WITHER HILLS RESERVOIR UPGRADE

FLETCHER BRUCE1, SIMON EDMONDS2, DOUG STIRRAT3

1. Structural Engineer, Beca Ltd. 2. Technical Director - Structural Engineering, Beca Ltd.

3. Technical Director - Civil Engineering, Beca Ltd.

SUMMARY The Wither Hills (Weld Street) Reservoir is an approximately 50 year old concrete reservoir which stores potable water for the town of Blenheim, see Figure 1 for a view of the reservoir above the town. In 2018 the reservoir underwent an upgrade with CH2M Beca as the designer and Fulton Hogan as the contractor. Marlborough District Council required outcomes were: unchlorinated potable water storage safeguarded from contamination, enhanced durability and seismic resilience for a further 50 year working life. These were all achieved via upgrades and additions to the existing concrete structure as much as possible resulting in what could be considered an example of a more sustainable solution than demolition and replacement. The construction had many challenges such as a tight programme, differing as built details and existing and new concrete issues. These challenges were overcome by the established track record from previous projects of collaboration and innovative problem solving between Marlborough District Council, CH2M Beca and Fulton Hogan. The upgrade was completed at the end of 2018.

Figure 1 – Wither Hills (Weld Street) Reservoir looking North to Blenheim before upgrade.

BACKGROUND The Wither Hills (Weld Street) Reservoir stores 5.3 ML of potable water for Blenheim. The reservoir is situated on a platform cut into the lower slopes of the Wither Hills to the south of Blenheim. The reservoir was designed by Steven and Fitzmaurice consulting engineers and constructed circa 1970. It is a prestressed concrete circular water retaining structure with an internal diameter of 36m and water depth of 5.2m. The floor slab on grade is reinforced concrete with circumferential and radial joints. The walls are conventional precast concrete, vertically prestressed and horizontally post tensioned with narrow in-situ stitch joints. Six pilasters around the reservoir provide anchorage for the post tensioning. A ring of six columns near the centre of the reservoir support the radially configured roof structure. The design is considered leading edge of the era due to the long span radial roof system made up of precast, prestressed tee beams with narrow in-situ stitch joints and no topping slab. In the 1990’s seismic strengthening was carried out on the reservoir including anchor bolts through the roof tee beam flanges into the wall and steel brackets glued to the base of the wall and bolted to the foundations to enhance the shear transfer connections at roof and base level. Bearing pads were also installed under the roof tee beams replacing the existing grout bedding. UPGRADE DESIGN In 2018 the reservoir was approximately 50 years old and starting to suffer from durability issues. Rainwater was suspected to be entering the reservoir through the roof and bringing in contaminants requiring the reservoir to be taken offline and disinfected several times. The cement paste on the internal wall face and columns had softened and dissolved leaving an exposed aggregate finish that could potentially accumulate pathogens. The internal pipework was also nearing the end of its life due to corrosion of the ductile iron and connections. Marlborough District Council also wanted to improve the seismic resilience of the reservoir. Following AS/NZS1170.0 - Structural Design Actions, Part 0: General Principles, the reservoir was assessed considering a 50 year residual design life with an Importance Level of 4 (IL4) considering the stored water is to be available after an earthquake and for firefighting. The seismic return periods considered based on these parameters were a Serviceability Limit State 2 (SLS2) of 1 in 500 years and an Ultimate Limit State (ULS) of 1 in 2,500 years. The SLS2 is considered to target operational continuity with the reservoir being required to retain water after the event. The ULS targets non collapse or avoidance of sudden, uncontrolled water loss albeit with sustained damage that may allow leakage. Blenheim is near several major active faults (Wairau and Awatere) resulting in a high seismic demand.

The slope behind the reservoir was assessed incorporating learnings from the 2010-2011 Canterbury earthquakes specifically ridge amplification effects (NZTA, 2017). Reprofiling of the slope above the reservoir was recommended with a flatter slope cut and a mid-height bench to minimise impact of slope failure onto the reservoir or falling debris created from an earthquake. The re-profiling design can be visualised in Figure 2.

