Separating Bit and Lithology Effects from Drilling Mechanics Data

Abstract

Abstract Most real time drilling interpretation is performed using some sort of normalized rate of penetration to infer changes in lithology and diagnose the condition of the bit. This frequently leads to ambiguous or confusing results because it is impossible to separate all bit effects from lithology effects with only one measurement. This paper shows how a second measurement, downhole torque, can be used with rate of penetration to confidently separate the bit effects from the lithology effects when drilling with milled tooth or PDC bits. Simple drilling equations for rate of penetration and bit torque are derived and compared with well known formulae. These lead to simple "quicklook" interpretations which can be made at the wellsite with a hand calculator. A detailed analysis can be performed on a wellsite computer. Many field case studies were made to examine the effects of lithology changes on the drilling response of milled tooth and PDC bits. A few of the studies are presented The results indicate that:changes in bit torque can be used to broadly classify the lithology into three categories, namely: porous, argillaceous (shaly), and tight, corresponding to high, medium and low torque respectively,trends in bit torque and rate of penetration (ROP) in shale type formations can be used to separate the wear of milled tooth and PDC bits from changes in shale strength,it was not possible to interpret bit wear in non-shaly type formations,surface drilling measurements can be insensitive to major formation changes (eg. sand/shale boundaries), particularly in deviated wells. Introduction In recent years there have been many papers describing models and methods to predict and interpret the drilling response of bits. Winters et al refer to most of the roller cone models. Gault et al and Glowka and Stone refer to many of the PDC models. Few of these models have proven to be practical in the field, or to have found validity over a wide number of environments. There are several reasons for the apparent lack of commercial success. Many models have been derived using data from laboratory drilling tests. Since representative rock samples are hard to find, the experiments tend to cover a wide range of weight on bit and rotary speed but on a narrow range of rocks. This tends to result in models which attempt to simultaneously describe a wide range of effects (such as individual crater formation at low weights, through crater indexing, through to poor bit cleaning at the upper end of weight on bit). This leads to fairly complex models sometimes requiring four or more empirically derived parameters. Clearly this is impractical in field operations. Furthermore laboratory data tend to be somewhat ideal. Field data, on the other hand, tend to contain a significant percentage of noise from the rather crude manner in which rig measurements of weight, or rate of penetration, are made. For example pipe stretch is not taken into account in the rate of penetration; drag results in the surface measured weight being higher than the bit weight; depth and weight measurements might not be made and averaged simultaneously etc. It does not make sense to fit a complicated model through a noisy cloud of data. Unlike laboratory experiments, the driller tends to drill a wide range of formations with a fairly narrow commercial range of weight on bit, rotary speed, and pump strokes. What is required is a model that is robust for formation changes, rather than a model which is robust for large changes in inputs such as weight, rotary speed and flowrate. Furthermore the driller tends to measure only one dependent variable, the rate of penetration. These reasons explain why Jordan and Shirley's well known d-exponent method has had so much success - it computes one parameter from one dependent variable, and is fundamentally simple! P. 123^