Background

In April 2018 the largest insurable loss in construction history occurred with the collapse of the Auxiliary Diversion Tunnel (ADT) of the $4 Billion USD 2400 Megawatt Ituango Hydroelectric Project in Colombia. After 2 years of ongoing various technical and legal investigations many questions continue to await answers from EPM (Empresas Públicas de Medellín), it’s shareholder the city of Medellín, design & construction contractors, and all parties involved with the project.

ADT Design and Construction

The ADT was designed as a single, horseshoe-shaped tunnel of a width and height of 14 m (~190 m²) and a length of about 2.3 km and was constructed by drill and blast methods with an 8 m top heading and a 6 m bench stage. The hydraulic design criteria adopted for the sizing of the original twin diversion tunnels was based on the diversion of an average flow of 700 m³/s and peak flows of 1000 m³/s, which for each of the original diversion tunnels equates to a flow velocities of 1.8 m/s and 2.6 m/s which is accepted practice for tunnels without concrete lining. It is however surprising that a single diversion tunnel of similar size of 190 m² was accepted to replace the original twin diversion tunnels, each of a size of 190 m², thus implying that the ADT was expected to operate on its own with a 50% hydraulic capacity for the final duration of the diversion period for a further one year including two intense periods of rainfall . The average design flow velocity was 5.1 m/s which is more than typically acceptable by industry standards for a tunnel without concrete lining but was considered to be acceptable by the project designers and also not require a concrete invert based on precedent cases and the assumed geotechnical conditions.

The opinions, conclusions, findings & statements in this guest contribution are those of the author and not necessarily those of Finance Colombia.

The portal for the ADT was located approximately 125 m upstream from the portals of the original twin diversion tunnels and adjacent to topographic depression in the valley slope of the right bank as shown in Figure 1 and unlike the downstream area of sub-vertical cliffs of shallow bedrock. The slope rises gently from the portal of the ADT to provide a cover over the ADT of about 100 m after the curve.

The project site is located within a geo-compressional district of northwestern Colombia associated with numerous north-south oriented major geological faults as well as east-west minor faults resulting in an area of stress relief with highly disturbed fractured rock conditions with a total of 7 fracture sets and of overall high permeability.

The encountered rock conditions along the ADT comprised moderately to highly fractured gneissic and muscovite schist bedrock containing continuous sub-horizontal foliation and frequent parallel shear zones and the rock quality was generally described as varying from fair to good. The distribution of tunnel support installed in the ADT mainly comprised Class 3 with some Class 2 and required the installation of spiling, that was not part of the ground support design, to attempt to control the tunnel profile.

Tunnel support for the ADT was based on the application of the Q-System of the Norwegian Geotechnical Institute with Classes 1 to 3 of support varying from spot requirements of rock bolts and shotcrete to pattern resin end anchored rock bolts of 6 m length of varying spacings in conjunction with a fixed thickness of shotcrete of 10 cm. Class 4 comprised HEB 160 steel ribs with shotcrete of 10 cm. Excavation of the ADT was carried out with variable blast rounds from 4-6 m and included the installation of spiling due to significant overbreak that had occurred that increased the overall profile with heights of overbreak that were estimated to be over 7 m and in some locations not measurable. The amount of overbreak that occurred is not surprising for the prevailing highly fractured geotechnical conditions and since the top bench was excavated with a height of 8 m in to provide adequate space in order to install the required 6 m long rock bolts. In order to have limited the amount of overbreak it would have been necessary to limit the size of the excavation stages and for example require the use of lateral drifts and this would have had a significant impact on progress.

Figure 2 presents the highly irregular profile and very large overbreak along a section of the ADT. The survey records confirm extensive overbreak some of which could not be measured.

Geotechnical mapping was carried out by both the Contractor and the designer and was in general agreement for the conservative installation of tunnel support. The selected tunnel support was to be agreed by both. It does not appear that a Geotechnical Baseline Report (GBR) was prepared as part of the contract as per good industry practice by International Tunnel Insurance Group to serve for bidding purposes as well as a dispute resolution function.

