激光焊接时光感检测信号具有混合信息数量多、 时域及频域变化复杂、 信号随机波动剧烈等明显的混沌信号特征, 常规检测手段对激光焊接熔透特征信息难以有效分离。 在激光焊接熔透特征信号同轴增效提取技术基础上, 针对光谱透视成像中提取出的复杂混合信号的分布特征, 提出一种大功率固体激光焊熔透特征混沌信号解析方法。 具体通过统计方法提取混沌检测信号的各瞬态概率密度函数, 然后动态形成一系列具有趋势性特征的简化图形模式, 并根据图形的多项形貌特征与熔透行为建立复合关联识别。 文中重点介绍了基值、 离散度和畸变率3种特征参数的定义及数学模型, 这些特征参数不但可以对特征模型的重心位置、 离散情况和单向畸变情况进行准确描述, 而且通过实验方法还充分验证了不同特征参数与激光焊接熔透状态的关联特性。 如基值特征参数在焊缝熔透程度较大时具有明显的识别效果, 但是熔透程度较小时其波动幅值较大; 而离散度特征参数只对较小熔透状态测试敏感, 可以对基值进行有效补充, 但是对较大熔透状态的识别存在不足; 畸变率特征参数虽然在适度熔透附近时数值波动剧烈, 但是对于焊缝熔透程度较小或较大时都能通过畸变率数值关系对熔透程度进行较好关联识别, 有效弥补了基值和离散度的检测盲区。 因此, 基值、 离散度及畸变率特征参数间具有相互补充、 相互验证关系, 通过它们的特征曲线进行复合识别可以对激光焊接熔透状态实现较好的定性及定量检测。 实验结果表明, 混沌信号解析方法结合激光焊接熔透特征信号同轴增效提取技术后, 可实现光谱学、 光学、 统计学、 抽象化解析等的多维复合识别效果, 可有效增强大功率固体激光焊的熔透检测能力, 实现可靠熔透在线识别。

Because the optical signal has the characteristics of chaotic signal obviously in laser welding, such as large number of mixed information, complex changing in the time domain and the frequency domain, fluctuation violently of random signals, and so on, it is difficult to separate the characteristic information of penetration by conventional means in laser welding. So, according to the distribution characteristics of complex mixed signals extracted from spectral perspective imaging, a new chaotic signal analytical method for high power solid state laser welding is proposed in this paper, which bases on the synergistic extraction method for coaxial penetration characteristic signal. The method includes extracting each transient probability density function of the chaotic signal, dynamically forming a series of simplified feature models with trend recognition feature, and establishing compound correlation recognition between the multi-morphological characteristics of the feature models and penetration state. In this paper, three kinds of characteristic parameters and their definitions and mathematical models were introduced in detail, and there are basic value, dispersion degree and distortion rate, which not only can accurately describe the different characteristics of the feature model, but also have much relationship with different penetration states of laser welding, by experiment verifying. For example, the base value characteristic parameter has obvious recognition effect when the weld penetration degree is bigger, but has much fluctuation amplitude when the penetration degree is smaller. The dispersion degree characteristic parameter is only sensitive to the small penetration state test, which can effectively supplement the basic value, but there are some shortcomings in the larger penetration state. The distortion rate characteristic parameter although fluctuates violently near the moderate penetration rate, when the penetration degree of weld is smaller or larger, the characteristic parameter can get better recognition, which the characteristic parameters of the basic value and the dispersion degree can't do. Therefore, in the characteristic parameters of basis value, dispersion degree and distortion rate, there are relationship of complementary and mutual verification. By the composite identification of their characteristic curves, the laser welding penetration state can be detected qualitatively and quantitatively .The experimental results showed that, after the chaotic signal analytical method combined with the coaxial synergistic extraction method, it can realize the multidimensional compound recognition effect of spectroscopy, optics, statistics, abstract analysis, etc, which can enhance the penetration detection ability of high power solid laser welding effectively, and realize reliable on-line identification of penetration.