Fault diagnosis method of rolling bearing of mine main fan based on transfer learning
-
摘要:
矿用主扇风机滚动轴承的状态监测与故障诊断研究对煤矿生产安全具有重要意义。现有的滚动轴承故障诊断方法在实际工况中进行直接应用时存在训练不足、故障诊断准确率不足的问题,且矿用主扇风机滚动轴承长期处于正常运行状态,正常样本的数量远多于故障样本,即存在样本不平衡问题。因此,提出一种基于迁移学习的矿用主扇风机滚动轴承故障诊断方法(TLCNN+加权交叉熵损失)。该方法将常规滚动轴承数据作为源域数据,将矿用主扇风机滚动轴承数据作为目标域数据。首先利用对称极坐标(SDP)方法将振动信号转换为SDP图像;然后利用充足的源域图像样本对常规滚动轴承故障诊断模型进行训练,训练完成后将诊断模型的参数迁移至矿用主扇风机滚动轴承故障诊断模型中;其次迁移过程中对低层网络进行锁定并通过目标域图像样本对模型的高层网络进行微调,便可得到参数权重优化后的矿用主扇风机滚动轴承故障诊断模型。最后,为了解决样本不平衡问题,在模型中添加了加权交叉熵损失函数进行训练,使诊断模型对作为少数类的故障样本赋予更高的权重并在诊断过程中更加关注故障样本,从而提高诊断准确率。为了验证提出方法的有效性,通过常规滚动轴承故障试验台与实际工况中的矿用主扇风机滚动轴承数据进行了试验验证。结果表明所提方法可以对矿用主扇风机滚动轴承的运行状态进行准确识别分类,准确率达99.28%。
Abstract:The condition monitoring and fault diagnosis of the rolling bearings of the main fan in the mine are significant to the safety of coal mine production. The existing fault diagnosis methods of rolling bearing have the problems of insufficient training and accuracy when applied directly in actual working conditions. Moreover, the rolling bearings of the mine main fan are in normal operation for a long time, and the number of normal samples is much more than the faulty samples, so there is a sample imbalance problem. Therefore, this paper proposes a fault diagnosis method for rolling bearings of mine main fan based on transfer learning. The method takes the conventional rolling bearing data as the source domain data and the mine main fan rolling bearing data as the target domain data. Firstly, the one-dimensional vibration signal is converted into two-dimensional SDP images using the SDP method, and then the conventional rolling bearing fault diagnosis model is trained using sufficient source domain image samples. After training, the parameters of the diagnostic model are transferred to the mine main fan rolling bearing fault diagnosis model, and the lower layer network is locked and the higher layer network of the model is fine-tuned by the target domain image samples during the transfer process, and finally the mine main fan rolling bearing fault diagnosis model with optimized parameter weights is obtained. Meanwhile, in order to solve the sample imbalance problem, a weighted cross-entropy loss function is added to the model for training, so that the diagnosis model gives higher weights to the fault samples as a minority class and pays more attention to the fault samples in the diagnosis process, thus improving the diagnosis accuracy. In order to verify the effectiveness of the proposed method, this paper uses a conventional rolling bearing fault test bench and the rolling bearing data of the mine main fan fan in actual working conditions for experimental verification. The results show that the proposed method can accurately identify and classify the operating status of the mine main fan rolling bearings, and the accuracy rate is 99.28%.
