煤矿井下环境复杂恶劣，存在各种有毒、易燃、易爆等有害气体，威胁井下工作人员生命安全。矿井主通风机作为煤矿通风系统的核心设备，能够有效排出有害气体、供应新鲜空气，保证井下人员作业环境，是煤矿安全生产的重要保障。矿井主通风机的健康状态关乎井下工作环境状况，一旦主通风机运行过程中出现故障，将严重影响煤矿的安全生产甚至引起重大事故。因此，本文对矿井主通风机的运行状态监测、诊断与预警进行研究。通过分析矿井主通风机常见故障机理及特征，制定矿井主通风机设备运行状态监测方案，提出基于振动信号的故障特征提取、智能故障诊断与趋势预警方法，构建矿井主通风机状态监测、诊断与预警的一体化系统，保障矿井主通风机的安全运行。

针对矿井主通风机振动信号的非平稳性导致故障特征准确提取困难的问题，提出一种依据转速变化幅度分工况进行特征提取的思路，将非平稳工况分为转速波动和变转速两种。针对转速波动工况，采用小波包分解提取不同故障下振动信号频带能量作为故障特征；针对变转速工况，采用阶次分析将时域非平稳信号转为角域伪平稳信号，对比不同故障信号角域特征得到故障特征参量。通过两种方法的结合，实现非平稳工况下矿井主通风机的振动故障特征精确提取。

针对基于数据的矿井主通风机故障诊断非线性问题，引入BPNN方法，将提取得到的特征值输入BPNN进行分类训练与识别，实验结果表明BPNN存在收敛速度慢、易陷入局部极小值的问题。本文通过引入粒子群算法优化BPNN，研究结果表明本文所提方法在故障识别准确率、算法收敛速度方面均有明显提高。

针对矿井主通风机故障预警存在的时间序列非线性拟合求解问题，在研究BPNN的基础上改进输出层激活函数增强网络动态跟踪性能。提出结合动态时间序列的PSO-BPNN故障趋势预警方法，通过实验数据及矿井主通风机现场数据验证了该方法的有效性。

        最后，开发了矿井主通风机状态监测与故障预警系统，包括硬件平台和软件两部分。硬件平台以普通PC为上位机，由传感器、采集卡、调理电路等模块组成完整信号传输路径；软件部分采用LabVIEW和MATLAB联合编程，实现矿井主通风机的状态监测、振动特征提取、故障诊断与预警、数据管理等功能。系统已在煤矿生产现场完成了各模块功能的验证，结果表明本文提出的特征提取、故障诊断与预警方法均实现预期功能，满足现场应用需求。

The underground environment of the coal mine is complex and harsh, and there are various toxic, flammable and explosive harmful gases, which threatens the safety of the underground workers. As the core equipment of the coal mine ventilation system, the mine's mine ventilator can effectively discharge harmful gases and supply fresh air to ensure the working environment of underground personnel, which is an important guarantee for coal mine safety production. The health status of the mine's main ventilator is related to the working environment of the mine. Once a fault occurs during the operation of the mine's main ventilator, it will seriously affect the safe production of the coal mine and even cause major accidents. Therefore, this paper studies the operation status monitoring, diagnosis and pre-warning of the mine's main ventilator. By analyzing the common fault mechanism and characteristics of the mine's main ventilator, the operation status monitoring program is formulated, and the vibration signal-based fault feature extraction, intelligent fault diagnosis and trend pre-warning method are proposed, and an integrated system of the mine's main ventilator condition monitoring, diagnosis and pre-warning to ensure its safe operation.

Aiming at the problem that the non-stationarity vibration signal of the mine's main ventilator makes it difficult to accurately extract the fault features, an idea of feature extraction based on the working conditions of the speed variation is proposed. Non-stationary operating conditions are divided into two types: speed fluctuation and variable speed. For the speed fluctuation conditions, the wavelet packet decomposition is used to extract the energy of the vibration signal frequency band under different faults as the fault characteristics and for the variable speed conditions, the order analysis is used to convert the non-stationary signal in the time domain to the pseudo-stationary signal in the angular domain, and then compares the angular domain characteristics of different fault signals to obtain the fault characteristic parameters. Through the combination of the two methods, the vibration failure characteristics of the mine's main ventilator can be accurately extracted under non-stationary conditions.

In order to solve the non-linear problem of the mine's main ventilator fault diagnosis based on data, the BPNN method is introduced, and the extracted feature are input into BPNN for classification training and recognition. The experimental results show that BPNN has the disadvantages of slow convergence speed and easy to fall into a local minimum. In this paper, the particle swarm optimization (PSO) algorithm is introduced to optimize BPNN. The research results show that the proposed method has significantly improved the accuracy of fault recognition and the speed of algorithm convergence.

In view of the time series non-linear fitting and solving problems of the mine's main ventilator failure pre-warning, based on the study of BPNN, the output layer activation function is improved to enhance the network dynamic tracking performance. A method for pre-warning combined with dynamic time series and PSO-BPNN is proposed, and the correctness of this method is verified by the data of laboratory and ventilator field operation.

        Finally, the condition monitoring and fault pre-warning system of the mine's main ventilator is developed, including the hardware platform and software. The hardware platform uses an ordinary PC as the host computer, and consists of sensors, acquisition cards, and conditioning circuits to form a complete signal transmission path. The software uses LabVIEW and MATLAB joint programming to realize the condition monitoring, vibration feature extraction, fault diagnosis and pre-warning, data management of the mine's main ventilator. The function verification of the system each module was completed in the coal mine production site. The results show that the proposed feature extraction, fault diagnosis and pre-warning methods all achieve the expected functions and meet the needs of field applications.