Robot assisted deconstruction of multi-layered façade constructions on the example of external thermal insulation composite systems
- 853 Downloads
The past years have seen major changes in society that have modified the requirements of residential architecture. To meet the new standards while preserving existing built resources, the building stock has to be generally reorganized. Resulting changes to buildings’ layout and structure in turn necessitate changes to the building envelope e.g. modification of openings or construction setups. Within this context, we investigate the automated disassembly of multi-layered façade constructions, specifically thermal insulation systems using expanded polystyrene foam panels. Our approach goes beyond efficient demolition, instead proposing automated deconstruction strategies that allow for effective material recycling due to a high degree of varietal material purity or even reuse of panel elements. The paper describes the development of the robot assisted deconstruction process including robotic setup, motion programming as well as initial experiments. Moreover, the described research results show the demand for agile process development to successfully deal with the unstructured environment of (de-)construction sites. Therefore, force controlled robot motion programming was tested to introduce industrial robots to a new working environment overcoming traditional factory based automation procedures.
KeywordsAutomated deconstruction Robot assistance Force control Unstructured environment Recycling Reuse Varietal material purity
Deconstruction and refurbishment projects within the building sector drastically increased in recent years and are expected to further rise in the near future due to various reasons (Lublasser et al. 2016). A great majority of refurbishment projects include (partial) deconstruction of façade elements or at least external finish layers because of a shorter life span in contrast to the buildings sub-structure, layout changes and energy optimization demands among others.
To efficiently handle this rising number of façade refurbishments in high wage countries such as Germany, deconstruction and refurbishment processes have to be generally reorganized and optimized. One approach in respect thereof is construction automation through application of new technologies in the building sector such as robotic assistance for automated building processes. But since robotized automation in the building industry lags behind this of other sectors such as the automotive industry a major research potential is initialized here (Bock 2015).
In this context, the research project “Robotic Façade Disassembly and Refurbishment System (RFDRS)” proposes as one goal the development of robot assisted, automated disassembly processes on the example of the deconstruction of multi-layered façade constructions. Initial development and experiments focus on external thermal insulation composite systems (ETICS) based on expanded polystyrene foam (EPS) insulation panels. Nevertheless, a transferability of the resulting process is expected to support the superior goal of developing safe and controllable deconstruction processes for constructions contaminated by hazardous substances such as HBCD, asbestos or other sediments of heavy duty or chemical industry. In relation to that, the research project not only deals with the demand for future deconstruction techniques but also with intensifying recycling and life-cycle regulations of recent European legislation (European Union 2008). Therefore, the deconstruction process for ETICS is not only developed regarding aspects of automation but equally focuses on harvesting the inbuilt materials with the highest achievable varietal purity. This is necessary to guarantee effective material recycling, separation of hazardous substances and to prepare for element reuse of specific components.
Additionally, the research issue of the project addresses general problems of the implementation of robotic assistance in the building sector and on the building site. Well-known problems here are among others the unstructured environment, high tolerances and the lack of a closed digital data flow in planning and execution (Brell-Cokcan and Lublasser 2016).
The following paper describes the approaches of the research project considering real world scenario observation on the construction site as a basic method for the development of a general process strategy for automated deconstruction of multi-layered façade constructions under aspects of material harvesting in high varietal purity as well as general problems of automation in the building sector. Furthermore, details on the robotic setup and robot programming as well as test results of the prototypical process application are provided and discussed.
2 Process parameters
To plan an adequate process strategy, the construction system of ETICS as well as the real world scenario considering the deconstruction site for automation were analyzed to evaluate essential criteria as a basis for process development, robotic setup, partitioning into sub-steps as well as robot motion programming demands.
2.1 General information on ETICS
The application of EPS based ETICS drastically increased since 1990 (Albrecht et al. 2015). ETICS is mostly applied to new and existing building facades in the course of energy-oriented refurbishment processes to reduce internal thermal loss while increase the life span of the buildings sub structure (Deutsche Energie Agentur GmbH). Nevertheless, recycling and refurbishment strategies for the future life cycle of ETICS are not fully developed but are being researched (Albrecht et al. 2015). However, due to the high degree of material compound of ETICS, a material extraction with high purity of variety is so far only expected to be achievable to a minor degree.