Figure 1 – Portal Location of Auxiliary Diversion Tunnel

Figure 2 – Very Large Overbreak within the ADT

Tunnel support Class 3 was installed in the area before, within, and after the collapse that included a total of eight, 6 m long rock bolts at spacings of 1.5 m together with two layers of 5 cm of shotcrete with a layer of mesh between. The findings of the root cause reports present that the tunnel support design was less than recommendations in the industry whereby only 75% of the rock bolts were installed and also that the mesh layer was not installed. In addition, the installation of the rock bolts was based on the use of cement grout that required a longer time to be effective versus the use resin and therefore allowed relaxation of the tunnel thereby reducing its stability during construction. It is also noted that no concrete invert was placed in the ADT in comparison to the original diversion tunnels.

The ADT operated successfully for about 9 months from August 2017 until late April 2018 and was designed in accordance with the “art” of engineering as stated by the project designer in response to recent criticisms from the national press.

Collapse of the ADT

Normal flows occurred in the Cauca River during late 2017 and early 2018. However, heavy rainfall occurred during April 2018 causing the reservoir level to increase steadily and submerge the ADT and cause pressure flow and high velocity conditions in the ADT. Actual flows during April 2018 varied from 1400 m³/s and increased to a maximum of 2800 m³/s with a second peak reaching up to 2400 m³/s for multiple days equating to flow velocities in the ADT varying from 7 m/s to almost 15 m/s during a 15 day period. In late April 2018 the flow measured through the ADT suddenly decreased followed by the formation of a very large circular crater shaped hole along the slope immediately above the portal with an estimated location of the collapse of about 150 m from the portal. The estimated volume of the crater shaped hole was 120,000 m³ extending over a lateral distance of about 100 m at surface. The crater shaped hole that formed revealed highly altered and discoloured poor quality rock of soil-like conditions extending to great depth of possibly 50 m as shown in Figure 3. With a maximum amount of overbreak of about 7 m noted in the ADT, the minimum cover of rock over the ADT after the curve is about less than 70 m. The circular crater shaped collapse is noted to be consistent and similar to the recognized failure of crown pillars in underground mines that are typically associated as failures through a highly fractured rock mass and not associated with discrete geological faults or shear zones. The height of the collapse is also similar to other examples such as occurred in the 16 km headrace tunnel at the Kemano Hydroelectric Station in Canada in the late 1950’s where there was a circular cavity with a height of 55 m.

Figure 3 – Collapse Crater at Surface of ADT

In order to prevent overtopping of the nearly completed dam the decision was made to open four of the power intake tunnels and bypass the flood flows through the nearly completed powerhouse which resulted in significant erosion along the floor of the unprotected tunnel inverts as well as significant damage to the electro-mechanical equipment.

Body photos & diagrams have been provided by the guest author. Headline photos courtesy EPM.

Root Cause Analyses Reports

Two separate reports have been completed as part of the investigations of the collapse as per typical industry practice and have been referred to as “root cause analyses”.

Both of the root cause reports appear to have focused on identified technical factors contributing to or responsible for the tunnel collapse rather than highlighting the fundamental reason for the collapse which is deemed to have been the high flood flows that occurred during the diversion period and the inadequate hydraulic (capacity) design of the ADT. It is noted that the identification of these technical factors was based on indirect evaluations since no inspection of the ADT has been possible yet, but valid information has been considered including drilling and blasting overbreak, rock quality records, installed support, and hydraulic conditions.

The main conclusion of both reports is that the root cause of the collapse of the ADT was due to erosion of a discrete shear/weakness zone which is considered to be speculative for the following reasons:

