套管井阵列声波测井效果评价

套管井声波测井有着重要的意义和广泛的用途。套管直径增大，套管波的幅度减小；水泥胶结越好，提取的纵波时差就越准确；水泥环厚度越大，纵波时差受水泥环的影响越大；下套管后，地层本身的变化也会引起声波时差的变化。实际测井资料显示，尽管套管尺寸不一，但固井质量是影响套管井声波测井的主要因素。在固井质量好的井段，提取的纵波时差能准确反映地层的纵波时差，可以不做校正直接应用。如果固井质量较差，则套管井声波测井受套管和水泥环的影响大，提取的纵波时差已不能准确反映地层的纵波时差。设计三层BP神经网络，对不同套管尺寸的声波样本数据进行网络学习，学习结束后得到相应的模型参数，输入过套管纵横波时差资料或过套管纵横波时差及中子测井资料，使用网络模型参数就能对过套管的声波纵波时差进行校正。校正结果表明，提取的横波时差越准确，套管内中子测井效果越好，校正效果就越好。

EVALUATION OF ARRAY SONIC LOGGING EFFECTS IN CASED HOLES

Sonic logging in cased hole is of great significance and has extensive application. The amplitude of casing wave decreases with increasing casing diameter; the better the cement bonding is, the more accurate the compressional wave slowness-time extracted is; the influences of cement sheath on the compressional wave lowness-time get larger along with the increasing thickness of cement sheath; and the change of the strata themselves after running casing also can contribute to the variation of SDT. Real logging data indicate that cement job quality is the major influential factor of sonic logging in cased holes regardless of the casing sizes. In intervals with good cement job quality, the compressional wave slowness-time extracted can accurately represent the compressional wave slowness-time of the strata, thus can be directly used without any correction. While in intervals with poor cement job quality, the influences of casing and cement sheath on sonic logging in cased holes are large, and the compressional wave slowness-time extracted cannot accurately represent the compressional wave slowness-time of the strata. A three-layer BP neural network is designed and the sampled acoustic wave data with different casing sizes are used to perform network training. The relevant model parameters are obtained after network training and can be used to correct the through casing compressional wave slowness-time when through casing compressional and shear wave slowness-time or through casing compressional and shear wave slowness-time and neutron logging data are input into the model. Correction results show that the effects of correction get better with the increasing accuracy of shear wave slowness-time extracted.