-
0. 引 言
我国1 500~3 000 m的煤层气地质资源量约为30.37×1012 m3,为1 500 m以浅煤层气资源量的2倍[1-2]。随着深部煤层气勘探开发深入,游离气含量受到关注,游离气含量高的干煤系统中具有见气时间早、见气产量高的特点[2]。游离气的保存除了受煤岩的自封闭作用,顶板的封盖性也是重要的影响因素。煤层气富集受构造特征影响明显,在鄂尔多斯盆地东缘发育有缓倾斜、单斜、背斜、向斜、推覆构造、逆断层和挤压型层滑等7种有利的富气构造[3]。不同微构造部位的煤层含气性、渗透性、压裂改造效果和气水产出存在明显差异,同时构造高部位会形成动态气藏,影响煤层气开发全周期[4]。尽管煤层气是连续型的天然气藏,但是也具有“源岩控储”和“物性控藏”的特征,在高孔高渗区容易形成富集甜点[5]。可见,顶板封盖性是影响煤层气聚集的重要影响因素,关联和影响甜点区优选和开发井网设计。
煤层直接顶板包括泥岩、灰岩和砂岩等,均可以形成封盖条件,但是具体封盖效果受裂缝发育情况、储层物性等因素综合影响。通过鄂尔多斯盆地的煤层气和煤层邻近致密气勘探开发证实,部分地区煤层直接顶板砂岩具有良好含气性,开发效果优于远距离砂岩[6]。深浅侧向、密度、声波时差和补偿中子等测井数据可以构建孔隙率和裂缝模型来评价顶底板封盖条件,形成了结合测井动态弹性模量和岩心静态结合的裂缝强度指数计算模型[7]。地震数据也可以用来解释顶板岩性和封盖性,但是总体精度不高,很难解释微小构造变化影响的煤层顶板差异。微电阻率成像测井可以采集更多地层信息,且处理后的动态图像可以直观反映裂缝发育情况等信息[8-11]。裂缝的发育方向可以指示地应力的方向,地应力研究对地质与工程具有重要的意义,可为后期的压裂射孔措施提供有效依据,提高工程作业质量和效率[12-16]。
为推动煤层开发地区深部煤层气勘探开发突破,笔者基于微电阻率成像测井手段,刻画不同岩性在成像测井上的响应特征,建立顶板岩性识别图版;明确高导缝与诱导缝在成像测井上的识别特征,分析煤层顶板裂缝发育情况;进一步结合顶板岩性、厚度与裂缝发育情况等,划分不同封盖性的煤层顶板组合条件类型,相关工作可以有效支撑深部煤层气甜点区带优选和井位部署。
1. 区域地质背景
鄂尔多斯盆地总面积约为25×104 km2,横跨山西、陕西两省,整体轮廓呈现矩形,构造位置属于华北地台西部。盆地内划分6个二级构造单元,包括伊盟隆起、渭北隆起、晋西挠褶带、天环坳陷、西缘冲断带以及伊陕斜坡[17-20]。本文所用钻井和测井数据主要来自于鄂尔多斯盆地东北缘,构造上属于晋西挠褶带,整体为单斜构造,地层倾角小[20]。鄂尔多斯盆地整体发育太原组和山西组2套主力煤层,下部太原组的8号煤层及其顶板是此次分析重点。
2. 顶板岩性识别
利用电成像测井,可以较为精确地识别煤层顶板岩性,评价顶板裂缝发育情况与连通性,为煤层及顶板封盖性精细评价提供有利支撑。综合参考钻井取心资料、常规测井图像特征等划分岩性,共识别出砾岩、粗砂岩、中砂岩、细砂岩、粉砂岩、泥质粉砂岩(砂质泥岩)、碳质泥岩、泥岩、灰岩及煤等10种岩性(图1)。
综合常规测井资料与电成像特征等,对以上10种岩性的特征进行了系统总结。砾岩,常规测井曲线上GR为低值,密度一般在1.5~4.5 g/cm3,声波时差较低,电成像静态图像以暗色、褐色为主,动态图像上可具有亮色斑点状特征,发育块状层理;粗砂岩,GR为低值,密度通常在2.5 g/cm3,声波时差较高,静态图像呈暗色,动态图像上发育交错层理,具有砂质感;中砂岩,GR为低值,密度介于2.2~2.8 g/cm3,声波时差较高,静态图像上呈暗色,动态图像上发育交错层理,砂质感较粗砂岩略细;细砂岩,GR为低值,密度一般大于2.4 g/cm3,声波时差较高,静态图像以黄褐色为主,动态图像上发育交错层理,层理细腻,颗粒感较差;粉砂岩,GR值较高,密度介于2.0~2.4 g/cm3,声波时差较高,静态态图像以橘黄色为主,动态图像上发育波状层理;泥质粉砂岩(砂质泥岩),GR较高,密度介于2.6~2.9 g/cm3,声波时差较高,静态图像上以亮黄色或亮白色为主,可见暗色或黄色块状,动态图像上可见黄色斑点和亮色条带;碳质泥岩,GR为高值,密度介于2.66~2.77 g/cm3,声波时差较高,静态图像以黄褐色为主,动态图像上发育水平层理,见亮色条带及黑色斑点;泥岩,GR为高值,密度介于2.2~2.7 g/cm3,声波时差高,静态图像以亮黄色为主,动态图像上发育水平层理,见白色条带;灰岩,GR为低值,密度一般为2.7 g/cm3,声波时差较高,静态图像上呈亮色,动态图像上呈层状特征;煤,GR为低值,密度介于1.3~1.4 g/cm3,声波时差较高,静态图像呈亮色,动态图像上呈层状、块状特征。可总结出如下规律:碎屑岩静态图像由亮变暗,泥岩最亮,砾岩最暗。通过对该区深煤层顶板岩性的识别统计发现,煤层的顶板岩性大部分为泥岩,约占78%,其次为砂岩,约占15%,灰岩和其他岩性发育较少(图2)。