• There does not appear to be any direct and convincing evidence of a discrete shear zone presented in the reports at or near the location of the collapse,
• There does not appear to be any direct and convincing evidence of a discrete shear zone presented in the Victims Report (see endnote 3);
• The Independent Advisory Panel also concludes the absence of any weak conditions in the ADT at the collapse location (see endnote 4);
• There is no direct evidence at the site of erosion in the ADT since there is no access.
• The possibility of erosion is inconsistent with the findings of 2 internationally recognized methods of empirical erosion analyses.
• Extremely large overbreak occurred along the roof area of the ADT over a 6 m section from 0+542 to 0+548 and over an 8 m section from 0+584 to 0+592
• The collapse location is positioned in between the two locations of very large overbreak which represent weak areas for high pressure flow and high velocity conditions to concentrate to destabilize the rock and manifest upwards to form a chimney to surface as shown in Figure 4;
• A similar collapse occurred within the intake shafts of 60,000 m³ without erosion that was a result of overpressure of the highly permeable rock surrounding the unlined shafts.
• The circular crater shape of the collapse is consistent with a failure of a highly fractured rock mass from experience in the mining industry rather than due to a discrete shear zone.
• Recent observations (after completion of these root cause reports) from the intake power as well as discharge tunnels where emergency flows were diverted clearly indicate that erosion was limited to along the floor of the tunnels without any collapses as shown in Figure 5.

There are clear differences in the root cause reports compiled as the report for the insurers infers that erosion of a shear zone at 0+540 was responsible for the collapse whereas the root cause report compiled for the project client infers that is was a shear zone at 0+525.

The root cause report compiled for the insurers appears to include a key contradiction whereby it first concludes that the design was in accordance with industry standards but then states that the tunnel was not constructed to the design in terms of the support requirements for the number of rock bolts, shotcrete thickness, and absence of a concrete floor thus implying indirectly that the Contractor was at fault for the works. This report also implies that the Contractor was responsible for the large overbreak due to excessive blasting. However, it is fundamental industry practice that the project designer and construction supervisor are responsible to strictly enforce the technical specifications regarding the procedures of excavation, the installation and confirmation of all tunnel support and to document all non-compliances and confirm the correction of all defective workmanship.

Figure 4 – Collapse in Between Locations of Large Overbreak

Figure 5 – Significant Erosion along Floor of Power Intake Tunnels without Collapse

Of particular interest is that both reports surprisingly do not present the original geotechnical mapping record of the collapse location near 0+525 and 0+540 but rather the record from 0+600 to 0+700, an area beyond from the collapse, without any explanation. This is considered to be quite strange and not typical for such important investigative reports as the collapse location is the relevant location along with an evaluation of the as-built geotechnical conditions and installed support at the collapse location. It is possible that the exact record was not available but should have been made available for such important reports or it was not permitted to be presented which would imply an omission of critical information for the investigative review.

Overall, both reports do not categorically state that the collapse was the responsibility of any particular party involved however based on standard engineering and construction practices in the industry it is quite obvious that there were multiple and serious shortcomings with regards to the engineering design of the ADT.

Finally, the technical expert board that was involved during the construction of the project recommended to the project client not to build the ADT.

 Alternative Root Cause and Associated Technical Reasons of Collapse

As noted, high floods with flows up to 1900 m³/s occurred during April 2018 which based on a review of annual hydrology data appears to be common for the Cauca River. However, the size of the single diversion tunnel of the ADT only provided for 50% of the diversion capacity of the design flows. As such, the ADT was therefore expected to operate under both high velocity and pressure flow conditions however it was not designed with adequate rock support and an impermeable smooth hydraulic lining to prevent the transfer of the internal pressures into the surrounding rock mass and accommodate the high velocity flows. Given the acceptance of the single diversion tunnel of the ADT by the project team it was therefore necessary to design the tunnel with a concrete lining for the hydraulic operating conditions of high velocity and pressures flow conditions.

The prevailing geotechnical conditions along the ADT of highly fractured and highly permeable conditions are recognized in the industry to be not acceptable for a shotcrete lined tunnel operating under high velocity and pressure flow conditions. In addition, the installation of the correct amount of rock bolts (+25%) or even tunnel support Class 4 comprising HEB 160 steel arch ribs at 1.0 m spacings with 10 cm of shotcrete (incomplete encasement of HEB) as well as a concrete floor also would not have been effective to prevent the collapse. Furthermore, spiling as pre-support is not an effective method to limit overbreak and control stability of such large tunnels for the prevailing highly fractured geotechnical conditions when such large excavation stages are allowed by the design. The ADT operated successfully from September 2017 to April 2018 during the non-flood flow period where there were only very limited durations of pressure flow conditions.

