Fault detection and diagnosis (FDD) modules in a modern control system are effective in detecting and identifying abnormal process behaviours in a timely manner, ensuring the high-performance of large-scale engineering systems. The detection and isolation of faults is essentially built on the characterisation of the observed behaviour of a system. However, due to the large number of technical indicators available for measurement, as well as the various constraints of sensor installation, monitoring all the operating parameters of a large-scale engineering system is not feasible. Therefore, locating sensors optimally in a large-scale system, to achieve a comprehensive description of an abnormality, becomes a key issue to successfully apply diagnostic technologies to real world situations. In this paper, a fault feature reduction (FFR) based sensor location approach is proposed for optimal sensor placement so as to achieve the desired performance of fault detection and isolation. The behaviour of faults is firstly analysed using a fault tree to obtain a comprehensive understanding of the multi-dimensional relationships between faults and symptoms. A Boolean matrix is then constructed to represent the corresponding relations around faults and potential sensors. All the alternative configurations of sensors, for a desired diagnosis of a system, are obtained by eliminating the redundant fault features. The trade-off without a certain sensor is also attained using the following proposed approach. Three large-scale systems, including, a diesel engine system and two chemical systems, are used to illustrate the proposed approach. Comparisons to existing competitive techniques indicate the enhanced abilities of the proposed approach to meet the varying requirements of a real-world monitoring network. The analysis of sensor placement can be performed at the design phase of a large-scale engineering system, to locate the preset measured hole, or, during the life-cycle, to perfect an incomplete or redundant monitoring system.

现代控制系统中的故障检测与诊断（FDD）模块可有效地及时检测和识别异常过程行为，从而确保大型工程系统的高性能。故障的检测和隔离基本上建立在对观察到的系统行为的表征上。但是，由于有大量可用于测量的技术指标以及传感器安装的各种限制，因此无法监视大型工程系统的所有运行参数。因此，在大型系统中最佳地定位传感器，以获得对异常的全面描述，成为成功地将诊断技术应用于现实世界情况的关键问题。在本文中，提出了一种基于故障特征减少（FFR）的传感器定位方法，以实现最佳的传感器位置，从而实现所需的故障检测和隔离性能。首先使用故障树分析故障的行为，以全面了解故障和症状之间的多维关系。然后构造一个布尔矩阵来表示故障和电位传感器周围的对应关系。通过消除冗余故障特征，可获得传感器的所有替代配置，以进行系统所需的诊断。使用以下建议的方法也可以在没有特定传感器的情况下进行权衡。使用三个大型系统，包括柴油发动机系统和两个化学系统，来说明所提出的方法。与现有竞争技术的比较表明，所提出方法的功能得到增强，可以满足现实世界监控网络的各种要求。传感器放置的分析可以在大型工程系统的设计阶段进行，以定位预设的测量孔，或者在生命周期内，完善不完整或冗余的监控系统。