I think you may have find the answer in this standard: EN 1926 Natural stone test methods. Determination of uniaxial compressive strength.

http://shop.bsigroup.com/ProductDetail/?pid=000000000030149201

i feel percentage error should be within the range of 5 % to 10 %..

Thank you for your response Prof. Joanna and Prof. Karthigeyan.

Error can be intended also as uncertainty. The related data are linked, for example, with the rock volume  involved  in the studied failure process: small volume requeste careful data, large volume can balance wider data distribution. Other conditions influence the acceptable range of geomechanics parameters, such as the level of stress and the relevance of the rock structure for the safety.

the maximum amount of error is acceptable depends on the type of sample, preparation of sample and also the experimental condition. Normally in statistical results is more significant when error varries from 1%  to 5%.

Thank you Prof Claudio and Prof. Bijay. for your valuable suggestions

It is logical that good model fit have less percentage of error. So my query is what is the maximum percentage of error that is acceptable to comment that this criteria is fitted well with the experimental results.

Is there any reference regarding this which is used for rock mechanics purpose.

Prof. Joanna suggested one reference, which is actually inaccessible in my place

I had seen in a paper that %20 of error is quite acceptable in rock engineering projects (Daido et al., 2003).  But I think there is no standard value for this. You may dive into sources of errors in rock mechanics for this respect, and there is a nice quality paper (Hadjigeorgiou and Harrison, 2012) about his subject which may be useful for your concern.

Daido et al. (2003). An experimental study on the validity of Kaiser effect in stress measurements b yAcoustic Emission Method (AEM). Rock Stress '03, 271-274.

Hadjigeorgiou and Harrison (2012). Uncertainty and sources of error in rock engineering. 12th ISRM Congress. 2063-2067.

Dear Dr Ram

sources of erros should be avoided (unrespect of standard, lack of calibration, misunderstanding in readings, poor sampling). Variability is inherent to rock mechanics, and the rock parameter distribution  assesment is one of our main goals.

What is important is the reliability of the model and of the data: try at least to evaluate how much is the output range due to a certain percentage of input data in your geomechanical or numerical model. Be careful on this, because it can determine the passage from stability to failure.

Some interesting comments on reliability can be found in

EINSTEIN, HERBERT H. Uncertainty in Rock Mechanics and Rock Engineering—Then and Now.

ISRM 2003—Technology roadmap for rock mechanics, South African Institute of Mining and Metallurgy, 2003.

Best Regards

I do not feel that a general percentage of error/uncertainty can be applied to rockmechanics or hydrodynamic parameters. Every physical quantity has his practical characteristics. For instance, a permeability (or conductivity) is expressed in m/s and the variability is a power function. The exponent of the value has physical meaning. for instance: between k = 10e-3 m/s and 0.8x10e-3 m/s there is 20% difference, but has no significant practical significance in most of the cases. For the estimation of rock strength, in the ingeneering applications, a variability of 20-30 % is offen acceptable, sometimes even more; if not, the needs of extensive in situ and laboratory investigations are compulsory and rhe results are statiscically based, with great standard variations, RMS, geometrics or arithmetics mean values, and so on. It costs a lot and in the majority of cases, this illusory precision has sense only in front of a court. Not really for serious engeneering. Rely best on reasonable safety factors.

The alignment of a railway line of an high speed railway do not suffer tolerances of very few miliimeters/km, but this should not be based on rockmechanics or geotechnics. It must be trimmed after the layout of the steel tracks, considering for instance settlements that sure can not be predicted with such an accuracy.

Of course, in most numerical models, you have to introduce values without variability. But again this precision is illusory in most of the cases. It is recommended to try different inputs and to plot sensitivity.

Good luck

In the geomechanical design a variability of the compression resistance parameter of 20-30% is fairly frequent and acceptable. When it has to do with anisotropic rocks, variability can reach up to 50%. Generally, variability tends to be much lower for rocks with high average compression strength values. I recommend using the Bayesian approach to obtain the characteristic value of compression strength of experimental data. Then compare that with the calculated values with different criteria and choose the value of the criterion that best approximates the characteristic value of the Bayesian approach.

I also attach the BS EN 1926: 2006 file that you can not find.

Best regards

Thank you Prof. Serdar Prof. Claudio, Prof. Luca and Prof. Bruno for your valuable suggestions and providing some  of the relevant references.

As there is no such suggested values of error percentage, so I think a model or criteria with 10-20% error with respect to experimental data (for isotropic rock)is acceptable in rock mechanics.

You also have to consider what kind of test you perform. Simpler tests (water content, point load strength) can be repeated more often than sophisticated tests (triaxial tests, hollow cylinder tests etc.). And the more often you repeat a test on the same material the variance of the results will narrow.

Form  the reserach data it has been concluded that +-5%is the most acceptable error in modeling in rock mechanic', 

5 to 10% error is acceptable in this case base on available litrature. It also depend on the relibility of testing and as well as estimation based on relible parameters. 

Considering the safety factor the error should not be more than 10%

Just a practical note to lab rock strength tests.  Make sure that you carefully select data from samples that failed as intact material.  Discard any data from samples failed along existing discontinuities or other planes of weakness.  The latter is particularly important with stratified materials such as gneiss, siltstones, greywackes or any mixed facies samples (e.g. flysch), where the orientation of those planes is of great importance.  After all is almost impossible to model rock behaviour without detailed examining its fabric.  Take a look at

Saroglou H. (2013). "Engineering behaviour of anisotropic and heterogeneous layered rocks", in Global view of Engineering Geology and the Environment - Wu & Qi (eds), Taylor & Francis Group.

Saroglou H, Tsiambaos G. (2006), "Point load strength determination of anisotropic rocks".  EUROCK 2006.

Stewart S. (2007), "Rock mass strength and deformability of unweathered closely jointed New Zealand greywacke". Phd Thesis, University of Canterbury, Christchurch, New Zealand.

Thank you Prof. Christian, Prof. Zahid, Prof. Sajjad and Prof. Georgios for your responses.

Dear Bikash I am not a Professor but a practising engineer. 

Giovanni Bruno

the error for my study (prediction of unconfined compressive strength - 1120 data points) is about 30%

do you have any reference regarding your answer that may help me