The actual design of the ADT with traditional rock support and a shotcrete lining with a extremely irregular profile with very large overbreak would have resulted in highly turbulent flow conditions and the development/build-up of overpressures within the surrounding rock as much as 35 m above the level of the top of the ADT for over a period of 3 weeks and occurring under relatively low cover near the portal location which caused saturation of the rock, lubrication and loss of integrity along weak fractures with an imbalance between the hydraulic and geotechnical conditions leading to a large instability resulting in the formation of a chimney extending upwards over 100 m to surface.

While the many of the historical hydroelectric tunnel collapses have been associated with specific zones of weakness such as at Kemano, Glendoe, Higuera and Shuakhevi (Brox, 2019)⁵ the geotechnical conditions at the ADT do not appear to clearly  reveal such a discrete main weak zone but rather comprises numerous continuous rock fractures and foliation with interconnectivity and with deep weathering. These geotechnical characteristics represent weak rock mass conditions with high permeability that allow for the transmission of water under pressure from the ADT extending significantly into the rock mass as effective saturation with lubrication of the rock fractures as indicated from the predictions in the first root cause analysis report and overall weakening causing eventual instability. These conditions with the experience of very large overbreak during excavation represent the underlying cause of the collapse.

After review of the available root cause reports together with additional information and reports discovered online the following alternative and additional factors are considered to be the primary and important reasons for the collapse of the ADT:

• Unacceptable portal location and initial tunnel section under low cover of fractured rock.
• Single, undersized tunnel for hydraulic operations of high flood flows (50% capacity);
• Unacceptable tunnel profile with large overbreak for hydraulic operations, and.
• Deficient final tunnel lining for hydraulic operations of high pressures and velocities.

All of the above listed reasons are related to the design of the ADT and none of these reasons were the responsibility of the Contractor as the applicable geotechnical and hydraulic conditions were previously known by the project designer and grossly misunderstood and unacceptable and unprecedented critical design decisions were made by the project designer. In addition, from a review of the information that has been available, there does not appear to be any indications of sub-standard work by the Contractor. A large collapse was therefore to be expected given the ADT design aspects and the actual high flood flows that occurred and should have been correctly mitigated through ongoing risk management practices for the project.

Preliminary Lessons Learned

Following a review of some significant relevant information of the project the collapse of the diversion tunnel can be regarded as an error of grand proportion involving what appears to be almost every major party included in the project. The occurrence of construction quality issues cannot be identified from the information and reports compiled to date and it appears that the Contractor carried out the works in accordance with the project requirements.

The geotechnical conditions of the ADT exposed during excavation with multiple locations of large overbreak (and high permeability to allow overpressures to develop within the rock) should have been immediately recognized by the designers to require a new design solution for enhanced support and lining for the anticipated hydraulic conditions of high flow velocities and high pressures. This omission represents serious negligence by the project designers since it is recognized that the design of the ADT with 50% hydraulic capacity did not follow typical good industry practice but rather represented unprecedented and high-risk practices in the hydroelectric industry. There does not exist an “art” of engineering for the design of hydroelectric tunnels.

The occurrence of the collapse of the ADT is a very unfortunate situation and highlights the importance of both risk management throughout the life of a project as per the recommendations of the International Tunnel Insurers Group. A preliminary key lesson learned is that the design of diversion tunnels that represent a critical component, especially for major hydroelectric projects, must be compatible with and respect the anticipated hydraulic conditions including annual intense precipitation to function safely during the entire diversion period. If adverse geotechnical conditions are realized during construction, then it is necessary to include special procedures for excavation and methods to achieve acceptable stability for the installation of the final lining. The typical industry practice comprises the design of fully concrete lined tunnels with control gates.

The author, Dean Brox, can be contacted directly via Linkedin or email (click either).

End Note References:

1. Adjusters Root Cause Analysis Report for Re-Insurers, Collapse of the Auxiliary Diversion Tunnel, August 2019, Ituango Hydroelectric Project, authors Snee, C., Guilherme de Mello, L., Murphy, B., and Prieto, R.
2. Study of the Root Physical Cause and Complementary Report, February 2019. Ituango Hydroelectric Project. Skava Consulting (In Spanish).
3. Fierro Morales, J., Aponte Rojas, D., and Quintero Chavarría, E. 2019. Analysis of geological, geomorphological, hydrogeological, and geotechnical information related to the rock mass and stability of the Ituango Hydroelectric Project, Technical Advice to the Victims, Ituango Hydroelectric Project (In Spanish) riosvivoscolombia.org/documentos
4. Independent Advisory Panel Report to International Development Bank Invest, Report No. 1, September, 2018.
5. Brox, D. 2019, Hydropower Tunnel Failures – Risks and Causes, International Tunneling Association World Tunnel Congress, Naples, Italy.

