Fault Localization Techniques Research Articles

Multi‐objective optimization‐based and fault localization‐oriented test case generation for novice programs

SummaryOnline judgment (OJ) systems are capable of evaluating program results by automatically executing test cases, significantly improving the efficiency of traditional guidance approaches. Moreover, existing studies attempt to assist novices through automated fault localization techniques to provide feedback to novices, which can help them quickly find the location of faulty statements. Among them, spectrum‐based fault localization (SBFL) techniques have been widely used for their lightweight and efficiency, which only requires coverage information and test results of test cases to conduct fault localization. However, manually constructing high‐quality test cases for a large number of OJ questions is tough work to complete. To solve this problem, we propose the novice program‐oriented Multi‐Objective Optimization‐Based Fault Localization‐Oriented Test Case Generation (MFTCG) for automatically generating test inputs. Specifically, we use multi‐objective optimization algorithms to evolve the test case in terms of both fault localization and faulty code detection capability. We conduct experiments with 8911 programs from the well‐known public OJ platform AtCoder. The results show that our proposed approach MFTCG can achieve the best fault localization performance compared with existing automated test case generation approaches in most cases and can achieve the similar faulty code detection capability compared to manually designed test cases.

Boosting fault localization of statements by combining topic modeling and Ochiai

Context:Reducing the cost of maintenance tasks by fixing bugs automatically is the cornerstone of Automated Program Repair (APR). To do this, automated Fault Localization (FL) is essential. Two families of FL techniques are Spectrum-based Fault Localization (SBFL) and Information Retrieval Fault Localization (IRFL). In SBFL, the coverage information and execution results of test cases are utilized. Ochiai is one of the most effective and used SBFL strategies. In IRFL, the bug report information is utilized as well as the identifier names and comments in source code files. Latent Dirichlet Allocation (LDA) is a generative statistical model and one of the most popular topic modeling methods. However, LDA has been used at the method level of granularity as IRFL technique, whereas most existing APR tools are focused on the statement level. Objective:This paper presents our approach that combines topic modeling and Ochiai to boost FL at the statement level. Method:We evaluate our approach considering five different projects in Defects4J benchmark. We report the performance of our approach in terms of hit@k and MRR. To study the impact on the results, we compare our approach against five baselines: two SBFL approaches (Ochiai and Dstar), two IRFL approaches (LDA and Blues), and one hybrid approach (SBIR). In addition, we compare the number of bugs that are found by our approach with the baselines. Results:Our approach significantly outperforms the baselines in all metrics. Especially, when hit@1, hit@3 and hit@5 are compared. Also, our approach locates more bugs than Ochiai and Blues. Conclusion:The results of our approach indicate that the integration of topic modeling with Ochiai boosts FL. This uncovers the potential of topic modeling for FL at statement level, which is valuable for the APR community.

Application of a Centroid Frequency-Based Back Propagation Neural Network Fault Location Method for a Distribution Network Considering Renewable Energy Assessment

The distribution network is a crucial component of the power system as it directly connects to users and serves the purpose of distributing power and balancing the load. With the integration of new energy sources through distributed generation (DG), the distribution network has undergone a transformation from a single power radial network into a complex multi-source network. Consequently, traditional fault location methods have proven inadequate in this new network structure. Therefore, the focus of this paper is to investigate fault location techniques specifically tailored for DG integration into distribution networks. This paper analyzes how fault conditions impact the characteristics of single-phase grounding faults and extracts spectral feature quantities to describe differences in zero-sequence currents under various fault distances. This paper also proposes a fault location method based on centroid frequency and a BPNN (back propagation neural network). The method uses centroid frequency to describe the features of zero-sequence currents; to simulate the mapping relationship between fault conditions and spectral features, BPNN is employed. The mapping relationship differs for different lines and distribution networks. When a line faults, the spectral features are calculated, along with the transition resistance and fault closing angle. The corresponding mapping relationship is then called upon to complete distance measurements. This location method can be applied to various types of distribution lines and fault conditions with high accuracy. Even with insufficient training samples, sample expansion can ensure accuracy in fault distance measurement. We built a distribution network consisting of four feeders with different types and lengths of each line on Simulink and verified the effectiveness of the proposed method by setting different fault conditions. The results suggest that the method has a clear advantage over other frequency domain-based approaches, especially for hybrid lines and feeder lines with branches in distribution networks. Additionally, the method achieves a measurement accuracy within a range of 100 m.