This catalog of layers can be extended by an additional joining layer of the insulation material via metal or plastic anchors but is not considered in this research project. The application of all layers is executed in an extensive manner but the adhesive-layer b. This layer can be applied in different ways. In this research project, the bead/spot method was examined.
2.2 System compound and extensive material application
The difficulties of automated deconstruction of ETICS are the strong material compound of the systems layers using adhesives for joining (along the depth of the system) and a lack of visible joining positions or module transitions (along the width of the system). This occurs because of extensively applied finish and hidden adhesive layers. Therefore, the system components provide no defined position for grabbing, partitioning and controlled separation neither along the width nor the depth of the system.
Furthermore, the problem of hidden adhesive layers in combination with an unknown pattern of the bead/spot method complicate a precise and continuous tool motion programming especially influencing the insulation deconstruction.
In manual deconstruction of ETICS, this problem can be circumvented because of the human sensitivity to feel minor differences in the materials varying resistance of adjacent layers. Accordingly, human workers are able to insert deconstruction tools into minor gaps between layers or to generate those gaps with only little demolition or mixing of adjacent layers as basis for further processing.
For this purpose, suitable approaches of robot motion programming and process planning have to be evaluated and implemented (see Sect. 4.1).
2.3 Unstructured environment
Additionally, the unstructured environment of the building/deconstruction site globally affects the process planning and motion programming. The core issue here is the deviation of the actual position of the building system components as well as construction site equipment (e.g. scaffolding) of the ideally planned position. The construction sector is known for its high tolerances due to on site manual construction. What is more, the deviation not constantly occurs but greatly varies throughout the total façade surface. Hence, the relative position of robot and system components is undefined if no extra techniques such as camera systems are applied. To avoid complex sensor based environment detection systems such as listed in Pentti et al. (2013) to minimize costs, system error rates and maintenance efforts, agile methods for flexible process planning and programming have to be implemented in this research project to be able to flexibly react on the building situation and varying component positions (see Sect. 4.2).
2.4 Layer based
To be able to deal with the complex system structure including the various layers, we at the beginning of the project decided to develop the robotic deconstruction process separately for the diverse inbuilt layers. This not only decreased complexity but allowed for implementation of most suitable tools as well as adjustable process operation for each layer.
Base coat with reinforcement and surface finish can be regarded as one layer (Fig. 1d, e). Manual deconstruction experiments showed that a separation of adhesive, finish and reinforcement is particularly time consuming because of the high degree of material compound along the fabric structure of the reinforcement. Additionally, partial demolition of the materials occurred frequently. For the reduction of work effort and because of uncertain varietal purity of the materials a separation of these layers was abandoned,
Insulation material (Fig. 1c),
Adhesive (Fig. 1b).
3 Experimental setup
3.1 Construction setup
Substructure (a) = Sand-lime bricks,
Adhesive (b) = weber.therm. 300; thick-film, mineral adhesive and base coat mortar,
Insulation (c) = weber.therm EPS 035, 140 mm,
Base coat (d) = base coat mortar as in (b) and weber.therm 310 Armierungsgewebe; highly tear-resistant glass fibre fabric with alkali-resistant finish,
Finish coat (e) = weber.pas 430; dispersion-based render.
3.2 Robotic setup
As robotic setup for automated deconstruction in our robot laboratory a robot tandem was used with the robotic arm KUKA iiwa in combination with the mobile platform KUKA KMR to develop the general deconstruction strategy. Standard industrial robot equipment was considered to use a robust well-known environment. Proven hardware allows for concentration on the process planning and motion programming and is our favored method to bring such processes to the future construction site near-term.
This robotic setup is used for prototypical process development in small scale. For actual on site application, the robot arm size and large scale kinematic have to be revised and probably generally redesigned in consideration of the harsh environment. Nonetheless, the robot tandem KUKA KMR iiwa was chosen from the variety of available robots because it provides two promising features for dealing with unstructured environments and on site processes.
First of all, the mobile aspect of the KUKA KMR allows for process fragmentation into sub-processes. These sub-processes can then be separately optimized to feature either specialized small-scale material demands or large-scale organizational demands for element handling, task sequencing and site logistics. Furthermore, the KUKA iiwa is ex-factory equipped with torque sensors at every robot axis. This provides a starting point for implementing aspects of human sensitivity into robot motion programming.