The existing reservoir roof detailing meant it did not have an adequate diaphragm to transfer seismic loads. The flange tips of the precast roof tee beams were connected via hook bars in a narrow in-situ stitch joint between adjacent roof tees. A reinforced concrete topping slab was proposed to provide a roof diaphragm. This had the advantage of also sealing the reservoir contents from rainwater ingress. However, this posed a gravity issue by increasing the mass significantly which the roof tee beams were not originally designed for. External post tensioning was designed for the roof tee beams to compensate for the increased mass and maximise reuse of the existing structure. Staged stressing of the post tensioning was required to avoid overstressing the existing roof tee beams. A new stainless steel bracket at the outer end of the beam was designed to enhance the shear connection between the roof tee beam and the new topping slab while providing a guide for the external post tensioning live end anchorage cast into the roof ring beam. The bracket also provided additional shear capacity to the roof tee adjacent to the wall support. Two post tensioning deviator brackets at approximately third point locations along each roof tee beam generates an applied stress profile along the length of the beam that balances the gravity bending stress profile produced by the new topping slab. The dead end anchorage consists of a steel anchor bracket each side of the beam web and a stress bar in a cored hole through the web which was pretensioned to achieve shear friction to resist the applied load produced from the post tensioning. Refer to Figure 3 for 3D views of the roof beam external post tensioning components.

Figure 2 – 3D model of slope re-profiling (CH2M Beca, 2018).

Figure 3 – Roof external post tensioning shown in 3D. Clockwise from top left: live end anchorage cast into new external roof ring beam, roof tee beam shear bracket with strand guides (outer end), strand deviation bracket (two at third locations), dead end anchorage bracket and anchor (inner end). (Fulton Hogan, 2018).

The roof tee beams webs are supported in cut outs in the top of the wall. Strengthening in the 1990’s included addition of vertical anchor bolts through the beam flanges into the top of the wall for transfer of seismic loads from the roof into the wall. The anchor bolts were assessed as no longer providing adequate shear transfer from the roof to the wall due to the increase in design seismic accelerations. To address thermal movements between the roof and the wall the anchor bolts were removed and a new reinforced concrete roof ring beam was proposed to replace the shear load transfer. The ring beam wraps around the roof perimeter and is reinforced to provide hoop tension capacity to support load transfer of the new diaphragm. Its primary function is to provide a shear key to transfer the seismic inertia of the mass of the roof by the trailing side of the roof ring beam bearing against the top of the wall. An analogy is a ‘jar lid’ on the top of a drinking glass. The reservoir wall is set into a slot in the foundation and floor slab with no direct dowelled connection between these components. The external brackets installed in the 1990’s to strengthen this connection were starting to lose their bond due to UV degradation of the epoxy and thermal expansion cycles of the steel. An internal reinforced concrete floor ring beam was proposed with dowels cored through the walls around the perimeter and connected to the existing floor slab and foundation with reinforcement starters to provide a new base shear transfer mechanism. In a previous assessment in 2010 hydrodynamic seismic sloshing of the water contents was assessed and found to be an issue, and the top water level was lowered to provide freeboard for these waves. This was reassessed in 2017 considering updates to the seismic design standards since the Canterbury 2010-2011 earthquakes with no change required. The existing wall post tensioning was sufficient for Serviceability Limit State 2 seismic requirements (1 in 500 year return period event) but insufficient for the Ultimate Limit State (1 in 2,500 year return period event) in terms of hoop tension capacity. External wall post tensioning was designed to compensate for the shortfall in the bottom third while also

providing additional useful hoop compression to the wall joints for leak mitigation at the vertical construction joints, see Figure 4 for visualisation. The pipework was considered to be vulnerable from breakage during an earthquake putting the contents at risk of draining out. The pipework inside the reservoir also had significant corrosion and was essentially reaching its design life. New stainless steel pipework was designed of a larger diameter to provide additional flow capacity. A new valve chamber was constructed to house new seismic isolation valves to the inlet and outlet and flexible bellows. A sensor measures shaking and over a certain acceleration level triggers the valves to be automatically shut off via a battery powered actuator. This minimises the risk of losing the entire water contents in the case of a pipe breakage. Condition remedial work to the reservoir structure was highlighted as provisional items to be confirmed during construction. This included: spot surface remediation of the internal wall surface and columns, spot repair of concrete holes or spalling in the floor and roof, floor slab sealant replacement and crack injection using polyurethane based sealant of active leaks in the walls. The design was modelled in 3D on Autocad Revit incorporating the existing structure, new structure and pipework. An internal view of the reservoir model is shown in Figure 4 featuring the new inlet pipe which runs across the floor to keep it separate from the outlet pipe.

Figure 4 – Wall external post tensioning shown in 3D (top), (Fulton Hogan, 2018). 3D model showing new inlet pipe inside the reservoir (bottom), (CH2M Beca 2017).

