specimens in the fully threaded and partially threaded cohorts were otherwise respectively comparable in regards to age, gender, and bone density. Five of the 10 TI-TS partially threaded specimens experienced simulated clinical failure with >1 cm displacement at the S1 foramen compared with 0 of the 10 TI-TS fully threaded cohort (50% vs. 0%, P = 0.03). The mean maximal displacement at the S1 foramen was greater in the partially threaded cohort (9.3 mm) compared with the fully threaded cohort (3.6 mm; P = 0.004). Fully threaded specimens also demonstrated greater mean force to failure than the partially threaded specimens (461 N vs. 288 N; P = 0.0001). Fully threaded TI-TS screw fixation seems to be mechanically superior to partially threaded fixation in a cadaveric vertically unstable transforaminal sacral fracture model with significantly less displacement of the posterior pelvic ring and greater load to failure.

Stochastic Risk Assessment with a Lagrangian Solution for the Optimal Cost Allocation in High-Speed Rail Networks

In large-scale high-speed rail networks (HSRNs), the occurrence of occasional malfunctions or accidents is unavoidable. The key issue considered in this study is the optimal allocation of the maintenance costs, based on the stochastic risk assessment for HSRNs. Inspired by the theoretical risk evaluation methods in the complex network, three major factors, including the local effects, global effects, and component self-effects are considered in the process of assessing the impact on the network components (nodes or lines). By introducing the component failure occurrence probability, which is considered to be an exponential function changing with the component maintenance costs, a feasible stochastic risk assessment model of the HSRNs together with the component impact assessment is proposed that can better unify the impact assessment of both the high-speed rail stations and railways. An optimal allocation algorithm based on a Lagrangian relaxation approach is designed. Correspondingly, the optimal cost allocation scheme can be determined using the algorithm to eliminate the various HSRN risks under the given costs. Furthermore, a real-world case study of the HSRNs in eastern China is illustrated. Compared with the genetic algorithm, the simulation shows that the approach can solve the optimal cost allocation problem to more effectively reduce the risks of large-scale HSRNs in practice.

Application of machine learning methods for predicting hazardous failures of railway track assets

The Aim of the paper is to reduce the number of hazardous events on railway tracks by developing a method of prediction of rare hazardous failures based on processing of large amounts of data on each kilometre of track obtained in real time from diagnostics systems. Hazardous failures are rare events; the set of variate values of the number of such events for an individual kilometre of track per year is: [0, 1]. However, for a railway network as a whole the yearly number of such events is in the dozens and efficient management requires the transition from the estimation of the probability of hazardous failure occurrence to the identification of the most probable location of failure. Methods . The problem of identification of rare, but hazardous possible events out of hundreds of thousands of records of non-critical railway track parameter divergences cannot be solved by conventional means of statistical processing. Hazardous events are predicted using the above statistics and artificial intelligence. Big Data and Data Science technology is used. Such technology includes methods of machine learning that enable item classification based on characteristics (features, predicates) and known cases of undesired event occurrence. The application of various algorithms of machine learning is demonstrated using the example of prediction of track superstructure failures using records collected between 2014 and 2019 on the Kuybyshevskaya Railway. Findings and conclusions . The result of facility ranking is the conclusion regarding the location of the most probable hazardous failure of railway track. That conclusion is based on the correspondence analysis between the actual characteristics of an item and conditions of its operation and the cases of adverse events and cases of their non-occurrence. The practical value of this paper consists in the fact that the proposed set of methods and means can be considered as an integral part of the track maintenance decision-making system. It can be easily adapted for online operation and integrated into the automated measurement system installed on a vehicle.

Electronic medical records-based comparison of glycemic control efficacy between sulfonylureas and dipeptidyl peptidase-4 inhibitors added on to metformin monotherapy in patients with type 2 diabetes.