In the project, two software tools were used. For robot motion simulation and basic motion programming KUKA|prc was used. Complex motion programming including implementation of force control via torque sensors was handled through KUKA Sunrise Workbench.
For the selection of adequate deconstruction tools for utilization as robotic end-effector the same strategy as for the robotic setup was pursued—using well-known existing tools to focus on detailed process planning instead of hardware development. As basis for specific tool choice, a number of deconstruction tools were tested manually to analyze effective motion sequences and general handling features. Thereby, the essential characteristics for motion programming were estimated. Furthermore, the effort for tool adjustments to suit robot flange attachment were evaluated. Additionally, the tendency of the tools for particle emission, tool size, universal utilization for various layers as well as possible flat tool alignment along the layers surface were taken into account for final selection.
Layer based tool choice
Multitool with radial diamond saw blade
Multitool with trowel
Insulation partitioning and removal
Prototypical hot wire cutter
Multitool with trowel
The hot wire cutter was designed prototypically to suit the pursed way of deconstruction progress and motion programming. The difficulty here was to find a hot wire cutter, that is capable of steadily cutting along the depth and the width of the ETICS setup at the same time to reduce processing time and to ensure reachability of necessary robot positions. The hotwire cutting process proves, that further research has to be pursued due to a lack of robustness and flexible application options in the existing hardware.
4 Process strategy
The following section describes the implementation of the evaluated process parameters for a general process strategy of robot assisted deconstruction of ETICS layers.
4.1 Dealing with strong material compound
4.2 Force controlled programming
To program an automated tool motion to guarantee the required flat alignment while dealing with the unstructured environment including unspecific positions of construction layer positions, force controlled robot motion programming using the torque sensors of the KUKA iiwa was implemented. Specifically, the force controlled motion is deployed as starting motion of each layer based deconstruction program and consists of a linear motion of the end effector onto the lower layer’s surface stopped by a force condition in the moment of tool-surface-collision (see Fig. 3 left, collision point referenced with “x”). Further layer based program parts are then depending on this evaluated position of surface contact.
The force controlled start motion obviates the need for detailed determination of construction element locations as well as a fine positioning of the robot in relation to the construction elements. Therefore, the layer based deconstruction programs are most agile and highly adaptable to local and constructional differences. Nonetheless, a rough positioning of the robot for approximate tool alignment as well as for the assurance of reachability of all deconstruction program positions is mandatory.
The force controlled robot motion programming can equally be applied within the layer based deconstruction programs if necessary. Detailed descriptions of the differences of the layer based deconstruction programs are explained in Sect. 5.
4.3 Process sub-steps
General process tasks
Rough orientation of robot tandem in front of test wall
Force controlled start movement to layer surface for definition of locally adaptable working environment and actual starting position of deconstruction tasks
Program execution of layer based deconstruction tasks
4.4 Different use of KMR and iiwa movement
The resulting differentiation of rough positioning and deconstruction task execution led to the assignment of particular tasks for the two robot tandem components KMR and iiwa (see Table 2, right column). This is addressing different, required characteristics of kinematics for optimized execution of small or large scale operations. Both robot components are replaceable with other suitable kinematics (e.g. linear axis instead of KUKA KMR).
In this project, the KUKA iiwa handles all layer based deconstruction operations with need for high accuracy (small scale) containing only one task specialized motion program sequence (e.g. rectangular cutting motion for plaster partitioning). In addition, the iiwa is used for adaptable definition of the specific working position for the given task within the unstructured environment via force controlled programming (e.g. start movement of each layer based program). On the other hand, the mobile KUKA KMR is used for rough positioning with low accuracy (large scale) and position changes for task sequencing (e.g. first plaster partitioning at position one, second plaster partitioning at position 2, and so on).
In accordance to Table 1 in Sect. 3.2.3, the tests were executed layer and task based. Details on the layer based deconstruction program, findings and problems are described. The provided figures of each section depict the required robot motions for each deconstruction program with black arrows. The force controlled motions are highlighted in blue. Positions based on the force controlled motion are presented with “x”.
