In recent years, Convolutional Neural Network (CNN) has achieved a great success in computer vision. However, at present, for an image classification task, there is no CNN model that can perform 100% accurately due to insufficient or excessive feature learning. Once a CNN model deployed to perform tasks online, misclassified samples might lead the system with the CNN model deployed to enter an unsafe state such as collisions. To assess the performance of such online models, we, in this paper, propose Parallel Signal Routing Paths (PSRP) method to identify misclassified samples by extracting execution paths for each sample and comparing inherent feature differences in terms of CNN nodes between misclassified and well-classified samples, for the ultimate aim of addressing the challenge of test data not having ground-truth labels in online environment where the CNN models are deployed, and give availability results for applying PSRP on 3 public datasets and 3 typical CNN models.

微服务系统每轮迭代过程中都需要进行回归测试,大量重复测试会造成资源浪费,可通过减少测试用例集的规模来降低成本,以提高测试效率.现有测试用例集缩减技术主要依赖系统规约和架构描述作为输入,对于具有服务自治、调用关系不确定等特点的微服务系统实用性受限.并且,现有测试用例集缩减技术很少考虑使用场景,测试用例集很难反映用户关切.提出了 一种基于API网关层日志挖掘的测试用例集缩减技术,从API网关层日志中挖掘出能够反映服务使用场景的频繁调用路径,建立频繁路径与测试用例的关联关系,进而构建搜索图,并基于启发式搜索生成测试用例缩减集.描述了该技术的完整过程,并基于一个集成办公微服务系统进行了实验.实验结果表明:该技术能够缩减测试用例规模40％以上,且缺陷发现能力降幅不超过10％.

在飞机维修与保养过程中,航空维修公司已积累了大量经验性的维修日志数据.合理利用该类维修日志,结合机器学习方法,可以辅助维修人员做出正确的故障诊断决策.首先,针对维修日志的特殊性,提出一种迭代式的故障诊断基本过程;其次,在传统的文本特征提取技术的基础上,基于领域内信息,提出一种基于卷积神经网络(convolution neural network,简称CNN)的小样本文本特征提取方法,在样本量较少的情况下,利用预测目标将字向量作为输入,得到更为充分的文本特征;最后,使用随机森林(randomforest,简称RF)模型,结合其他故障特征判别飞机设备的故障原因.卷积神经网络以故障原因为目标,预先对故障现象中的字向量进行训练,从而得到更能反映该领域的文本特征.与其他文本特征提取方法相比,该类方法在小样本数据上得到了更好的效果.同时,将卷积神经网络与随机森林模型应用于飞机设备的故障原因判别,并与其他文本特征提取方式和机器学习预测模型进行对比,说明了该类文本特征提取方式和故障原因判别方法的合理性和必要性.

Effectively forecasting the failure data in the usage stage is essential to reasonably make relia-bility plans and carry out reliability maintaining activities.Beginning with the historical failure data of complex system,a long short-term memory(LSTM)based recurrent neural network for failure time series prediction is presented,in which the design of network structure, the procedures and algorithms of network training and forecasting are involved.Furthermore,a multilayer grid search algorithm is proposed to optimize the parame-ters of LSTM prediction model.The experimental results are compared with various typical time series predic-tion models,and validate that the proposed LSTM prediction model and the corresponding parameter optimiza-tion algorithm have strong adaptiveness and higher accuracy in failure time series prediction.

面向标准的目标符合性论证是适航认证领域的重要研究内容。论证过程中,在已有证据不足以表明目标符合性论证结果满足期望要求时,有必要进一步收集证据。然而,由于缺少有效的方法帮助划定证据收集范围和收集力度,导致证据收集的结果往往是高成本低效力的。因此,必须避免收集任何低效力且高收集成本的证据,以确保在目标符合性论证结果满足期望要求时证据收集总成本较低。针对定量评估的目标符合性论证方法,以证据收集成本最小化为规划原则,提出了一种成本优化的证据收集方案,以便于明确需收集的证据项及收集力度,进而规避掉不必要的时间、人力、物力的消耗。结合RTCA DO-178C适航认证标准,通过一个覆盖多分支情况的案例,评估了方案的有效性。

编译原理的新概念、新算法和新理论比较多,被计算机及软件专业学生视为最难以掌握的专业基础课之一.针对课程教学中难点比较集中的问题,为了让学生能平缓地接受该课程的基础理论和知识,提出两遍教学法的思路,阐述如何将编译原理的实践环节和教学环节有机结合,对教学和实践的各个环节进行合理安排,并说明已取得的教学效果.