Sulfonylurea (SU) and dipeptidyl peptidase-4 (DPP-4) inhibitors are most common secondary agents that are added to metformin monotherapy. Real-world studies have become increasingly important in providing evidence of treatment effectiveness in clinical practice and real-world data could help appropriate therapeutic information. Therefore, this study aims to compare the glycemic effectiveness of SU and DPP-4 inhibitors, which are added to metformin monotherapy in real clinical practice using electronic medical record (EMR) data. EMR data of type 2 diabetes patients treated at Seoul National University Hospital from December 2002 to December 2012 were retrieved and analyzed. The patients were divided into three groups: patients who maintained metformin monotherapy (M), and patients who added SU (MS) or DPP-4 inhibitors (MD) to metformin monotherapy. The mean change in HbA1c level, the proportion of patients achieving the HbA1c target < 7.0%, proportion of patients with treatment failure, and probability of treatment failure occurrence and changes in prescription were evaluated to compare glycemic control efficacy between SU and DPP-4 inhibitors. The MS showed significantly greater reduction in the Hb1Ac level than MD. The proportion of patients achieving HbA1c < 7.0% is higher in MD, whereas the proportion of patients with treatment failure was greater in MS. The probability of the treatment failure and probability of changes in the prescription were lower in MD than MS with hazard ratio of 0.499 and 0.579, respectively. In conclusion, this real-world study suggested that DPP-4 inhibitors are expected to show more durable glycemic control efficacy than SU in long-term use.

Optimal Allocation of Dynamic Var Sources for Reducing the Probability of Commutation Failure Occurrence in the Receiving-End Systems

Commutation failures (CFs) are usually triggered by the voltage dips caused by the faults in the receiving-end systems. To reduce the probability of CFs, counter measures to reduce the voltage dips should be implemented, among which adding dynamic var sources (DVSs) is very helpful. In this paper, an optimal allocation method for DVSs is proposed to maximize the benefit of DVSs to be installed. First, the probability of CFs is quantified based on the concept of area of vulnerability where the detailed ac network is considered. In order to find the most effective locations for installing DVSs, a 0-1 nonlinear programming model is formulated based on the quantitative index. Then, we introduce penalty functions to transform the binary programming to an unconstrained optimization problem, which can be solved by the gradient iteration method efficiently. Moreover, the sensitivity analysis in the calculation of gradient provides a better understanding of the relationships between the DVSs and CFs. In case studies, the 37-bus system and 1648-bus system are used to show the effectiveness of the proposed method.

Basic approaches to ensuring structural integrity of technical systems

Structural reliability of technical systems is to be insured in the presence of high level of uncertainty related to natural variability of loads acting on the system, scatter of mechanical properties of structural materials, inaccuracies of geometrical dimensions, imperfections of test equipment and design models. In the view of these uncertainties two basic approaches to securing structural integrity can be distinguished: deterministic (standard-based) approach founded on the application of the so called safety factors that are introduced into the design equation to compensate uncertainties; and probabilistic (reliability-based) approach that requires that the probability of structural failure occurrence not to exceed some allowable limiting value considered acceptable by society at the current stage of the technological development, available resources and the willingness of the society to pay for the implementation of protection measures. The paper presents a comparative assessment of deterministic and probabilistic approaches to securing structural integrity of technical systems subjected to various types of loading regimes. The condition of the equivalence of these two basic approaches to securing structural integrity is considered.

Predictive Analysis of Settlement Risk in Tunnel Construction: A Bow‐Tie‐Bayesian Network Approach

A hybrid method consisting of bow‐tie‐Bayesian network (BT‐BN) analysis and fuzzy theory is proposed in this research, in order to support predictive analysis of settlement risk during shield tunnel excavation. We verified the method by running a probabilistic safety assessment (PSA) for a tunnel section in the Wuhan metro system. Firstly, we defined the “normal excavation phase” based on the fuzzy statistical test theory. We eliminated the noise records in the tunnel construction log and extracted the occurrence probability of facility failures from the denoised database. We then obtained the occurrence probability of environmental failures, operational errors, and multiple failures via aggregation of weighted expert opinions. The expert opinions were collected in the form of fuzzy numbers, including triangular numbers and trapezoidal numbers. Afterwards, we performed the BT‐BN analysis. We mapped the bow‐tie analysis to the Bayesian network and built a causal network PSA model consisting of 16 nodes. Causes of the excessive surface settlement and the resulting surface collapse were determined by bow‐tie analysis. The key nodes of accidents were determined by introducing three key measures into the Bayesian inference. Finally, we described the safety measures for the key nodes based on the PSA results. These safety measures were capable of reducing the failure occurrence probability (in one year) of excessive surface settlement by 66%, thus lowering the accident probability caused by excessive surface settlement.