3. 裂缝类型识别及应用
煤层顶板裂缝会影响力学性能与渗透率,降低顶板岩石强度,导致煤层气逸散等情况,降低封盖能力,不利于煤层气的保存[21]。为评价深煤层顶板的封盖性优劣,需对裂缝的存在及裂缝的类型进行精确判断。基于成像测井资料对研究区内煤层顶板的裂缝发育情况进行综合分析可知,主要存在裂缝类型为构造缝与非构造缝,其中构造缝也称为天然裂缝,非构造缝也称为诱导缝,构造缝按充填特征可分为高导缝和高阻缝2类,高导缝充填物一般为泥质充填和黄铁矿充填等,高阻缝充填物一般为方解石、白云石和石英等非导电矿物。非构造缝缝根据成因可分为钻具震动缝、泥浆压裂缝、应力释放缝与井眼崩落,如图3所示。
构造缝与非构造缝在电成像上的成像特征有明显区别。构造缝在电成像图像上表现为较规则的正弦线,以高导缝为例(图4a),泥质充填缝表现为暗色连续的正弦曲线,缝面较规则,轻微溶蚀;张开缝表现为暗色断续的正弦曲线,缝面宽窄不一,局部有明显的溶蚀扩大现象。
诱导缝在电成像图像上多呈羽状,分布于对称的2条极板上。钻具振动缝在电成像图像上微小且延伸很短,呈羽毛状或雁行状(图4b);应力释放缝,表现为一组呈180°或接近180°对称分布的羽状纹理(图4c);井眼崩落,表现为2条较宽且呈180°或接近180°对称分布的暗色或黑色垂直条带或斑状(图4d);泥浆压裂缝,表现为2条呈180°或接近180°对称分布的黑色垂直条带,延伸较长,方向基本稳定。
诱导缝的形成与地应力有密切关系,因此借助诱导缝的发育方向可以有效判断地应力方向。通过对井壁应力进行分析,在最小水平主应力方向上有最大剪切应力。当应力超过岩石的抗剪强度,井壁就会产生崩塌。因此,井眼崩落的方向即为最小水平主应力方向(图5a)。由于裂缝较为发育,古构造应力大多被释放,地应力基本平衡,但在致密地层中裂缝不发育,且构造应力未被释放,因此地应力较大。当地层被钻开时,地应力释放,进而产生一组应力释放缝,该裂缝的方向即为最大水平主应力方向(图5b)。
4. 有利顶板组合条件
煤层含气量高低不仅受到煤层顶板封盖条件的影响,还受到煤岩演化程度、构造条件、水动力条件、地应力条件等地质因素综合影响,此外,煤层顶板中存在的裂缝对其力学性能与纵向渗透率影响较大,导致顶板岩石强度降低,井眼失稳,以及煤层气逸散等情况出现,降低封盖能力,不利于煤层气保存,进而影响含气量[22]。但在同一地区,其他地质因素相似的情况下,煤层含气量主要与煤层顶板封盖性优劣有关,因此,需划分出有利的煤层顶板组合条件,为深部煤层气甜点区带优选和井位部署提供支撑(表1)。
表 1 煤层顶板组合条件分类Table 1. 1 Classification of coal seams and roof conditions级别 岩性特征 裂缝特征 厚度 含气量 静态图像 顶板厚度 直接顶板 岩性 裂缝 I 顶板以灰岩、泥岩等渗透性极差的
岩性为主,且厚度大裂缝发育很少或者基本无裂缝发育 大 高 亮色 厚 泥岩、灰岩 不发育 II 顶板以泥岩、碳质泥岩等渗透性极差的
岩性为主,但是厚度较薄由于厚度较薄,受诱导缝
影响易产生裂缝较大 较高 橘黄色 较薄 粉砂岩 较少 III 顶板岩性以砂岩为主 由于岩石强度较低,受诱导缝
影响易产生裂缝薄 低 褐色 薄 砂岩 发育 I类组合。顶板裂缝发育很少或者基本无裂缝发育, 此类顶板渗透性极差且厚度大,基本无裂缝发育,是最好的顶板封盖层。煤层厚度大(一般大于6 m),煤层含气量高(大于15 m3/t),在成像测井中,其静态图像呈现亮色,煤层直接顶板为泥岩、灰岩等,裂隙不发育(图6a)。II类组合。顶板层渗透性较差,但是厚度较薄,受诱导缝影响发育较孤立或细碎的裂缝。煤层厚度较大(3~6 m),煤层含气量较高(10~15 m3/t),在成像测井中,其静态图像呈现橘黄色,煤层直接顶板为粉砂岩,发育较少裂隙(图6b)。III类组合。顶板岩石在诱导缝的影响下裂缝发育。煤层较薄(小于3 m),煤层含气量较低(小于10 m3/t),在成像测井中,其静态图像呈现褐色,煤层直接顶板为砂岩,裂隙发育(图6c)。
