无监督稀疏模式诊断：感应热成像系统的缺陷检测

131 浏览量更新于2024-08-29 收藏 1.74MB PDF 举报

"感应热成像系统对缺陷的无监督稀疏模式诊断" 本文"Unsupervised Sparse Pattern Diagnostic of Defects With Inductive Thermography Imaging System"发表在2016年2月的《IEEE Transactions on Industrial Informatics》第12卷第1期，由Bin Gao、Wai Lok Woo、Yunze He和Gui Yun Tian等几位IEEE资深会员共同撰写。该研究论文提出了一种无监督的方法，用于通过感应热成像系统进行缺陷诊断和监控。该方法的创新之处在于其完全自动化，不需要用户手动选择特定的热帧图像来进行缺陷诊断。文章的核心是结合了基于物理的感应热机制与基于信号处理的模式提取算法。具体来说，它采用了一种称为稀疏贪婪主成分分析（SGPCA）的算法。SGPCA是一种统计分析工具，能够从复杂的数据集中提取出最重要的特征。在本研究中，SGPCA被用来自动识别和定位缺陷，同时，研究者在算法内部构建了一个功能，以控制SGPCA的稀疏度，从而提高缺陷尺寸估计的准确性。感应热成像是利用电磁场引起材料温度变化来检测内部缺陷的技术。当材料被加热并随后冷却时，其内部的缺陷会影响热量的分布，从而在热成像中显示出来。通过无监督学习，这种方法可以无需先验知识或标记数据就能识别异常模式，即潜在的缺陷。为了证明所提方法的有效性，研究人员在金属样本上进行了自动诊断缺陷的实验，并对比了其他方法的性能。实验结果表明，提出的无监督方法在识别和准确估计缺陷尺寸方面表现出色。这些实验和比较进一步证实了该方法在工业应用中的潜力，尤其是在自动化检测和监测缺陷的场合，例如质量控制、结构健康监测等领域。这篇研究论文提出了一种利用感应热成像系统和无监督学习技术的新方法，该方法能够在没有人为干预的情况下有效地识别和量化材料的内部缺陷，具有较高的精度和自动化程度，对于工业界和科研领域都具有重要的实践意义。

GAO et al.: UNSUPERVISED SPARSE PATTERN DIAGNOSTIC OF DEFECTS 373

Fig. 1. ECPT system.

where T = T (x, y, z, t) is the temperature distribution, k is the

thermal conductivity of the material (W/m · K), which is depen-

dent on temperature. ρ is the density (kg/m

), C

is speciﬁc

heat (J/kg · K). q(x, y, z, t) is the internal heat generation func-

tion per unit volume, which is the result of the EC excitation.

The variation of temporal temperature depends on the spatial

temperature variation for heat conduction. According to (2),

heat conduction is inﬂuenced by T (x, y, z, t), ξ, υ, σ, μ, and

l, where ξ denotes the sensor geometry factor; υ denotes the

parameters of the excitation (frequency and amplitude) and l

denotes the lift off (distance between the sensor and sample).

From the above analysis, it becomes clear that the variation of

temperature spatially and its transient response recorded from

the IR camera directly reveals the intrinsic properties variation

of the conductive material.

2) IT Defect Detection: Fig. 1 shows the diagram of IT

defect detection system. The excitation signal generated by

the excitation module is a small period of high-frequency cur-

rent. The current in the coil will induce the ECs and generate

the resistive heat in the conductive material. The heat will

diffuse in time until the heat reaches equilibrium in the mate-

rial. If a defect (e.g., crack and fatigue region) is present in

the conductive material, EC distribution as well as heat diffu-

sion process will vary. Consequently, the spatial distribution of

temperature on the surface of material and the temperature tran-

sient response will show the variation, which is captured by an

infrared camera. It can be divided into two phases: 1) heating

phase; and 2) cooling phase. As an example, we take a ﬁnite

length sample with small penetrated slot as a defect testing sam-

ple. The resultant heating frame from IT (0.1 s) is presented in

Fig. 1 right bottom panel. In the heating phase, different heat

generation rates enlarge the t emperature spatial variation. Hot

spots are observed around the slot tips and the cool areas locate

at both sides of the slot. In the cooling phase, heat diffuses from

high-temperature area to low-temperature area, and reduces the

contrast. In addition, the area located far away from excitation

coil will continually rise temperature because of its heat dif-

fusion. These different areas can be considered as the pattern

regions which share the similar transient responses in the sam-

ple. The infrared camera functioned as a temperature spatial

image signal recorder along with time ﬂowing. In addition, the

camera actually records the mixed image signal correspond-

ing to the signal image from the thermal pattern regions at

each time point. These regions are termed as thermal patterns

in IT.

