Dr. Gennaro Senatore - Swiss Federal Institute of Technology (EPFL), School of Architecture, Civil and Environmental Engineering (ENAC), Applied Computing and Mechanics Laboratory (IMAC)
Dr. Philippe Duffour - University College London
Dr. Pete Winslow - Expedition Engineering
Prof. Chris Wise - Expedition Engineering
Context and Motivation
Designing strength into buildings to prevent collapse has been the underlying tenet of structural engineering since before the ancient Egyptians built the first pyramids. A second fundamental principle, which appears in all modern design codes, is to build structures that are stiff enough to prevent excessive movement and deformation under statistically calculated worst-case loads such as high winds, heavy snow, and large crowds. This is not really a question of safety, rather it is about the usability and serviceability of buildings.
However, when it comes to movement, how much is too much? How often do these expected worst-case loads really happen? For example, football stadiums may only have to cope with a full crowd 20 times per year for 90 minutes on a Saturday afternoon. Most of the time such structures experience loads significantly lower than the design load, which means they are effectively over-designed for most of their working life. Could building structures be adaptive rather than relying only on passive load-bearing capacity?
The Adaptive Structures Design Method
Active control has been used in engineering structures for a variety of purposes. The most widespread application in civil engineering has been in vibration control. A few have investigated the potential of using adaptation to save material, but whether the energy saved by using less material makes up for the energy consumed through control and actuation has so far received little attention. Gennaro Senatore formulated a new methodology to design adaptive structures during his doctorate, which was a research project collaboration between the University College London and Expedition Engineering.
Figure 1 illustrates this method diagrammatically. It shows the total energy of the structure as a function of some notional degree of active control of the structure. The total energy is made of two components: embodied energy and operational energy. Embodied energy refers to the energy consumed by all of the processes associated with the production of a building, from the mining and processing of natural resources to manufacturing.
For a completely passive design, the embodied energy dominates the total energy: members are designed to bear 100 percent of the design loads to meet strength and serviceability requirements. By contrast, for a highly active design, the embodied energy may be small, but the lifetime operational energy necessary to control and actuate the structure is high. Therefore, an active-passive system that corresponds to the minimum total energy is the optimum solution.
In this context, an adaptive structure should be designed to passively withstand ordinary loading, whilst relying on both passive resistance and active control to deal with rare, unpredictable events (for example, wind storms, snow, earthquakes, unusual crowds, and moving loads such as trains). Instead of increasing the stiffness using material mass, strategically located active elements (actuators) manipulate the internal flow of forces and change the shape of the structure. In this way, we can homogenise stress and keep deflections within desired limits. However, the actuators only activate during rare occasions when external loading reaches a certain threshold, so the operational energy is only used when necessary.
As an analogy to the natural world, consider an adaptive structure like a human arm. If one holds an empty glass in their hand, then fills it with water, the hand does not move. As the weight of water increases, the brain senses this and, through the nervous system, it tells the arm muscles to compensate thereby holding the glass in position. However, unlike the human arm, which expends energy simply holding its own weight, adaptive structures only need to activate in rare loading scenarios.
Extensive numerical simulations, which compare modern passive structures with the equivalent adaptive solution, show that we could reduce the total energy (embodied + operational for a 50-year lifespan) by up to 70 percent for slender structures.
We present two examples here. Figure 2 shows a structure that represents a section of a typical hangar building made of planar trussed portal frames. Figure 3 and Figure 4 show a structure that is a simplified model of a London skyscraper informally known as the Gherkin. This building is assumed to carry external loads only using its own exoskeleton as structural system (that is, no cores).
We identified an optimal region in which adaptive structures outperform their optimised passive versions_ in terms of energy and monetary cost savings. This is broadly the region of stiffness-governed structures (for example, long-span structures and high-rise buildings).
The Adaptive Truss Prototype
We built a large scale prototype (Figure 5), designed using this new methodology, at the University College London structures laboratory.
The prototype is a 6 m cantilever spatial truss with a 37.5:1 span-to-depth ratio (Figure 6). The truss consists of 45 passive steel members and 10 electric linear actuators strategically fitted within the tension diagonal members. We designed the structure to support its own weight (102 kg including actuators and cladding) and take a live load of 100 kg at the tip of the cantilever (person walking along the deck).
We fully instrumented the frame (Figure 7) to monitor the stress in the passive members, the deflected shape, and the operational energy consumed by the active elements. We sized the truss’s passive steel members to prevent collapse, but instead of adding more material, a state-of-the-art control system governs the more onerous requirements of deflection and movement. Due to the fail-safe nature of the actuators, if the power is cut, the actuators simply stop moving with no compromise of load carrying capacity.
We designed the control system to respond to loading without user intervention or predetermined knowledge of the external load.