A double-ended third harmonic method for vegetation high impedance fault location in overhead distribution systems

Vegetation-related high-impedance faults (VHIFs) on overhead distribution systems can cause damage, maintenance issues, and even wildfires, impacting both utilities and ecosystems. Existing fault location techniques struggle to accurately locate VHIFs. In this paper, we propose a novel double-ended terminal location method based on the third harmonic component of the fault system. The proposed technique determines the distance of the VHIF by approximating the third harmonic voltage at the fault local, utilizing measurements from the remote terminal located at the far-end of the feeder. This method efficiently locates VHIFs with just two measurement terminals, cutting the need and cost of extensive sensor deployment. Moreover, the proposed method overcomes the limitations of traditional fault location techniques by using a third harmonic model of the faulted system, eliminating the need for complex mathematical models to approximate the fault behavior and voltage at the fault local. Simulation tests were conducted to validate the proposed method, demonstrating its effectiveness in accurately locating VHIFs. The technique achieved error rates below 1.6 %, making it a promising solution for real-world applications in power distribution systems. The ability to precisely locate VHIFs meets a proactive protection philosophy and contributes to improved power system protection and ensures uninterrupted power supply.

A novel fault location method for the active distribution network based on dynamic quantum genetic algorithm

AbstractThe fault location of an active distribution network is a vital analysis to prevent major outages in the power system. Considering the influence of renewable distributed generations on fault characteristics, this paper proposes a novel location method based on a dynamic quantum genetic algorithm to solve for fault locations in active distribution networks. In the method, the fault current code is measured based on feeder terminal units. A universal switching function is presented to convert the feeder switch status into an uploaded fault current code. The fault location model is defined as an optimization problem that presents the evaluation objective function with an anti-false-positive factor. The dynamic quantum genetic algorithm is developed to locate the fault feeder according to the uploaded fault current code of the feeder terminal unit. The algorithm adopts dynamic rotating gate strategy and adaptive quantum crossover strategy to satisfy the requirements of quickness and accuracy for fault location. Moreover, the method avoids easily falling into a local optimum by integrating the discrete quantum mutation. The proposed fault location technique is tested and compared to other existing techniques on a 33-bus active distribution network. The simulation results show that the proposed fault location method can locate fault feeders accurately with fast computational times under conditions of single or multiple faults and with an information distortion of the feeder terminal unit.

A Fault Location Algorithm for Multi-Section Combined Transmission Lines Considering Unsynchronized Sampling

This paper introduces a fault location technique for multi-section combined transmission lines, utilizing unsynchronized fault record data. The approach determines the intersection points of the voltage magnitudes computed using the fault current and voltage records of each terminal and the constant line parameters of each section. Finally, the real fault location is identified based on the calculated inverse angle of equivalent impedance. To evaluate the performance of the proposed algorithm, a three-section combined transmission line is constructed using the PSCAD/EMTDC software version 5.0.0, and various fault scenarios are simulated and tested employing the proposed algorithm. The results indicate that the proposed algorithm is applicable to various types of faults in a multi-section combined transmission line; the estimations are accurate.

A deep semantics-aware data augmentation method for fault localization

Context:Fault localization (FL) techniques are employed to identify the relationship between program statements and failures by analyzing runtime information. They rely on the statistics of input data to explore the underlying correlation rooted in it. Consequently, the quality of input data is of utmost importance for FL. However, in practice, passing tests significantly outnumber failing tests regarding a fault. This leads to a class imbalance challenge that can adversely affect the effectiveness of FL. Objective:To tackle the issue of imbalanced data in fault localization, we propose PRAM: a deeP semantic-awaRe dAta augmentation Method to improve the effectiveness of FL methods. Method:PRAM utilizes program dependencies to enhance the semantic context, thus showing how a failure is caused. Then, PRAM employs mixup method to synthesize new failing test samples by merging two real failing test cases with a random ratio to balance the input data. Finally, PRAM feeds the balanced data consisting of synthesized failing test cases and original test cases to FL techniques. To evaluate the effectiveness of PRAM, we conducted large-scale experiments on 330 versions of nine large-sized real programs for six state-of-the-art FL methods, two data optimization methods and two data augmentation methods. Results:Our experimental results show that PRAM outperforms in most cases for Top-K metrics and reduces the number of checked statements from 40.38% to 80.04% compared with the original FL methods. Furthermore, PRAM reduces the checked statements from 16.92% to 56.98% for data optimization methods and from 12.48% to 26.82% for data augmentation methods. Conclusion:The experimental results show that PRAM is not only more effective than the original FL methods but also more effective than two representative data optimization methods and two data augmentation methods, which indicates that PRAM is a universal effective data augmentation method for various FL methods.