The hot spots are used specially for defect location and siz-

ing. Fig. 1 shows the example of temperature distribution at the

sample surface as the lift-off distance is set d =4mm. When

the inductor is close to the tip of the defect (d =4mm), it is

seen that signiﬁcant EC ﬂows around the tip of the defect and

the defect behaves predominantly as a slot.

3) Relationship Between Excitation System, Heating

Phase, and Cooling Phase With Respect to Material

Variation: IT uses inductive heater as the excitation system.

Hence, it is speciﬁc for conductive materials or multilayer sys-

tem with conductive layer. The inductive heat depends on the

parameters of material and excitation signal.

1) To optimize the signal-to-noise ratio (SNR), the heat

power should be maximized. Normally, the high-current

amplitude (hundreds of ampere) and great frequency

(>100 kHz) are used.

2) The longer heating time results in the accumulation of

large amount of heat Q. For detecting surface defects, the

long heating time is useful for good SNR and contrast

[18]. At the same time, the longer heating time is useful

for detecting the deep defects due to heat conduction from

surface to deep defects [51].

3) The long cooling time is useful for heat conduction to

detect deep defects. The thermal penetration depth for a

pulsed excitation is determined by the thermal diffusivity

of the material and by the observation time t (cooling

time) after pulse heating.

4) The small electrical conductivity can lead to a high heat

power and great EC penetration depth. The thermal diffu-

sivity depends on thermal conductivity, mass density and

speciﬁc heat. If the thermal diffusivity is large, the tem-

perature changes quickly. Hence the sampling frequency

of the camera must be high enough to capture the changes

of temperature.

4) Detectability: In general, the ECPT is valid for both

deep and shallow defects, which is based on the physics

principles of inductive heating, heat conduction, and infrared

radiation. According to skin effect in inductive heating, the

EC has a penetration depth δ, which is expressed by equation

δ =

√

fπμσ

, where f is the frequency of the pulsed excitation.

If the shallow defects are in this skin area, they will dis-

turb the EC distribution and then the temperature distribution.

Theoretically, the lower excitation frequency has deeper detec-

tion depth. In practice, in order to improve the heating efﬁcient

with IT, the frequency is about 100 kHz and the skin depth is

relative small. For example, ferromagnetic metals have a much

smaller skin depth (about 0.04 mm at 100 kHz). Therefore,

shallow/surface defects can be detected. The heat will con-

duct to the interior and lateral area of the material. If the deep

defects disturb the heat conduction process, the surface temper-

ature distribution will be different from the surrounding area.

The heat conduction is used to detect the deep defects [51].

The temperature will be captured by IR camera. If the surface

defects show a different emissivity value, the temperature will

be different from the surrounding area.

5) IT Multiphase Analysis: According to (2), the ﬁrst-

order derivative of the temperature response of transient

response, as shown in Fig. 2(b), is composed of heat diffu-

sion and Joule heating [52]. In Fig. 2(c), we can infer from the

thermal video that the heat conduction procedure can be divided

into six phases.

剩余12页未读，继续阅读

weixin_38710524

粉丝: 7
资源: 884

无监督稀疏模式诊断：感应热成像系统的缺陷检测

热成像系统介绍

基于MATLAB的热成像系统仿真和性能度量方法研究.pdf

基于Linux的嵌入式红外热成像系统设计.pdf

matlab热成像 缺陷检测

c# 红外热成像系统

红外热成像实时测温系统的未来规划

基于红外热成像的无人机检测与识别系统

基于stm32的amg8833热成像显示系统设计

dde热成像+matlab

最新资源

matlab热成像缺陷检测