5. 结 论
1)结合常规测井资料与微电阻率成像测井特征,基于静态电成像图可以有效区分砾岩、砂岩和泥岩等10类岩性,主要特征为碎屑岩静态图像由亮变暗,泥岩最亮,砾岩最暗,该煤层气开发地区煤层顶板以泥岩为主、砂岩次之。
2)煤层顶板裂缝主要发育有高导缝与诱导缝,高导缝包括张开缝和泥质充填缝,在成像图像上表现为正弦曲线;诱导缝可分为钻具震动缝、泥浆压裂缝、应力释放缝与井眼崩落,在成像图像上呈现为羽状或雁行状排列,可以用于判断地应力方向。
3)综合顶板岩性、厚度、裂缝发育情况等,该煤层气开发地区可划分出封盖性不同的3类煤层顶板组合,其中I类顶板电成像图像一般呈亮色且裂缝发育少;II类顶板发育较多裂隙;III类顶板图像显示裂隙切割且图像不清晰。
-
表 1 故障诊断试验数据集
Table 1 Experimental data set for fault diagnosis
数据集 数据来源 故障状态 训练集
样本数量测试集
样本数量I 常规滚动轴承 正常状态 1000 200 内圈故障 1000 200 外圈故障 1000 200 滚动体故障 1000 200 II 矿用主扇风机
滚动轴承正常状态 800 200 故障状态 200 200 表 2 故障诊断模型架构
Table 2 Fault diagnosis model architecture
网络架构 参数设置 输入层 Inputs 224×224 卷积模块1 Conv1 3×3-64 Conv2 3×3-64 Max pool 2×2 卷积模块2 Conv1 3×3-128 Conv2 3×3-128 Max pool 2×2 卷积模块3 Conv1 3×3-256 Conv2 3×3-256 Max pool 2×2 卷积模块4 Conv1 3×3-512 Conv2 3×3-512 Conv3 3×3-512 Max pool 2×2 卷积模块5 Conv1 3×3-512 Conv2 3×3-512 Conv3 3×3-512 Max pool 2×2 全连接模块 FC-layer-1 4096 FC-layer-2 4096 输出层 Outputs 故障类别数 表 3 不同诊断方法的平均准确率
Table 3 Mean accuracy and standard deviation of different diagnostic methods
诊断方法 平均准确率/% TLCNN+加权交叉熵损失 99.28 TLCNN+交叉熵损失 97.23 CNN+加权交叉熵损失 94.88 CNN+交叉熵损失 90.05 -
[1] 刘洪文,李志常. 大型矿用风机配套电机轴承结构型式的研究[J]. 防爆电机,2010,45(2):7−9. doi: 10.3969/j.issn.1008-7281.2010.02.003 LIU Hongwen,LI Zhichang. Research on structure type of motor bearing for large mine fan[J]. Explosion-proof Electric Machine,2010,45(2):7−9. doi: 10.3969/j.issn.1008-7281.2010.02.003
[2] ZHAO R,YAN R,CHEN Z,et al. Deep learning and its applications to machine health monitoring[J]. Mechanical Systems and Signal Processing,2019,115:213−237. doi: 10.1016/j.ymssp.2018.05.050
[3] LEI Y,YANG B,JIANG X,et al. Applications of machine learning to machine fault diagnosis:a review and roadmap[J]. Mechanical Systems and Signal Processing,2020,138:106587. doi: 10.1016/j.ymssp.2019.106587