A PLC communication characteristics‐based fault location method in medium voltage meshed distribution networks

AbstractA power line carrier (PLC) communication characteristics‐based method is proposed for single‐phase‐to‐ground fault location in neutral isolated medium voltage meshed distribution networks in this paper. The carrier signals with a time‐varying frequency and constant amplitude are processed by a set of PLC transmitters and receivers, whose placement is optimized by regarding the power network as an undirected graph. Two signal encoding and decoding algorithms for the PLC terminals are proposed to avoid using expensive timing systems between the terminals. The fault location technique is implemented by comparing the cosine similarity of amplitude attenuation and phase offset between the fault and a feature library. The node corresponding to the maximum cosine similarity of the characteristics between the present fault and the library is selected as the location of the current fault. Only one set of low‐cost PLC communication terminals and the widely available power lines are needed in the fault location system, making this approach highly practical. Numerical simulations using MATLAB/Simulink have been performed to verify the technique's feasibility. The results show that the method can accurately locate faults in neutral isolated medium voltage meshed distribution networks. Besides, the presented approach achieves a high level of accuracy in estimating transition resistance values.

Variable-based Fault Localization via Enhanced Decision Tree

Fault localization, aiming at localizing the root cause of the bug under repair, has been a longstanding research topic. Although many approaches have been proposed in past decades, most of the existing studies work at coarse-grained statement or method levels with very limited insights about how to repair the bug ( granularity problem ), but few studies target the finer-grained fault localization. In this article, we target the granularity problem and propose a novel finer-grained variable-level fault localization technique. Specifically, the basic idea of our approach is that fault-relevant variables may exhibit different values in failed and passed test runs, and variables that have higher discrimination ability have a larger possibility to be the root causes of the failure. Based on this, we propose a program-dependency-enhanced decision tree model to boost the identification of fault-relevant variables via discriminating failed and passed test cases based on the variable values. To evaluate the effectiveness of our approach, we have implemented it in a tool called VarDT and conducted an extensive study over the Defects4J benchmark. The results show that VarDT outperforms the state-of-the-art fault localization approaches with at least 268.4% improvement in terms of bugs located at Top-1, and the average improvement is 351.3%. Besides, to investigate whether our finer-grained fault localization result can further improve the effectiveness of downstream APR techniques, we have adapted VarDT to the application of patch filtering, where we use the variables located by VarDT to filter incorrect patches. The results denote that VarDT outperforms the state-of-the-art PATCH-SIM and BATS by filtering 14.8% and 181.8% more incorrect patches, respectively, demonstrating the effectiveness of our approach. It also provides a new way of thinking for improving automatic program repair techniques.

Data-driven fault location approach in AC/DC microgrids based on fault voltage and current differences

In recent years, various fault location techniques for microgrids have been presented. Some of these methods estimate the distance or section of the fault by assuming the availability of data from all measuring nodes, which is not economically viable. Moreover, the accuracy of these techniques is sensitive to the type of sources, fault location, resistance, inception angle, and they require accurate line models and enormous data banks. The prominent disadvantage of these methods is their performance for DC or AC networks separately. This article suggests a method to estimate both the distance of the fault and faulty section in a hybrid microgrid (DC/AC) using the π line model. This method begins by analyzing the voltages and currents of all buses based on the data collected at the points of common coupling and DG locations. Then, by analyzing the voltage and current difference at the buses of each branch, the fault location is determined. The proposed approach does not require DG models and needs a quarter of the data cycle. A 15-node hybrid microgrid is implemented in MATLAB, and a 7-node AC microgrid is implemented in the power systems simulator to evaluate the suggested technique. The results confirm the insensitivity of this approach to the different fault types, locations, inception angles, inductances, high impedance fault (500 Ω), DG operating conditions, and line parameters. In addition, the proposed algorithm is investigated by considering the measurement error. The results show the suitable accuracy of this scheme compared to other similar methods.