CONSTRUCTION The key target for a successful construction was meeting the programme due to water supply requirements. The start date for the works was agreed to enable the upgrade to be completed before the expected start of peak water demand. The target recommission date was end of October 2018. Due to various unexpected issues which are discussed further below the actual construction took longer than expected. There was a wet start to summer which was fortunate and the reservoir was recommissioned before the end of the year in time before the summer dry period. Before Fulton Hogan started work on site they used virtual reality from CH2M Beca to plan their work (see Figure 5). Fulton Hogan used Sketchup for their design elements of the post tensioning components which was imported into the 3D model for completeness. Sharefile was used as a depository for contract documentation and site records. A 360 degree camera was used during construction to capture general progress of the reservoir and site. Skype for business proved to be a useful tool as both companies had access allowing users to share screen to discuss issues and resolve problems efficiently.

Figure 5 – Fulton Hogan construction team using virtual reality.

An external inspection was carried out before draining the reservoir to highlight existing leaks in the wall, map existing wall cracks and confirm extent of remedial work required. A water drop test was carried out on the reservoir before draining began to determine the existing rate of leakage as a benchmarking exercise. A roof water test was not carried out as it was already suspected that rainwater was contaminating the supply requiring the reservoir to be taken offline for disinfection. A large access hatch was cut into the roof at the beginning of construction to provide ventilation, lighting and access for scissor lifts and other large construction equipment. The reservoir could now be classed by Fulton Hogan as a restricted space instead of confined space due to alternative access/egress routes and additional ventilation. An internal inspection was then carried out to confirm as built details and the extent of remedial work required inside the reservoir.

Scaffolding was installed around the roof perimeter and an access stair from the ground to the reservoir roof. An internal access stair was installed and scissor lifts craned in via the enlarged access hatch to allow access to the underside of the roof tee beams. Reinforcement scanning was carried out of the base of the reservoir walls to confirm location of horizontal post tensioning and vertical prestressing prior to coring holes for the floor ring beam dowels. The ring beam was poured in two stages to allow for installation of replacement pipes and surrounding replacement floor slab. The concrete mix included shrinkage compensating admixture, hydrophilic crystalline admixture and microsilica to improve the durability, water resistant and shrinkage cracking performance of the concrete. Figure 6 – Floor ring beam under construction (left) and completed (right). Figure 6 shows the beam under construction and completed.

Figure 6 – Floor ring beam under construction (left) and completed (right).

The top of the walls were cut out using overlapping cores either side of roof tee beams to achieve the design intent of modifying the roof shear transfer mechanism to the new roof ring beam. The roof ring beam was constructed in four equal segments. See Figure 7 for reinforcement and final construction. The live end anchorages of the roof tee beam external post tensioning were cast into the roof ring beam and is discussed further below.

Figure 7 – Roof ring beam under construction (left) and completed with rainwater scupper and post tensioning precast pocket covers shown (right).

The roof tee beam external post tensioning was constructed in several stages. The level of the bottom normal reinforcement of the roof tee beam was found to be higher than shown on the shop drawings (70mm cover compared to 40mm indicated). This resulted in clashes with the new anchorages into the beam. The dead end anchorage at the inner end of the beams was raised to avoid a clash with the bottom reinforcement as the bars are providing a shear transfer mechanism for new loads. However, a flow on effect from this modification was a clash with some of the existing prestressed strands. The beam was reassessed and external post tensioning redesigned to account for this. Refer to Figure 8 for photos of the components. The dead end anchorage hole was cored with a diamond bit on a specially designed jig due to the limited access between the roof tee beams at the inner end (around 600mm between side of roof tee beam webs). The coring was made more difficult due to introduced requirement to core through the high tensile steel. The time taken to core these holes was longer than originally allowed for and the diamond core bits were dulled much quicker requiring replacement more frequently. The live end anchorage at the outer end of the roof tee beam required coring through the bottom reinforcement for the same reason as noted previously. This was compensated by the new external shear brackets cast into the external roof ring beam. The deviator brackets were hot dip galvanized and once installed coated with epoxy mortar. Heat shrinking of a conduit sleeve was used for the junction between the strand PVC ducts to the steel guide tubes for durability protection of the strands.

Figure 8 – Roof tee beam external post tesnioning. Clockwise from top left: general view, shear bracket (outer end), deviator, dead end anchorage (inner end).