Safety-critical software has been widely used in many fields.As the specific requirement of safety-critical software is preventing catastrophes,this kind of software must comply with its relevant safety standards.But now it does not have any effective ways to construct objectives conformity argument model for standards.By analyzing the features of objectives of software testing process in DO-178C,an objective conformity argument pattern description framework based on GSN is proposed,and these patterns are described through four fields: the problems that we need to solve,the specification for the solution,the approach to use them and the effect after using them.At the same time,some extensions for safety case patterns are proposed to describe the objectives conformity argument patterns.On this basis,three objectives conformity argument patterns based on software testing process in DO-178C are proposed,which are code-requirement conformity argument pattern,test coverage of requirements argument pattern and test coverage of structure argument pattern.At the same time,the instantiated method to build the objectives conformity argument structure for a specific program based on these patterns is proposed.People can construct objectives conformity argument structure for objectives of software testing process in DO-178C effectively through the proposed way.At last,one case study,which is an embedded real-time operating system,indicates that the objectives conformity argument patterns proposed here are useful and effective.

Traceability is the mechanism or the ability to relate artefacts and the attached factors.Safety-critical system development,besides the general system development,contains more independent safety analysis which generates and verifies system safety requirements.At present,there are few traceability researches oriented to safety analysis process,which are of extremely challenging.Safety related standards,such as ARP-4761,DO 178C,provide guidelines for conducting safety analysis.However,some information may be neglected since there are a lot of concepts and methods.Besides,software safety requirement analysis should include both system to software and software to system safety analysis.Establishing bi-directional traceability of safety related information oriented to software safety requirement analysis process helps to simplify the verification and impact analysis.In this paper,we established a traceability model oriented to software safety requirement analysis process.

Conventional test data generation techniques based on random algorithm, constraint-solving and white-box testing oriented generation method ignore the dependency between test data and database, which are not suitable for test data reuse and web application testing. This paper presents a test data generation and selection approach based on data constraints in business processes, which improves the reusability of data by injecting semi-formalized constraints into web business pro-cess with data pool storage technique. The case study shows a set of test data with high usability can be generated by this technique with low consumption.

Test case is used for a particular destination and prepared in a set of test inputs, conditions and expected results in order to test whether a feature is to meet a specific demand [1]. In industrial applications, the description of the test cases is usually in the form of documents, these documents are usually short of a standard format [2]. The other hand, for real time systems, the test case description is more complex. In our work, with the investigation research on the real-time system in industrial field, and the study of the related standard and literature, we extracted out the key elements of the real-time system, constructed the domain model for the important concepts, and then based on the Restricted test case modeling (RTCM) [3], we proposed a real-time test case description method, named as the Real-time Restricted test case modeling (RT-RTCM). RTCM is a test case modeling approach which contains a templates and a group of restriction rules. Based on this approach, we define three templates and proposed new restriction rules, which can support and standardize the description of the real-time test case, while through on the constraints of the natural language, it can also effectively reduce the ambiguity of the test case description. And based on the description method, we proposed an approach about the real-time test cases generation and choice, this method is a combination of the branch coverage criteria of the test flows, the boundary coverage criteria of the test data and the status coverage criteria of the system resource. And in the end of this paper, we modeled the single elevator system, to prove that RT-RTCM is easy to learn, use and understand, and can be an effective technique for real-time test cases modeling.

如何确保机载软件满足其实时性需求一直是一个引人关注的研究问题.根据工业界的报告结果,缺陷发现得越早,用于修复缺陷以提高机载软件不超时的可能性的代价就越小.对于运行在由ARINC653标准所描述的综合模块化航电系统(IMA)上的机载软件可采用以下方法:将机载软件的设计模型(UML模型)转化成仿真模型(Simulink模型),通过在Simulink平台上运行仿真模型来发现潜在的实时性问题.由于机载软件可能与IMA平台之间具有大量的交互(例如接口层和操作系统层)用来申请使用资源或者与其它的应用程序进行通讯,因此设计了一个仿真内核来仿真IMA平台的行为.最后,使用一个工业案例来论证上述方法的有效性.

