| Letters
to the Editor: In response to the article, Geostatistics in Geotechnical Engineering by Dr. Reginald Hammah (RocNews Spring 2003), we received the four responses below. This piece concludes with a final comment by Dr. Hammah. RocNews welcomes and appreciates responses to our articles, as well as to those from contributing engineers and academics in the field of geotechnical engineering. An exchange of ideas and points of view is always interesting to our readers. We look forward to publishing more Letters to the Editor in the future. Dear Rocscience, I have been a Rocscience "user" since it's inception, and have used just about every Rocscience product and I am generally impressed with their user-friendly and intuitive functionality, making it very easy to integrate data from a variety of sources and to undertake a variety of scenarios in a short time. I have recently read your "Geostatistics" article, and I also have a keen interest in this field. I have been fortunate enough to work on a few projects for the Western Australian mining industry, that have allowed me to introduce some geostatistical techniques as part of my consulting work. Snowden is a consulting company to the mining industry in almost all disciplines, yet is particularly reknowned for being leaders in the field of geostatistics and resource estimation. We have been fortunate enough to take advantages of these skills in some of our geotechnical engineering projects. I am currently in the throws of "re-researching" the discipline of geostatistics, as I can only vaguely remember the geostats course work I did as an undergraduate all those years ago!! Your suggested list of references in your article is excellent, which, funnily enough has also been suggested by our geostats gurus! After reading your article, I can see that Rocscience is interested in keeping an eye on the development of geostatistical techniques in geotechnical engineering. As a practitioner of geotechnical engineering, I too can see the benefits of geostatistics and I can see that new tools are indeed required to take advantage of these new developments. In this regard, I hope to offer a few comments from my experience in this field that may assist you in any potential developments you may be considering in this subject. The following comments are based primarily from our experience in using geostatistics on geotechnical engineering projects. We have found that the use of geostatistics on geotechnical engineering projects, is only cost effective and beneficial where significant geotechnical information is available, that is, where the amount of geotechnical data available is at least around 30-40% of the amount of total geological resource data. Where geotechnical data is less than around 30%, it becomes difficult to justify the use of geostatistical techniques, in terms of effort for information/understanding gained. In addition, the robustness and reliability of any "models" produced are sometimes questionable. We have also come across the dilemma of trying to combine various data sources, each of variable quality/reliability and quantity, to represent a single geotechnical parameter (eg using point load tests, field index tests, schmidt hammer tests and UCS tests, all used to represent intact rock strength). We had to resort to some of the more exotic geostats estimation methods, including indicator kriging, to undertake this task, which was quite time consuming and required considerable skill (thankfully we had the in-house expertise to do this). The use of indicator kriging was fortuitous, as it also allowed us to estimate the reliability of the predicted values. This has enabled us to critically examine the study area, highlighting areas of "higher risk", which may require further investigation and/or lower probabilities of failure / more stringent control and management requirements. This example highted that it may take a lot more development to produce routine "Suggested" methods and tools for estimation of the various geotechnical parameters (based on their data sources) before they can be used regularly by geotechnical engineers. Another dilemma that we faced when trying to use geostatistical methods, was that some geotechnical parameters cannot be estimated correctly, as they are not simple point/scalar attributes. For example, RQD and fracture frequency are "tensor"-like, as their value is dependant on the direction it is viewed and/or sampled. RQD and fracture frequency vary according to the fracture system of the rockmass (ie its anisotropy). If you have two holes intersecting at the same location, however oriented at different directions, you will have two different values of RQD or fracture frequency. Which value do you use? If you decide to develop/use some sort of bias correction/weighting to your RQD values based on the orientation of your holes, you must also know the defect structural patterns and their characteristics in the all the structural domains that the holes pass through. This can be quite a time consuming process. Some of the "deliverables" produced for our "geostats" geotechnical projects included 3-dimensional block models of the rock mass, which contain parameters such as UCS, RQD, fracture frequency, GSI, density, rock type, and "rock mass classification" parameters (such as Jn, Jr, etc), etc. These models were used, not only to highlight variations in rock mass properties, but we invaluable tools in optimising slope designs, underground stope design and the estimation of rock reinforcement and ground support requirements. As block models essentially consist of small block shaped cells, with an "x,y,z" centroid, cell size extent (ie cell width, height and length), together with attributes (ie rock mass parameter values), it may therefore be foreseeable that Rocscience may be able to develop tools that allow for the import of such models into it's products (from standard mining software