登录 EN

添加临时用户

印刷电路板钻削钻头磨损与钻孔质量监测研究

Research on Drill Wear and Hole-drilling Quality Monitoring in the PCB Drilling Process

作者:谈齐峰
  • 学号
    2016******
  • 学位
    博士
  • 电子邮箱
    tqf******.cn
  • 答辩日期
    2023.05.26
  • 导师
    李勇
  • 学科名
    机械工程
  • 页码
    116
  • 保密级别
    公开
  • 培养单位
    012 机械系
  • 中文关键词
    PCB钻孔,钻头磨损,钻孔质量,多信号和多模型融合,监测模型
  • 英文关键词
    PCB drilling,drill wear,hole-drilling quality,multi-sensor and multi-model fusion,monitoring model

摘要

印刷电路板(Printed circuit board, PCB)是不可缺少的电子产品元部件,而PCB孔的加工是关键工序之一。PCB钻头在钻孔时发生磨损会导致钻孔质量下降,所以钻头磨损与钻孔质量的监测尤为重要。本文采用在线监测技术通过采集和分析加工过程中的传感器信息,对钻头磨损和钻孔质量进行及时准确的辨识与监测,从而保证钻孔质量,降低生产成本,提高生产效率。首先提出了一种基于电流信号的模型驱动钻头磨损监测方法,通过微元法建立了考虑钻削参数和钻头磨损的钻孔过程钻头所受负载的半经验解析模型。通过建立钻削负载与主轴电流的解析模型,最终建立了钻头磨损、钻削负载和主轴电流间的解析模型。通过变钻削参数实验采集了钻削过程中的钻削力和主轴电流,并且采用最小二乘法确定了解析模型中的系数。实验结果表明所提出的方法可以实现基于主轴电流信号的模型驱动钻头磨损监测模型的建立,为后续研究提供了理论准备和支撑。为实现钻头磨损特征的提取,提出了一种基于自适应多稳态随机共振模型的微弱信号处理方法。对多稳态随机共振的结构特点和性能优势进行研究,并通过信号仿真加以验证。采用参数补偿方法克服随机共振模型只能处理低频信号的问题,并且采用差分进化算法自适应地调整多稳态模型系统的参数。对PCB板钻头钻削过程中的振动信号进行处理,转频和齿频的平均信噪比提升,增强了钻头磨损的微弱特征。在主轴转频处,处理后钻头磨损的信号频谱值呈线性增大关系,说明采用该方法进行信号处理后的转频频谱可以提取为钻头磨损特征信号。为了增加钻头磨损的监测精度,提出了多信号和多模型融合的钻头磨损监测方法。根据电流、振动和声发射信号的特点提取了信号特征,通过深度神经网络模型实现了多信号数据驱动融合钻头磨损监测。并且采用改进的卡尔曼滤波算法实现模型和数据融合钻头磨损监测模型的建立,进一步提高了模型精度和泛化性能。最后提出了考虑钻头磨损和钻削参数的基于多输出支持向量回归模型钻孔质量多目标监测方法。该监测模型对孔壁粗糙度以及钉头厚度的监测精度高,对孔位精度的监测精度相对较低。与其他多输出机器学习模型相比,有着更好的监测性能。上述研究完成了PCB钻孔物理过程建模、钻头磨损微弱信号处理、多信号和多模型融合监测钻头磨损以及钻孔质量多目标监测,为实现PCB钻孔过程钻头磨损和钻孔质量的在线监测提供了可行的技术基础。

Printed circuit board (PCB) has become an indispensable part of electronic products, and the processing of PCB holes is one of the key PCB manufacturing processes. Drill wear in the PCB drilling process will lead to the degradation of drilling quality, so the monitoring of drill wear and hole-drilling quality is particularly important. In this study, online monitoring technology is used to monitor drill wear and hole-drilling quality timely and accurately by collecting and analyzing sensor information during processing, which will ensure hole-drilling quality, reduce production cost and improve production efficiency.First, an analytical model of the drilling load on the drill in the drilling of PCB considering the drilling parameters and drill wear was established by the micro-element method and semi-empirical formula. By establishing the relationship between drilling load and spindle current, the analytical model between drill wear, drilling load and spindle current was finally established. The drilling force and spindle current during the drilling process were collected through experiments with variable drilling parameters, and finally the coefficients in the analytical model were determined using the least squares method. The experimental results show that the proposed method can realize the establishment of drill wear monitoring model based on spindle current signal, which provides theoretical preparation and support for the subsequent research.To realize the extraction of drill wear features, a weak signal processing method based on the adaptive multi-stable stochastic resonance model is proposed. The structural characteristics and performance advantages of the multi-stable stochastic resonance were investigated and verified by signal simulation. A parameter compensation method was used to overcome the problem that the stochastic resonance model could only process low-frequency signals, and a differential evolution algorithm was used to adaptively adjust the parameters of the multi-stable model. The vibration signals of drilling process were processed by the model, and the average signal-to-noise ratios of the rotational and tooth frequencies were improved to enhance the signal feature of drill wear. At the spindle rotational frequency, the signal spectrum value of the processed drill bit wear increases linearly, which means the rotational frequency spectrum after signal processing by this method can be extracted as the drill wear feature.In order to increase the monitoring accuracy of drill wear, a multi-sensor and multi-model fusion method for monitoring of drill wear is proposed. The signal features were extracted according to the characteristics of current, vibration and acoustic emission signal, and the multi-sensor fusion was implemented to monitor drill wear by a deep neural network model. An improved Kalman filter algorithm was used to realize the model and data fusion for monitoring drill wear, which further improved the accuracy of monitoring model. The multi-objective model of hole-drilling quality considering drill wear and drilling parameters was performed based on a multi-output support vector regression model. The monitoring error of the hole wall roughness as well as the mean value of the nail head is small, and the monitoring error of the hole accuracy is relatively larger. Compared with other models, it has better monitoring performance.The above research has completed the process modeling of PCB drilling considering drill wear, weak signal enhancement of drill wear features, fusion model establishment of multi-sensor and multi-model to monitor drill wear and hole-drilling quality, which provides a feasible technical basis to realize online monitoring of drill wear and multi-objective prediction for hole-drilling quality in PCB drilling process.