A Hybrid Technique for Fault Classification and Location in a Jointed Overhead-Underground Distribution Line

The electrical distribution system is crucial for the utility grid to transmit power from generators to consumers. Considering the intricate structure of distribution systems and their significant role in power networks, establishing a robust fault classification and location scheme is vital. Due to ageing, distribution systems are often prone to faults from factors like poor operational conditions and wear and tear. The line faulting rate and the pertinent restoration epochs influence the frequency and duration of power disruptions. Thus, precisely locating the fault section is essential to minimize power restoration timeframes. This paper presents a hybrid fault classification and location technique in a combined continuous overhead and underground distribution line. A simulation of the hybrid model was designed in Simulink for an 11 kV combined continuous overhead and underground electrical distribution line, considering short circuit faults as they are the most predominant and cause massive damage in distributed systems. The proposed technique first classifies the fault using Discrete Wavelet Transforms (DWT) and Multi-layer Perceptron-Artificial Neural Networks (MLP-ANN). Next, the impedance and Adaptive Neuro-Fuzzy System (ANFIS) based technique is employed for fault location. At a sample rate of 50 kHz, the DWT was applied to current signals and the coefficients used for ANN training, while phase impedance values were used as input to the ANFIS for training. The simulation results showed accuracy of 96.6% for fault classification and 99.17% for fault location. The developed models can significantly enhance fault location for speedier outage resolution by promptly repairing the affected distribution lines.

Trace matrix optimization for fault localization

Fault localization (FL) techniques gather trace information as input data and analyze it to identify the relationship between program statements and failures. Therefore, the input trace matrix is essential for fault localization. However, the current trace matrix faces two main challenges. Firstly, the occurrences of coincidental correctness (CC), which refer to the execution of faulty statements that lead to correct program output, adversely impact the effectiveness of FL. Secondly, the significant disparity in the number of failing and passing test cases poses a data imbalance problem for fault localization. To overcome these issues, we propose TRAIN: a Two-stage tRace mAtrix optImizatioN method for fault localization. In the first stage of optimization, TRAIN leverages an improved cluster analysis to identify and exclude the CC tests to optimize the trace matrix. Subsequently, in the second stage, TRAIN utilizes data augmentation to enhance the failing test cases to further balance the trace matrix. The optimized trace matrix is then used as input data in the FL pipeline to locate the faulty statements. Through extensive experiments conducted on 330 faulty versions of nine large-sized programs (obtained from Defects4J, ManyBugs, and SIR) using six state-of-the-art FL methods, TRAIN demonstrates remarkable improvements in FL effectiveness.

High pass filter based traveling wave method for fault location in VSC-Interfaced HVDC system

Due to the difficult topography and bad weather conditions across large distances, dc line breakdown is a serious danger to the safety of high-voltage direct-current systems (HVDC). Thus, there is a need of an accurate and reliable fault location method to safely isolate the dc grid in the occurrence of a dc fault. Traveling wave (TW) fault location techniques have gained in popularity over the decades, with the expectation that power networks will become considerably more credible and accurate throughout. However, in the existing traveling wave based location, the signal processing methodology for extracting the wave front consumes too many computer resources, necessitating the development of a new location method with a fast response and high accuracy. In this paper, a new algorithm of TW based on high pass filter is proposed, it can differentiate the TW peak more easily by removing the oscillation in a signal. To ensure more accuracy double ended method (Type D) is used to obtain the location for the fault. The theory behind this new algorithm is presented in detail. High pass filter is designed for this purpose and its validity tested on point to point (P2P) dc simulation model which is created in PSCAD software to gather the voltage data from both terminals during transient. Following that the proposed fault location algorithm is executed in the MATLAB environment to get the fault location results. A comparison of the proposed algorithm with two popular methods namely the wavelet transforms (WT) and the single ended method highlights the superiority.

Visually impaired recognition of a fully loaded high and medium voltages electrical network for layout formation and fault monitoring

Faults in power systems are inevitable. These faults can cause undesirable outages that interrupt people's life, block factories’ production chain, and may even in worse case scenarios contribute to killing hospitalized patients. To minimize such catastrophes, many techniques were developed for the purpose of instantaneous monitoring and repairing of power systems. However, due to the increasing complexity of transmission line networks, enhancements to the fault location techniques became a necessity to align with the new challenges. This paper presents a technique designed to diagnose faults in a power network under test; it starts by a comparison between two commonly used Fault Location Algorithms: impedance and travelling wave-based methods. The latter have shown numerous advantages when compared to impedance-based methods that led us to the adoption of a travelling-wave variant Time Domain Reflectometry (TDR). TDR was incorporated with the graph theory and an optimization algorithm (Particle swarm optimization (PSO) in order to acquire the fault data rapidly, accurately, and with the least possible complexity.