针对软件工程课程实践性强、涉及知识面广和知识更新快等特点,阐述高级软件工程课程中,面向培养学生系统化建模能力和工程分析与设计能力的互动式教学模式,及其在课程项目实践中的应用,同时介绍互动式教学模式理念和相应设计以及教学辅助平台和工程实验环境,并通过3种评价方式展现该教学模式对提高学生工程能力的作用。

故障定位的目的是帮助程序员寻找引发失效的原因或故障位置,以加快调试过程.故障和失效间的关系往往非常复杂,难以直接描述故障到失效的转化.最新的研究多采用差异分析的方法,基于可疑模式,构建故障推理贝叶斯网络,其节点由可疑模式及组成可疑模式方法的调用者构成;定义了贝叶斯网络的构建算法、各个相关概率的定义及BBN中各个边的条件概率计算公式.提出基于该BBN的推理算法,推理得到包含故障的模块,并计算得到每个模块包含故障的概率.提出了评价方法,详细设计了参数调整与定位性能的关系实验和定位结果分析实验.实验数据表明,该故障定位方法取得了平均0.761的定准率和0.737的定全率,定位结果良好,具有较高的实用价值.

"编译技术"作为一门核心专业必修课,其实践环节对本科学生培养动手能力有重要作用。但在教学过程中,笔者发现对于这种软件类课程的实践环节而言,通常的题目设置和考核方案并不能真正确保学生独立完成、学有所获,在过程中加强监督和控制才是有效的办法。本文给出在教学实践中采用的质量管理体系,并通过两年来的应用情况分析了该体系的效果。

"编译技术"是一门公认的难教难学的课程,包含了看似晦涩的理论部分和覆盖面广的实践部分。如何根据本学校的培养定位组织教学内容,如何实施教学过程,是教师们普遍关注的主要问题。本文结合北京航空航天大学这类工科院校强调工程技术的特点,探讨了"编译技术"课程的组织方式,提出了以编译过程为主导带动课程知识点的课程安排模式,并针对不同类型高校给出了教学安排建议。

White-box testing is an effective means of ensuring high reliable software quality.Traditional white-box testing,having the higher costs and relative limitations,may not be practical in the large reliable software testing.Therefore,a gray-box testing method,which is an organic combination of black-box testing and white-box testing is needed in the course of testing.A gray-box testing model is defined,and more over,the model is validated by testing a large-scale object-oriented system.

软件产品开发规模的扩大和数量的增长迫切需要找到一种方法来增加软件测试的有效性.可测试性设计可以增强软件的可测试性,降低测试的强度.该文讨论了软件可测试性的特征和影响软件测试的因素,以及改进软件可测试性设计的几种方法.建议在软件开发的整个周期中融入软件的可测试性的设计.

1 引言软件测试过程中,测试用例的生成是软件测试的关键.目前,测试用例的生成主要靠手工完成,因此测试效率低下,软件成本居高不下.为此,迫切需要开发一些测试用例的自动生成工具,提高软件测试效率,降低软件成本,保证软件质量,提高软件测试的自动化程度.

首先分析了传统软件质量评价过程模型的不足,提出了可操作性更强的软件质量评价过程模型.在此基础上,明确提出了以软件度量为基础、软件质量预测模型为依托、基于用户评测历史信息库的预测模型调整技术为优化手段的完整软件质量评价体系,并以实验方法验证了上述体系的合理性与有效性.

度量技术是保证软件质量的重要手段,它定量地反映了软件某一方面的特征.随着面向对象技术的发展,传统的结构化度量不再适合面向对象的软件,因此出现了不少针对面向对象程序的度量,这些度量在一定程度上反映了面向对象软件的若干方面的特征.文章详细分析了面向对象软件的主要特征,在此基础上介绍了现在常见的几组面向对象度量和它们的应用价值,并且分析了它们的不足之处.最后,提出了几个构造面向对象度量应该遵循的准则.