The pocket for the post tensioning anchorage was cast into the concrete using a 3D printed insert to achieve the special shape. Each strand anchorage pocket was unique due to the different strand alignment angles. Inserts were colour coded for recognition during

construction. Precast concrete covers were cast in moulds produced by a 3D printer to create matching geometry for the cast in inserts. Refer to Figure 9 for example photos.

Figure 9 – 3D printed inserts for roof post tensioning anchorage pockets (left) and precast concrete pocket covers made in 3D printed moulds (right).

Ducts were run from the dead end anchorage through the deviator brackets to the live end anchorage. Strands were drawn through as greased and sheathed and barrels and wedges were installed. The strands were stressed from the live end in pairs and stages to avoid overstressing the beams at any point of time. The roof tee beam external post tensioning stressing sequence was interdependent with the topping slab pour sequence. Refer to Figure 10 and the next page for the topping slab pour sequence. The stressing sequence as a percent of final design load (60% of the ultimate tensile strength of the strands) was as follows:

1. Initially 25% for each pair then stressed again to 50% at end of stage 1 prior to topping slab pour 4.

2. After pours 4, 5 and 6 strands were stressed again in pairs first to 75% and then 100% of design load.

Survey was carried out multiple times on the roof tee beam web soffits to monitor deflection of beams throughout the stressing and concrete pour sequencing:

1. Original 2. Stage 1 post tensioning 3. Topping slab pour 4 4. Topping slab pour 5 and 6 5. Stage 2 post tensioning 6. Topping slab pour 7 and 8

Theoretical hog and sag deflections were calculated for each stage of stressing and concrete pours for comparison with the survey. Refer to Table 1 for summary of results. Differences in the estimated and actual deflections were generally accounted for as discussed in the comments.

Table 1 - Roof tee beam deflections at different stages (all dimensions are in millimetres taken at the midspan of the beam, positive hogging, negative sagging).

The existing roof surface undulated due to the pattern of precast roof tee beams and in-situ stitch joints. Some transitions were significant and some were known to be letting rainwater through. The topping slab thickness was required to be strictly controlled to avoid overloading the beams at any point of time but required a minimum thickness for effective diaphragm action and rainwater tightness. The original roof fall was also required to be maintained for rainwater runoff. The initial construction sequence was to pour the slab in four stages, each a quarter of the roof area. However, this was required to be revised due to the tight topping slab thickness tolerance. Fulton Hogan with CH2M Beca developed a revised construction sequence shown in Figure 10:

1,2. Roof Ring Beam 3. Central slab 4. Inner Ring Slab 5,6. Outer Ring Slab 7,8. Middle Ring Slab.

Figure 10 – Plan view of the roof topping slab construction sequence (numbered).

This order allowed a string line to be setup between subsequent pours to follow the existing surface to control the subsequent slab thickness. The topping slab concrete mix included hydrophilic crystalline admixture and microsilica to promote self-healing of potential cracking and enhance the durability. The original central joint between the roof tee beams and the central slab was leaking but required to be maintained for rotation and thermal movement actions between the precast roof beams and the in-situ central columns and slab above. A HDPE sheet custom membrane rearguard to seal the joint was fabricated, welded together onsite, fixed to the original concrete prior to being cast (refer to Figure 11). Typical PVC rearguard water stops were used under the normal construction joints.

Figure 11 – HDPE membrane sheet over the central roof joint fixed in position before topping slab pours.

The roof topping slab was found to have cracked more frequently and with greater crack widths in some pours. A roof water test was carried out and found that some of these cracks were allowing water to leak through into the reservoir. Fulton Hogan used a drone to take detailed aerial photos to map the extent and location of cracks. The roof was wet first and the photo taken while the water was drying which highlighted damp spots and therefore likely problem crack locations. Some concrete pours had more cracks than others, in particular the outer ring had the most cracking. It was found that hairline and thin cracks (0.1mm to 0.2mm) healed over time due to the hydrophilic crystalline admixture or autogenous healing. However, cracks greater than 0.3mm still needed addressing. Crack injection was carried out targeting the worst cracks and then a rerun of the roof water test to confirm performance. An investigation was made into possible causes of the extensive cracking which identifies the following potential factors:

• Staged post tensioning and concrete pours meant the support beams were continuously changing in stress state and opposing magnitudes of deflection.

• The pour dimensions (length to width ratio) were not advantageous for some slab segments due to the staged loading requirements.

• The external post tensioning adds strength to the existing beams but the stiffness remains unchanged meaning the beams are still as flexible as the original design.

• Concrete placement and curing on the roof top was very exposed to the wind and sun.

