基于改进贝叶斯算法的黏滑振动等级评估研究

Evaluation of Stick-Slip Vibration Grade Based on Improved Bayesian Algorithm

黏滑振动会导致钻井效率降低,是影响钻头和井下工具寿命的重要因素。为了评估黏滑振动严重程度,通过对井下近钻头测量参数与地面录井参数的综合分析,得到了衡量黏滑振动等级指标。通过对近钻头测量参数进行时频域分析,采用主成分分析法(PCA),建立了一种基于差分演化算法的属性加权朴素贝叶斯(DE-AWNB)改进模型,在朴素贝叶斯分类算法中加入属性权重,通过属性加权法估计后验概率,利用差分演化算法寻找最优权重属性。试验结果表明,DE-AWNB算法的分类精度可达92.38%,收敛时间可达4.95 s。改进贝叶斯算法在黏滑振动等级评估工程应用上明显优于传统贝叶斯算法、随机森林法和遗传算法属性加权朴素贝叶斯(GA-AWNB)算法。将该模型应用于实际钻井工程,能够有效提高黏滑振动识别水平,提高钻井效率。

Stick-slip vibration is an important factor affecting the service life of drill bits and downhole tools,which will leading to a decrease in drilling efficiency.In order to evaluate the severity of stick-slip vibration,an integrated analysis was conducted on the downhole near-bit measuring parameters and the surface mud logging parameters to obtain indicators for measuring the stick-slip vibration grade.By means of conducting time-frequency domain analysis on near-bit measuring parameters,the principal component analysis(PCA)was used to build an improved differential evolution algorithm based attribute weighted naive Bayes model(DE-AWNB);attribute weights were added to the naive Bayes classification algorithm,the attribute weighting method was used to estimate the posterior probability,and the differential evolution algorithm was used to find the optimal weight attribute.The test results show that the classification accuracy of the DE-AWNB algorithm can reach 92.38%,and the convergence time can reach 4.95 s;the improved Bayes algorithm is apparently superior to traditional NB,RF and genetic algorithm attribute weighted naive Bayes(GA-AWNB)algorithms in the evaluation of stick-slip vibration grade.Application of this model to actual drilling engineering can effectively improve the stick-slip vibration recognition level and improve the drilling efficiency.