5.1 Partitioning of finish (A)
5.1.1 Results and problems
Initially, the Multitool with saw blade was favored due to its quick tool change mechanism allowing minimum tool change times between various deconstruction tasks. However, uncontrollable wedging of blade and finish material occurred resulting in deviations of cutting directions and process breakdown. Furthermore, the additional rotation motion unnecessarily extends the processing time. Therefore, also the milling spindle was tested presenting less disadvantages.
In general, the orientation of the reinforcement fabric providing different resistances in horizontal and vertical direction is influencing the fluency of movement. This occasionally leads to a process breakdown. Future process optimization should hence include implementation of material specific features to e.g. adopt tool speed.
Furthermore, a force controlled start motion could not be implemented because of the tools’ fragility and geometry. A precise definition of working environment and starting position can, therefore, not be guaranteed. Nonetheless, the process could be handled because the tool characteristic of long cutting edges allows for handling emerging tolerances.
5.2 Removal of finish (B)
Rough positioning of tool in front of wall,
Force controlled starting movement towards insulation surface until tool touches surface,
Slight rotation of the tool to guarantee a parallel position of trowel and insulation surface,
Downward movement—trowel automatically inserts itself between layers and starts material separation,
Horizontal movement away from wall for final layer separation.
5.2.1 Results and problems
In principle, the developed process was repeatedly executed successfully overcoming the unstructured environment and reliably separating the materials. However, from time to time unintentional movement into insulation material instead of the layers’ separation gap occurred leading to increasing contamination of the finish layer with insulation material. Adjustments of the motion procedure adding force control to the rotation and downward movement provide promising prospects for further optimization.
Furthermore, the size of the gained finish elements strongly correlates with the length of the trowel. An approximate immersion depth of 2/3 of the element’s height was examined to ensure correct material separation currently allowing a maximum element height of only 12 cm. To increase process efficiency different tools or trowel types have to be comparably tested.
5.3 Insulation removal (C)
Rough positioning of tool in front of wall,
Movement towards insulation/finish surface until tool touches surface, movement stops due to force control,
(manual insertion of hot wire cutter into slit within finish layer),
Switching on of hot wire cutter,
Movement horizontally into insulation material with pre-defined distance,
Movement of last couple centimeters force controlled until tool touches bricks/adhesive,
Vertical movement to separate insulation from adhesive/brick with pre-defined length,
Horizontal movement outside of insulation (and finish layer).
Initial experiments showed that the hot wire cutter with the given electrical settings for safe operation is not able to move through insulation material as quick as the programmed robot motion. This mainly influenced the vertical movement. Therefore, the movement was split up into smaller movements with either a waiting command or a tool rotation command in-between. Both interruptions of the movement generated a successful and controllable material separation.
5.3.1 Results and problems
Depending on the hot wire size and form, arbitrarily sized insulation material deconstruction was generally successful. However, due to the inhomogeneous material and layer positions a fluent process flow could not be guaranteed mainly because of the following issue.
During construction of the insulation panels, an infiltration of the minor joining gaps between panels with adhesive cannot be prevented. Resulting adhesive nozzles and smears disturb a fluent hot wire cutter movement and lead to either a jammed tool or greater movement delay. This similarly occurs for horizontal as well as vertical gaps. In addition, similar problems emerge due to the use of the bead/spot method instead of full-surface application of the adhesive layer. In both cases, the location of adhesive spots or nozzles cannot be precisely predicted. In respect thereof, an alternative motion programming has to be implemented. Various experiments introducing force controlled motion programming to achieve a bypassing of the adhesive were not convincing. Further adjustments require a preliminary fine adjustment of different force parameters for all possible inconveniences. This has to be elaborated with the help of additional experiments.
5.4 Adhesive removal (D)
5.4.1 Results and problems
The cohesion of adhesive and sand-lime bricks is stronger than this of insulation and finish layer. Due to high emerging forces, the iiwa robot with its 10 kg of payload is not capable to execute the programmed task properly but comes to a standstill in-between both layers. Furthermore, an identification of uncovered wall surface areas within the complex adhesive pattern for correct tool alignment during the procedure is additionally required for fully automated process execution and task sequencing.
