摘 要

船舶柴油机是船舶动力装置的关键设备之一，如果发生故障，会严重影响船舶的运行情况，可能造成各种设备的损坏，导致巨大的经济损失，甚至会危及船员的生命财产安全。而现实中整个柴油机故障原因中由燃油系统的故障导致的所占比例最大。传统排查方法由于没有智能化的设备，需要及时排除的故障难以立即处理。为保证船舶动力装置正常运行，进行快速诊断、明确原因、排除故障的意义重大。

本文在分析了国内外发展现状后，以BP神经网络学习为基础，以主机燃油系统为对象，开展了智能故障诊断研究。本文主要内容包括以下几个方面：

1. 回顾和分析了诊断技术的热点与现状，结合理论基础与燃油系统诊断的实际需求，确立了本文的主要研究方向与技术路线。
2. 介绍了神经网络技术作为当下热门技术之一，有着非线性、强连线、高适应的种种优点，可以用于训练复杂数据。本文研究了使用最广泛的BP神经网络，并成功用以数据训练。
3. 介绍了燃油系统运行时其高压油管压力波形会因故障而产生变化，可以相对直观的体现燃油系统的工作状态。通过这一特性，成功将BP神经网络技术运用到了燃油系统故障诊断上。在对系统进行了调整与改进后，取得了较好的诊断效果。

关键词：故障诊断，柴油机，燃油系统，压力波，BP神经网络

Abstract

With the development of modernization, the requirement of shipbuilding knowledge is higher and higher. As one of the main equipment of marine power station, marine diesel engine. If that fails, it will affect the operation of the ship. It may damage all types of equipment, endanger the life and assets of cruise individuals, often causing huge economic losses. In fact, the overall failure of diesel engine is caused by the failure of fuel system. There are no intelligent devices and troubleshooting methods in traditional methods, so it is difficult to deal with the shortcomings that need to be eliminated in time. In the actual management process of engine room, it is necessary to diagnose all kinds of faults quickly, find out the causes, eliminate the faults effectively and ensure the normal operation of offshore power station.

In this paper, after analyzing the development status at home and abroad, based on BP neural network learning, the main engine fuel system is taken as the object to carry out intelligent fault diagnosis research. This paper mainly includes the following aspects:

Firstly, the hot spot and current situation of diagnosis technology are reviewed and analyzed. The main research direction and technical route of this paper are established based on the theoretical basis and the actual demand of fuel system diagnosis.

Secondly, as one of the most popular technologies, neural network technology has many advantages such as non-linear, strong connection and high adaptability, which can be used to train complex data. In this paper, the most widely used BP neural network is studied and successfully used for data training.

Thirdly, it introduces that the pressure waveform of the high-pressure oil pipe will change due to the fault when the fuel system is running, which can reflect the working state of the fuel system relatively intuitively. Through this characteristic, BP neural network technology is successfully applied to the fault diagnosis of fuel system. After the adjustment and improvement of the system, a better diagnosis effect has been achieved.

Key words: fault diagnosis, diesel engine, fuel system, pressure wave, BP neural network

目 录

第1章 绪论 1

1.1研究背景 1

1.2.国内外发展现状 1

1.3.论文研究内容和技术路线 3

1.3.1论文主要研究目标 3

1.3.2主要工作内容与技术路线 3

第2章 BP神经网络原理 5

2.1.人工神经网络的基本理论 5

2.1.1人工神经网络的发展 5

2.1.2.人工神经网络的原理 6

2.1.3.人工神经网络的应用 7

2.2.BP神经网络 8

2.2.1.BP网络模型 8

2.2.2.BP网络算法与设计 9

第3章 基于神经网络的燃油系统故障诊断 11

3.1.柴油机燃油系统典型故障分析 11

3.1.1.柴油机燃油系统介绍 11

3.1.2.柴油机燃油系统压力波 12

3.1.3.燃油系统故障压力波特征值 14

3.2.基于BP神经网络的故障诊断 15

3.2.1.输入与输出设计 15

3.2.2.BP神经网络设计与优化 17

3.2.3.BP神经网络的仿真结果 19

第4章 总结与展望 21

中外文参考文献 22

附录 23

致谢 28

第1章 绪论

1.1研究背景

从国家近年推出的《“十三五”国家战略性新兴产业发展规划》、《智能船舶发展行动计划（2019-2021年）》等政策^[1]，可见船舶工业作为我国重要工业之一，创新劲头足，发展前景广。

柴油机是十分关键的一种动力设备，若发生故障，会影响船舶整体的运行，甚至危及船上工作人员人身安全。而现实中整个柴油机故障原因中由燃油系统的故障导致的所占比例最大。对于船舶柴油机的故障诊断，传统方法大多需要凭轮机员个人的经验来进行判断进而排除故障。但是由于没有智能化的仪器，现实中往往难以及时排除需要快速处理的故障。因此，在现代机舱管理中，如何智能化的判断故障在哪里发生，快速诊断故障的发生原因，并排除故障。是现代智能船舶的一个重要课题。

为契合国家智能化船舶战略，满足现代化船舶的智能化要求，达到船舶实际运行时要求的效率高成本低的目标，开展智能化柴油机故障诊断研究，对其运行情况进行智能化监测、发现征兆及时报险、故障发生以节省维修时间，降低维修人力成本，提高设备运行效率，具有重大意义。