Outliers in a dataset are unavoidable and pose a two-pronged problem: first, how do you define and detect an outlier, and secondly, how do you handle outliers? Rock engineering’s ap-proaches to answering these questions include relying on engineering judgement to determine what data points are outliers and removing these outliers to get a better fit. The “engineering judgement” portion of outlier definition and detection in rock engineering makes reproducibil-ity difficult, as engineering judgement is inherently subjective and often not made transpar-ent. Through the results of a survey, this paper aims to demonstrate that extreme caution should be used when i) employing engineering judgment for outlier detection and removal in rock engineering and ii) interpreting the results of a statistical analysis of data where engineer-ing judgment was used to remove outliers. This paper finds that more transparency is needed when engineering judgement is employed for outlier detection and/or removal.

Rockbolt is an integral support element in reinforcements and stability of Tunnels. It is observed that sometimes in tunnels contractual specifications, Factor of Safety (FoS) of rockbolts is defined in terms of loads with a value ranging from 1.5 to 3, and pullout tests are carried out to ascertain the contractual requirements. The author wishes to point out it is inappropriate to define FoS of rockbolts in terms of a load carrying capacity alone. The above FoS values are applicable in cases of wedge failures or low in-situ stresses (Li, 2017). The bolt installed in the tunnel periphery experiences different strain along its length, and axial forces coming on rock-bolt are unequally distributed. It is suggested that FoS should also consider the strain capacity of rockbolt system. This paper gives a review of interaction and decoupling behavior of rock-bolt-grout-rockmass, and highlights aspects related to rockbolt FoS for tunneling applications.

The aim of this paper is to introduce the concept of fracture index for performance pre-diction of Tunnel Boring Machines (TBM) in excavation of rock tunnels, and to examine the influence of rock fracture intensity and its frequency on machine performance. For this purpose, field data from six tunneling projects excavated by TBM were examined. The database included more than six hundred data points. The fracture index, which rep-resents the fracture count over an arbitrary length of scanline or borehole core with similar intensity of fracturing, provides an insight into the fracturing and joint frequency and state of rock masses. The analysis shows that fracture index can offer a critical quantification of rock mass cuttability, and can heavily impact mechanized rock excavation process, and significantly affect TBM performance.

ICL Súria & Sallent (ICL IBERIA) operates two salt and potash mines, Vilafruns, at the southern block of the Tordell fault and at a moderate depth, and Cabanasses, where depths of 1,000 m are reached, at the north of the aforementioned fault, in the province of Barcelona, Spain. The Cabanasas mine is currently accessed through Shaft 2; a 680-meter-deep shaft refurbished in 2004. In order to expand the mine and to increase its extraction capacity, the company decided to build a 5,023 m transportation ramp (19% inclination) in which a depth of 915 is reached. The section of the ramp allows the installation of the belt and the crossing of transport equipment during its excavation, being 9.4 x 5.5 m, with widths every 1,000 m and six By-passes. The construction of this mine ramp took place between 2012 to 2020. The ramp crosses the characteristic materials of the Catalan Potassic Basin. During the excavation of the ramp, several problems have appeared, derived, first, from the intersection of a calcareous aquifer (PK 0+640), from the Tordell fault (PK 1+150) and at a greater depth, above 600 m overburden, squeezing conditions were encountered. The squeezing conditions turned to be severe developing a creep behaviour. The deformation associated to this behaviour has ranged from 4 (0,5%) to 80 cm (8,5%) with creep velocities reaching 0,5 mm/day. The paper describes the support design methodology used, including creep stress-strain calculation with FLAC3D, developing a sequential application of the support elements in order to generate a yield response of the support was adopted, involving also a final structural shotcrete lining (SCL) usually 48 m behind the excavation face, including a strut slab/invert. A risk heat assessment of the ramp was done, and in the Suprasaline unit it was decided to adopt a temporary support that will have to be repaired periodically. Cabanasas ramp is operating successfully since April 2021.

Sea stacks like the E.T. formation at Hopewell Rocks Provincial Park, New Brunswick, Canada, are one of the world’s most popular geotourism landmarks. Failure of sea stacks poses a risk to public safety; this study uses 3D Finite Element geomechanics numerical models with input from UAV-based photographs, 3D Structure-from-Motion photogrammetry models, and erosion records to predict time of failure of the E.T. formation. A novel aspect of this study is the comparison between simplified symmetric and more realistic asymmetric erosion patterns, and impacts on simulated stability. Using symmetrical erosion, the predicted time of failure is within the years 2071 and 2123, but asymmetric erosion models show significantly earlier failure that may occur between the years 2056 and 2078. Ongoing work aims to address sensitivities of input properties, and possible future accelerated erosion rates due to climate change.

In this paper, a Discrete Fracture Network (DFN) model was created by numer-ical simulations which can be used to quickly and precisely estimate several input parameters necessary for rock mass classification. The paper shows how a large amount of useful information from the DFN model can be incorporated into the process of calculating the RMR value. The concept was tested by using the point clouds of real rock slope, acquired with LiDAR laser scan-ning technology. The fitting, calibrating, and generating of DFN model was performed in special-ized ADFNE MATLAB package. In this work, a rock slope on which conventional RMR classi-fication protocol was performed, is used to compare conventional versus DFN-based RMR classification. The results show how and to what extent the RMR values differed depending on procedure. The calibrated DFN approach was more reliable for scoring specific RMR parameters.