Enhanced artificial intelligence technique for soft fault localization and identification in complex aircraft microgrids

- In recent years, the aviation industry has witnessed a substantial integration of power electronics technology within Aircraft Microgrids (AMs). Consequently, the extension of electrical wiring networks has expanded, resulting in heightened intricacies within these systems. Therefore, the identification and location of faults in wiring networks have become an important topic in AMs to guarantee the safety of the electrical power systems. Time Domain Reflectometry (TDR) is widely used to locate and recognize electric wire faults. However, soft fault location on complex electrical networks using TDR is perplexing due to its weak effect on the reflected signal. Moreover, the existence of noise in the environment can worsen the TDR's performance. In this paper, a new approach based on TDR, along with the Subtractive Correlation Method (SCM) and Neural Network (NN), is proposed. The TDR response of the complex wiring network is determined using a newly proposed model based on the Finite Difference Time Domain (FDTD) method. The validity of the proposed model is established through experimentation including two distinct cable types also the introduced model notably enhances computational efficiency, a fact substantiated by our experimental findings and an extensive benchmarking against recent publications. These evaluations collectively underscore a significant reduction in computational time. Then, the reflected signal undergoes processing through SCM, a technique employed to amplify the subtle influence of the soft fault in two scenarios: one accompanied by noise and the other noise-free. Furthermore, NN is used to handle the inverse problem of localizing and characterizing the soft faults by their exact resistance values. Even within noisy environments, the proposed methodology excels in accurately locating and characterizing soft faults with a high degree of precision, all in real-time diagnostic scenarios.

Understanding the use of spectrum‐based fault localization

SummaryDevelopers spend significant time locating and fixing bugs, which is often performed manually. Although spectrum‐based fault localization (SFL) techniques aim at helping developers to locate faults, they are not yet used in practice. Recent studies have investigated how developers use SFL, presenting different conclusions about their effectiveness and usefulness. We carried out a user study to further enhance the understanding of SFL. We assessed whether SFL improves the developers' performance and to what extent SFL leads developers to inspect faulty code excerpts. We also investigated the intention of the developers to use SFL and how they interact with SFL. Twenty‐six participants performed debugging tasks using real programs, with and without using the Jaguar SFL tool. Using SFL, more developers located and fixed the bugs. SFL also led more developers to inspect the faulty code and locate the faulty method. However, they did not spend less time locating the faults. SFL was well‐accepted by the participants, who showed intention to use it in their daily activities. Our results indicate that SFL is useful even when the fault is not ranked among the first positions, leading developers to reach faulty code regions and find the bugs.

Model-domain failing test augmentation with Generative Adversarial Networks

It is undoubtedly that test suites are crucial for fault localization techniques. Through the running of a single test case in the test suite, a fault localization technology is able to obtain the dynamic coverage information of the statements, which is a vector and could be analyzed to identify the corresponding coefficients between statements and program failures. This vector is commonly referred to as a model domain test sample. There are two types of model domain test samples: passing ones and failing ones. However, the number of passing test samples often far exceeds the number of failing ones, which undoubtedly causes the asymmetry phenomenon. Previous studies have proved that the imbalance of model domain test samples may hamper the effectiveness of fault localization techniques. In practice, it is difficult to create failing test cases in input domain. With the rapid development of deep learning techniques, Generative Adversarial Networks (GAN) have achieved promising results in many fields, which may bring a new perspective for augmenting model domain failing test samples. Thus, we try to provide a new method named MAG: Model-domain failing test Augmentation with Generative Adversarial Networks. MAG first constructs a Generative Adversarial Network, and then generates vectors with the abstract representation of execution information as input. At last, MAG utilizes influential global and local contexts to enhance the generated vectors, and to formulate model domain test samples with failing labels. MAG augments model domain failing test samples. Different from traditional methods of generating test cases directly from input domain, MAG seeks to augment failing test samples of model domain, which could improve the effectiveness of existing fault localization techniques. Experimental results illustrate that MAG significantly improves the effectiveness of 13 typical fault localization approaches and is better than two representative data optimization techniques.