block model formats: eg Datamine, Surpac, Vulcan, Gemcom). For example, a section through the block model could be imported into, say, SLIDE, which would represent the section of the block model as rows and columns of different material properties. Another example, perhaps, Phases2-d could use the imported block model section to modify the properties of the mesh elements?? In the future, it may be possible to develop SLIDE-3D, which imports the full 3-d block model of material properties, together with the proposed 3-d pit shell (as DXF?). The user can then highlight/restrict volumes and areas of the pit surface for analysis.... Now that would be neat! I am currently preparing a paper for Ground Support 2004 (5th Int. Symp. on Ground Support in Mining & Underground Construction, September 2004, Perth, WA) that touches on some of these issues. I hope that these comments may be of some use to you at Rocscience. Cheers Peter Peter Cepuritis Consultant Geotechnical Engineer and Manager PCepuritis@kal.snowdenau.com Dear Sirs, My name is Eduardo Medina and I am the Senior Geotechnical Engineer for Minera Barrick Misquichilca S.A., which operates in Huaraz - Peru. I have read your article about Geostatistics and found it very interesting to me. I am also interested to start learning its concepts and tools that will permit myself using the powerful of the methodology to improve my analysis techniques because I understand the importance of this methodology when applied to the mining industry and guess what should be the benefit when applied to Geotechnical. I will appreciate very much if Roc Science could keep myself informed about the progress with this important Project that will improve the analysis techniques of the interested personnel into the geotechnical community. Also if you are planning to provide courses dealing with this matter and/or to develop software that be necessary for the geotechnical applications I would like to be informed about it, With my best regards, Eduardo Medina Senior Geotechnical Engineer Minera Barrick Misquichilca S. A. 01-217-0140 Extension: 3315 emedina@barrick.com Thanks for the article in RocNews Spring 2003. I think that probabilistic analysis is becoming more acceptable especially for the simple issues we have to deal with that lend themselves to statistics - wedges and other limit equilib calcs. For some of the large slopes on my current project I have done the standard deterministic calcs and then checked them against a prob approach to see the possibility of failure. As always, the database of design parameters is either very small or completely absent. We are left making correlations from quantitative estimations and using published formulas to get design values. The prob approach included in Swedge and RocPlane are very user friendly and "empowering". Although trying to explain the output to engineers is a bit difficult!! Factors of safety still rule supreme. However, how widely useable geostatistics is I am not sure. The spatial distribution of properties is so geologically controlled that one has to split the problem into geological domains first. Otherwise you will be comparing apples and pears or drawing wrong conclusions. In a geologically very simple environment (like the Channel Tunnel where I was a geologist), geostats to determine bedding levels may have been useful, but the primary control of major geological structure had to be understood first and the spatial info superimposed on that. For me, even though I am not a rock-tapper, I remain unconvinced about geostats but firmly in favour of probabilistic tools. One has to be careful not to try and feed the geostats thirst for data at the expense of well founded geological analysis. In the situations where gathering data is very expensive and the geology can be homogenised maybe geostats can help. Thanks again for your excellent software and your obvious interest in serving the profession, Ralph JH PARKIN Resident Excavation and Construction Engineer Chartered Engineer and Chartered Geologist BSc, MSc, MIMM, CEng, FGS, CGeol Stucky Consulting Engineers Ltd Deriner Dam and HEPP ERG 3 Kampi Deriner Baraji TR-08000 Artvin Turkey email: rparkin@stucky.ch Dear Rocscience, I would first like to commend Rocscience on the very informative Geostatistics article in the Spring newsletter. This is a subject I have a very keen interest on and I am busy completing my PhD on the development an application of a 3D geotechnical model. I have attached a recent paper which summarises some of the research to date (published in the 2002 IAEG conference held in Durban). As my work progresses I would be happy to share the learning's with Rocscience in support of your initiative to develop this field. Regards Alan Bye abye@ampnet.co.za As seen from the three sample responses above, geotechnical engineers are most certainly keeping an eye on developments in Geostatistics. We thank all respondents for their contributions. Frankly, the level of interest and awareness of the topic exhibited by readers amazed us outright. The responses sent in reflected a rich diversity of opinions on the possible benefits of the science to geotechnical engineering. These have helped confirm three important facts that will shape all future Rocscience developments on Geostatistics:
geotechnical engineering only with the creation of simple, easy-to-use computational tools 2. Teaching of the basic concepts and applications of Geostatistics in geotechnical engineering education is critical in helping accelerate the adoption and exploitation of the science's capabilities, and 3. Geostatistical tools, like practically all other engineering approaches, will bring the greatest benefit only in the hands of skillful practitioners. Dr. Reginald Hammah Rocscience Inc. rocnews@rocscience.com Again, RocNews thanks Peter Cepuritis, Eduardo Medina, Ralph J.H. Parkin, Alan Bye and Dr. Reginald Hammah for their contributions. |