[4] LIU R,YANG B,ZIO E,et al. Artificial intelligence for fault diagnosis of rotating machinery:a review[J]. Mechanical Systems and Signal Processing,2018,108:33−47. doi: 10.1016/j.ymssp.2018.02.016
[5] WANG Y,HUANG S,DAI J,et al. A novel bearing fault diagnosis methodology based on SVD and one-dimensional convolutional neural network[J]. Shock and Vibration,2020:1−17. doi: 10.1155/2020/1850286.
[6] SHAO H,JIANG H,ZHANG H,et al. Rolling bearing fault feature learning using improved convolutional deep belief network with compressed sensing[J]. Mechanical Systems and Signal Processing,2018,100:743−765. doi: 10.1016/j.ymssp.2017.08.002
[7] 侯文擎,叶鸣,李巍华. 基于改进堆叠降噪自编码的滚动轴承故障分类[J]. 机械工程学报,2018,54(7):87−96. doi: 10.3901/JME.2018.07.087 HOU Wenqing,YE Ming,LI Weihua. Fault classification of rolling bearings based on improved stack noise reduction self-coding[J]. Chinese Journal of Mechanical Engineering,2018,54(7):87−96. doi: 10.3901/JME.2018.07.087
[8] SUN Y,LI S. Bearing fault diagnosis based on optimal convolution neural network[J]. Measurement,2022,190:110702. doi: 10.1016/j.measurement.2022.110702
[9] LI X,ZHANG W,DING Q. Cross-domain fault diagnosis of rolling element bearings using deep generative neural networks[J]. IEEE Transactions on Industrial Electronics,2018,66(7):5525−5534.
[10] ZHANG W,LI C,PENG G,et al. A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load[J]. Mechanical Systems and Signal Processing,2018,100:439−453. doi: 10.1016/j.ymssp.2017.06.022
[11] 张旭辉,潘格格,郭欢欢,等. 基于深度迁移学习的采煤机摇臂部滚动轴承故障诊断方法[J]. 煤炭科学技术,2022,50(4):256−263 ZHANG Xuhui,PAN Gege,GUO Huanhuan,et al. Fault diagnosis method of roller bearing of shearer rocker arm based on deep transfer learning[J]. Coal Science and Technology,2022,50(4):256−263.