Improving fault localization via weighted execution graph and graph attention network

AbstractSoftware fault localization is commonly recognized as arduous and time consuming. Spectrum‐based fault localization (SBFL) has been widely used due to its lightness. However, the effectiveness of SBFL is limited since it only considers simple statistics on the coverage information, ignoring the tie problem that the spectrum matrixes of some statements are the same. Most existing deep learning‐based fault localization (DLFL) techniques convert the coverage information into a vector, which utilizes the spectrum in a simplified manner and still has limitations in practice. To solve the above problem, we propose an approach via the weighted execution graph and graph attention network (WEGAT). We use a graph structure to represent the coverage information between test cases and program elements. Then, we generate a weighted execution graph by applying the predicate execution sequence. Furthermore, we combine the weighted execution graph with the AST as an integrated graph, which is the input of the GAT for fault localization. We evaluate WEGAT in within‐project and cross‐project prediction scenarios on the Defects4J benchmark. Experimental results show that our approach outperforms traditional SBFL (Ochiai, DStar and Tarantula) and DLFL (TraPT, CNN‐FL, Grace, and AGFL) methods, effectively improving the accuracy of fault localization.

Fault localization in DSLTrans model transformations by combining symbolic execution and spectrum-based analysis

AbstractThe verification of model transformations is important for realizing robust model-driven engineering technologies and quality-assured automation. Many approaches for checking properties of model transformations have been proposed. Most of them have focused on the effective and efficient detection of property violations by contract checking. However, there remains the fault localization step between identifying a failing contract for a transformation based on verification feedback and precisely identifying the faulty rules. While there exist fault localization approaches in the model transformation verification literature, these require the creation and maintenance of test cases, which imposes an additional burden on the developer. In this paper, we combine transformation verification based on symbolic execution with spectrum-based fault localization techniques for identifying the faulty rules in DSLTrans model transformations. This fault localization approach operates on the path condition output of symbolic transformation checkers instead of requiring a set of test input models. In particular, we introduce a workflow for running the symbolic execution of a model transformation, evaluating the defined contracts for satisfaction, and computing different measures for tracking the faulty rules. We evaluate the effectiveness of spectrum-based analysis techniques for tracking faulty rules and compare our approach to previous works. We evaluate our technique by introducing known mutations into five model transformations. Our results show that the best spectrum-based analysis techniques allow for effective fault localization, showing an average EXAM score below 0.30 (less than 30% of the transformation needs to be inspected). These techniques are also able to locate the faulty rule in the top-three ranked rules in 70% of all cases. The impact of the model transformation, the type of mutation and the type of contract on the results is discussed. Finally, we also investigate the cases where the technique does not work properly, including discussion of a potential pre-check to estimate the prospects of the technique for a certain transformation.

Code-aware fault localization with pre-training and interpretable machine learning

Following the rapid development of deep learning, many studies in the field of fault localization (FL) have utilized deep learning to analyze statements’ coverage information (i.e., executed or not executed) and test cases’ results (i.e., failing or passing), which have shown dramatic ability in identifying suspicious statements potentially responsible for failures. However, they mainly pay attention to the binary information of executing test cases but ignore incorporating code snippets and their inner relationships into the learning process. Furthermore, how a complex deep learning model for FL achieves a particular decision is not transparent. These drawbacks may limit the effectiveness of FL. Recently, graph-based pre-training techniques have dramatically improved the state-of-the-art in a variety of code-related tasks such as natural language code search, clone detection, code translation, code refinement, etc. And interpretable machine learning tackles the problem of non-transparency and enables learning models to explain or present their behaviors to humans in an understandable way.In this paper, our insight is to select a candidate solution that leverages the promising learning ability of graph-based pre-training techniques to learn a feasible model for incorporating code snippets as well as their inner relationships into fault localization, and then uses interpretable machine learning to localize faulty statements. Thus, we propose CodeAwareFL, a code-aware fault localization technique with pre-training and interpretable machine learning. Concretely, CodeAwareFL constructs a variety of code snippets through executing test cases. Next, CodeAwareFL utilizes the code snippets to extract propagation chains which could show a set of variables interact with each other to cause a failure. After that, a graph-based pre-trained model is customized for fault localization. CodeAwareFL takes the code snippets and their corresponding propagation chains as inputs with test results as labels to conduct the training process. Finally, CodeAwareFL evaluates the suspiciousness of statements with interpretable machine learning techniques. In the experimental study, we choose 12 large-sized programs to conduct the comparison. The results show that CodeAwareFL achieves promising results (e.g., 32.43% faults are ranked within top 5), and is significantly better than 12 state-of-the-art baselines.