• The slab was thin at nominally 100mm thick.

• The concrete mix included supplementary cementitious material (microsilica) which may have greater autogenous shrinkage and can exhibit greater plastic shrinkage as they tend not to ‘bleed’. The effect of measures to reduce evaporation from the concrete surface to reduce plastic shrinkage may not have been sufficient under some adverse weather conditions. Additional shrinkage reduction admixture used in later slab pours may have reduced these effects.

Figure 12 – Drone aerial photograph showing damp areas highlighting potential cracks (Fulton Hogan, 2018).

It is likely the cause was due to a combination of the above reasons. Figure 12 shows drone aerial photograph taken by Fulton Hogan. The wall external post tensioning required to pass through the pilaster sections to minimize the strand deviation and local stress effects. Holes were cored through each pilaster away from the existing post tensioning anchorages. Four pairs of strands were installed in the bottom third of the reservoir wall. Two anchorages were staggered 180 degrees from each other around the reservoir for each pair of strands. The anchorage, shown in Figure 13, consists of a steel block with barrels and wedges which was encapsulated in concrete for durability. Friction losses were found to be typically low from stressing the greased and sheathed strand around the curvature of the reservoir wall.

Figure 13 – External wall post tensioning anchorages.

During the internal inspection if was found that the cement paste loss on the wall surfaces was to a greater extent than originally estimated and allowed for. The wall surface was re-finished with an acid resistant repair mortar to smooth the rough surface and minimise risk of pathogens accumulating. This proved to be a significant task considering the large surface area. The columns were fibre wrapped to refinish the surface, provide enhanced confinement for the additional axial load and increase the ductility under lateral loading. See Figure 14 for photos of the wall and column surfaces before and after. The floor slab sealants were also replaced.

Figure 14 – Clockwise from top left: wall surface before showing historic water level at beginning of cement paste loss, wall surface after refinishing, column before showing extent

of cement paste loss, column after with protective outer mortar (fire wrapping and mortar bedding beneath).

A final drop test for the reservoir was carried out using a fixed ruler dipped into the water from the large access hatch. Photos of the water level were taken each day of the test using a ‘go-pro’ camera on the end of an extendable pole. The result showed the water level was dropping approximately 3mm in 7 days. A new concrete reservoir is expected to drop less than 10mm in 7 days (Appendix A of NZS3106, 2009). For existing reservoirs the benchmark drop test is typically used as a performance measure. In this case, the performance exceeds the requirement for a new reservoir and therefore was accepted.

A point cloud scan was taken using a Leica BLK 360 camera inside the reservoir to record the final construction, refer to Figure 15. The point cloud was imported into a 3D model in Revit via Autodesk Recap and compared to the original 3D model created from original design drawings. The models were found to be very close indicating the structure was generally constructed following the original drawings and strengthening design.

Figure 15 – Snips taken from Autodesk Recap displaying the BLK scan photos (top) and the point cloud scan model (bottom), (CH2M Beca, 2019).

CONCLUSION The Wither Hills (Weld Street) Reservoir upgrade designed by CH2M Beca and constructed by Fulton Hogan achieved the required outcomes by Marlborough District Council of unchlorinated potable water storage safeguarded from contamination, enhanced durability and seismic resilience. These were accomplished by upgrades and additions to the existing concrete structure as much as possible resulting in a more sustainable solution than demolition and replacement. The construction had many challenges such as a tight programme, differing as built details and existing and new concrete issues. These challenges were overcome by the established track record of collaboration and innovative problem solving between Marlborough District Council, CH2M Beca and Fulton Hogan. The upgrade was completed at the end of 2018, refer to Figure 16 for a view of the upgraded reservoir.

Figure 16 – Upgraded reservoir.

ACKNOWLEDGEMENTS The authors would like to acknowledge the numerous and continuing contribution by Marlborough District Council and Fulton Hogan in the development and implementation of the innovative approach taken on this project. REFERENCES Brabhaharan, P., Mason, D., Gkeli, E., (2017) Seismic design and performance of high cut slopes. NZ Transport Agency research report 613. 149pp. Standards Australia and New Zealand, AS/NZS 1170.0:2002 – Structural Design Actions, Part 0: General Principles. Standards New Zealand, NZS 1170.5:2004 – Structural Design Actions, Part 5: Earthquake actions – New Zealand with 2012 amendments. Standards New Zealand, NZS 3106-2009 – Design of Concrete Structures for the Storage of Liquids.