[12] TONG Z,LI W,ZHANG B,et al. Bearing fault diagnosis under variable working conditions based on domain adaptation using feature transfer learning[J]. IEEE Access,2018,6:76187−76197. doi: 10.1109/ACCESS.2018.2883078
[13] TONG Z,LI W,ZHANG B,et al. Bearing fault diagnosis based on domain adaptation using transferable features under different working conditions[J]. Shock and Vibration,2018:1−12.
[14] DU Y,WANG A,WANG S,et al. Fault diagnosis under variable working conditions based on STFT and transfer deep residual network[J]. Shock and Vibration,2020:1−18.
[15] 雷亚国,杨彬,杜兆钧,等. 大数据下机械装备故障的深度迁移诊断方法[J]. 机械工程学报,2019,55(7):1−8. doi: 10.3901/JME.2019.07.001 LEI Yaguo,YANG Bin,DU Zhaojun,et al. Deep migration diagnosis method for mechanical equipment faults based on big data[J]. Chinese Journal of Mechanical Engineering,2019,55(7):1−8. doi: 10.3901/JME.2019.07.001
[16] 邵海东,张笑阳,程军圣,等. 基于提升深度迁移自动编码器的轴承智能故障诊断[J]. 机械工程学报,2020,56(9):84−90. SHAO Haidong,ZHANG Xiaoyang,CHENG Junsheng,et al. Bearing intelligent fault diagnosis based on uplifting deep migration autoencoder [J]. Chinese Journal of Mechanical Engineering,2019,56(9):84−90.
[17] 尹晓伟,江雪峰,王龙福. 风电机组轴承故障诊断与疲劳寿命研究综述[J]. 轴承,2022(5):1−8. YIN Xiaowei,JIANG Xuefeng,WANG Longfu. Review on fault diagnosis and fatigue life of wind turbine bearing[J]. Bearing,2022(5):1−8.
[18] 张运生,乔建伟,赵岩,等. YBF系列风机电机的滚动轴承[J]. 电气防爆,2011(2):12−15. doi: 10.3969/j.issn.1004-9118.2011.02.004 ZHANG Yunsheng,QIAO Jianwei,ZHAO Yan,et al. Rolling bearing of YBF series fan motor[J]. Electric Explosion-Proof,2011(2):12−15. doi: 10.3969/j.issn.1004-9118.2011.02.004
[19] 孟宗,关阳,潘作舟,等. 基于二次数据增强和深度卷积的滚动轴承故障诊断研究[J]. 机械工程学报,2021,57(23):106−115. doi: 10.3901/JME.2021.23.106 MENG Zong,GUAN Yang,PAN Zuozhou,et al. Research on fault diagnosis of rolling bearing based on secondary data enhancement and deep convolution[J]. Chinese Journal of Mechanical Engineering,2021,57(23):106−115. doi: 10.3901/JME.2021.23.106
[20] ZHANG X,MENG G,WANG A,et al. Study on fault diagnosis method for the tail rope of a hoisting system based on machine vision[J]. Advances in Mechanical Engineering,2022,14(8):1−13.
[21] XU X,LIU H,ZHU H,et al. Fan fault diagnosis based on symmetrized dot pattern analysis and image matching[J]. Journal of Sound and Vibration,2016,374:297−311. doi: 10.1016/j.jsv.2016.03.030
[22] SUN Y,LI S,WANG Y,et al. Fault diagnosis of rolling bearing based on empirical mode decomposition and improved manhattan distance in symmetrized dot pattern image[J]. Mechanical Systems and Signal Processing,2021,159:107817. doi: 10.1016/j.ymssp.2021.107817
[23] 邵思羽. 基于深度学习的旋转机械故障诊断方法研究[D]. 南京:东南大学,2019. SHAO Siyu. Research on rotating machinery fault diagnosis method based on deep learning [D]. Nanjing:Southeast University,2019.