In its analysis, Moody’s listed four factors as its key rationale to upgrade EPM’s ratings:

• Moody’s Baa2 rating for the city of Medellín (with a stable outlook)
• the “high level of dependence” between the company and Medellín
• the strong probability of “extraordinary support” from Medellín if EPM faces financial distress
• “EPM’s intrinsic credit profile,” which is based upon Moody’s baseline credit assessment (BCA) of baa3

Moody’s also noted the significance of EPM’s ability to reduce its debt last year after dealing with both a “severe El Niño phenomena” (that caused drought and lowered hydropower production) and a major disruption to its energy plant in Guatape. The company improved financial metrics in the second half of the year were also aided by Colombia’s $2 billion USD sale of Isagen, which EPM had a stake in and benefitted from, as well as a recovery in generation operations.

Between the debt reduction and a belief that EPM is now better prepared to withstand risks posed by the extreme weather cycles, Moody’s felt comfortable upgrading the rating.

“The raising of the baseline credit assessment to baa3, from ba1, reflects management’s commitment to maintain a prudent financial policy as evidenced by the group’s commitment to maintain the consolidated debt to EBITDA below 3.5x on a sustained basis,” stated Moody’s. “This was the case for the last 12 months ending February 2017.”

The upgrade also reflects optimism surrounding EPM’s Ituango hydroelectric plant, which is reportedly about two-thirds complete and is expected to provide some 2,400 megawatts of capacity after it begins operating (with a projected online date sometime in 2018). Moody’s was encouraged by the utility’s announcement that it plans to remain patient in terms of pursuing growth until after the Ituango plant comes online.

“The raising of EPM’s BCA to baa3 captures Moody’s expectation of the successful commission of Ituango first phase’s four units — of 1,200 megawatts — starting end of next year,” stated the rating agency. “The significant progress in the construction of the Ituango plant (over 65%) and the completion of the more complex milestones underpin this anticipation … The stable outlook captures mainly Moody’s expectation for the successful commissioning of Ituango”

Photo credit: Empresas Publicas de Medellín (EPM)

Bank of America and HSBC Securities served as global coordinators, joint bookrunners and mandated lead arrangers. The rest of the syndicate was made up of Bank of Tokyo-Mitsubishi, Sumitomo Mitsui Banking as mandated lead arrangers and joint bookrunners, and Citigroup Global Markets, JP Morgan Securities, and Mizuho Bank as lead arrangers.

The loan was authorized by the Ministerio de Hacienda y Credito (Finance Ministry) on December 22, by degree #4783, and the deal was signed on December 29 by the banks and Juan Esteban Calle Restrepo, general manager of EPM.

83% of the loan will go towards energy, such as the Ituango Hydroelectric Project, and 17% will go towards EPM’s waterworks businesses.

The loan is disbursable over a 12 month period and has a term of 5 years. The interest rate is LIBOR+1.40%. The loan will finance the investment plan and $12.4 billion 2016 budget for EPM, approved by the board of directors on November 24 of last year.

“This operation represents a diversification of the sources of financing for our enterprise, and the consolidation of a long-term commercial relationship with this important group of international banks that back the growth strategy and diversification of projects and businesses of Grupo EPM,” said Calle.

“The competitive conditions practiced in this operation ratify the trust and the solid reputation that EPM projects in the international financial environment, even amidst an economic landscape of high volatility in international markets.”

Over the next four years, EPM plans to make $2.4 billion USD in investments to develop expansion projects, modernization and growth in the energy, gas, and water sectors, along with social responsibility and environmental programs that will contribute to sustainable growth of the enterprise and the areas in which it operates.

Photos of loan closing courtesy of EPM