The control system (Figure 8) consists of a single controller for acquisition and processing, 10 linear actuators, a control driver board for each pair of actuators, 45 strain gage-based sensors (one for each structural member), two amplifiers for signal conditioning, 280 m of screened signal cables, and 45 m of power cables.
The main controller is a powerful cRIO-9024 real-time, embedded processing and acquisition system that is energy efficient and features 800 MHz and 512 MB DRAM. The real-time processor guarantees deterministic control by operating without buffering delays and minimising latencies caused by software interrupts and thread switching. These important characteristics empower us to control the structure in real-time in the event of sudden changes of external loads.
Aside from real-time processing, the CompactRIO system also integrates an all-programmable FPGA, which we used to provide in-line processing on acquired and generated signals. We used LabVIEW to take full advantage of the many benefits of FPGA (reliability, parallelism, processing speed), without having to deal with the intrinsic complexity of traditional low-level, hardware descriptor languages.
The CompactRIO real-time processor runs the main control routine, which requires input from the strain measurement and the actuator feedback positions (both of which have been preprocessed on the FPGA). We used this data to reconstruct the node spatial positions to compute minimum length actuator changes to bring the structure to the desired shape (in this case flat within ±2 mm accuracy end to end). A proportional-integral-derivative algorithm, implemented on the CompactRIO FPGA then computes the appropriate pulse-width modulation signals to move the actuators to the designed position. A switch on/off command sent to the actuator driver cuts off the power supply whenever the target position is reached.
We used LabVIEW to develop an efficient implementation that kept each full control cycle (read strains, reconstruct truss node displacements, adjust actuator lengths) within 40 ms, with less than 2 ms jitter.
Displacements Control | Infinite Stiffness Structure
Extensive loads tests showed that the displacements were practically reduced to zero, achieving an infinite stiffness structure (zero deflection under loading). We measured the displacements using a probe and an accurate, self-levelling laser. The difference in the vertical position between two consecutive nodes was within ±1 mm, and between the supports and the free end nodes within ± 2 mm. Figure 9 shows an example of the difference between uncontrolled/deformed shape (a) and the controlled shape (b).
Power Consumption | Total Energy Assessment
The objective of this experiment was to gather experimental data to validate the claim that adaptive structures allow significant total energy savings. We modelled the external load using a stochastic distribution, which is representative of typical building loading scenarios (design life of 50 years). The chosen displacement limits match the common serviceability criteria of a high-rise building. Due to pronounced slenderness, the adaptive structure prototype can be regarded as the scaled super structure of a tall tower subjected to wind load.
We recorded the structural adaptation power-consumption for all the electronic devices associated with the prototype and then benchmarked against two traditional passive structures (Figure 10) designed to cope with the same loads, whilst complying with the same serviceability limits used for the adaptive truss. The first structure is made of two steel I-beams. The second is an equivalent truss designed using state-of-the-art optimization methods. The adaptive truss achieves 70 percent energy savings compared to the I-beams and 40 percent compared to the optimised passive truss. This experiment confirmed that the new adaptive structure design methodology is reliable and can achieve substantive total energy savings compared to equivalent passive structures.
Adaptive structures represent a new design philosophy in structural engineering that can produce structures that: 1) have a low overall environmental impact; 2) have displacements that can be controlled within very tight limits; and 3) are extremely slender. These three performance objectives are normally mutually exclusive, so being able to combine all three is a unique benefit in structural engineering.
Scenarios where adaptive structures could bring significant benefits include:
- Stringent, high-performance requirements for deflection, such as laboratory buildings, gantry crane runway beams, and bespoke facades
- Structural designs governed by rare, but high, loads such as earthquakes and wind storms
- Times when very high slenderness/shallow structural depths are needed, which could be driven by space limitations or awe-inspiring aesthetics
- Long-span and high-rise buildings (skyscrapers, bridges, roofs) would benefit from the three main characteristics of adaptive structures (stiffer, lower weight, slender)
We exhibited the prototype at various key institutions, including the International Association for Shell and Spatial Structures (IASS) symposium held in Amsterdam, the London Building Centre, and University College London.
The work described in this article is part of an engineering doctorate (EngD) carried out at the department of Civil Engineering (CEGE) at the University College London. The project was managed and funded under the Engineering and Physical Sciences Research Council (EPSRC) Industrial Doctorate Centre: Urban Sustainability and Resilience. The industrial sponsor was Expedition Engineering. Further research funding was obtained for the prototype development by The Institution of Civil Engineers (ICE) and the Institution of Structural Engineers (istructe).
Dr. Gennaro Senatore
Swiss Federal Institute of Technology (EPFL), School of Architecture, Civil and Environmental Engineering (ENAC), Applied Computing and Mechanics Laboratory (IMAC)
Station 18, GC G1 597, CH-1015