6 Varietal purity of harvested material
The partitioned elements of the finish layer are obtained with minor pollution of EPS particles in areas where there was direct tool contact (Fig. 8c). Comparably, the front surface of the insulation panels is as well only slightly contaminated with adhesive residues of the finish layer. Having applied a different tool for the back side of the insulation, here, no contamination of the insulation occurs. However, both sides of the insulation panels show major geometric deviations of the originally flat surface texture of the panels displaying traces of the various used tools (Fig. 8d). A reuse of the insulation can, therefore, be recommended solely with subsequent trimming of the panels. Nevertheless, material loss is limited to a few millimetres of the elements’ width. The degree of pollution for both layers is equivalent to this of manually deconstructed elements documented by other researchers such as Graubner and Clanget-Hulin (2013) and Albrecht et al. (2015).
On the contrary, the adhesive layer is extensively covered with a thin layer of insulation material due to the applied hot wire cutter technique (Fig. 8a). Additionally, the contamination greatly varies due to indistinct adhesion application as described in Sect. 5.3.1. This leads to inhomogeneous tool motion and material separation including extra material contamination wherever adhesion nozzles occur (Fig. 8b). Manual chipping off of the insulation panel as executed in Graubner and Clanget-Hulin (2013) produced less contamination but was not transferred to a robot assisted automated process on account of required high load capacities of the robot as well as aspects of layer reachability and universally applicable movement execution within the unstructured environment. Additional experiments testing a wider range of tools, tool movement and deconstruction strategies have to be carried out comparing the resulting material contamination in relation to the deployed process efforts.
7 Conclusion and outlook
The results of the research project “RFDRS” show, that robot assisted automation of deconstruction processes is viable and distinctive layer separation and material harvesting is generally achievable. Moreover, the research developments demonstrate that flexible process procedures such as implemented via force controlled programming enable the automation of new sensitive deconstruction and separation tasks within an unstructured environment in the first place.
This lays out the foundation for the efficient, comprehensive handling of increasing refurbishment and end-of-life scenarios dealing with the growing number of obsolete and out dated building stock. Furthermore, these experiments present first achievements in overcoming traditional demolition approaches, instead guaranteeing a high degree of varietal material purity for effective material recycling. With minor effort in optimizing and upscaling the proposed automated deconstruction strategies a low contamination material harvesting with subsequent element reuse is expected in the near future.
Additional implementation of force controlled programming to overcome problems of uncontrollable tool wedging,
Upscaling tools and process procedures for supply of adequate element sizes for efficient recycling and reuse,
Upscaling robotic process to ensure task operation requiring higher load capacities such as adhesive removal,
Experimental verification of process application for complex constructional setups including façade openings and corner situations,
Validating the process in terms of costs and efficiency.
We thank Saint-Gobain Weber GmbH for sharing expert knowledge on ETICs and for providing the constructional setup for the test scenario.
- Albrecht W, Schwitalla C (2015) Rückbau, Recycling und Verwertung von WDVS: Möglichkeiten der Wiederverwertung von Bestandteilen des WDVS nach dessen Rückbau durch Zuführung in den Produktionskreislauf der Dämmstoffe bzw. Downcycling in die Produktion minderwertiger Güter bis hin zur energetischen Verwertung, Forschungsinitiative ZukunftBau F, vol 2932. Fraunhofer-IRB-Verl, StuttgartGoogle Scholar
- Brell-Cokcan S, Lublasser E (2016) Neue Technologien im Bauprozess als Qualitätssicherung ressourchengerechten Bauens. Bauingenieur 91(7/8):317–321Google Scholar
- Deutsche Energie Agentur GmbH The energy refurbishment top five. http://pressemitteilungen/the-energyrefurbishment-top-five.htmlGoogle Scholar
- EAE European Association for External Thermal Insulation Composite Systems European guideline for the application of ethics. http://www.ea-etics.eu/views/news-press/european-guideline.html
- European Union (2008) Directive 2008/98/ec of the European Parliament and of the council of 19 November 2008 on waste and repealing certain directives (waste framework directive)Google Scholar
- Graubner CA, Clanget-Hulin M (2013) Analyse der Trennbarkeit von Materialschichten hybrider Außenbauteile bei Sanierungs- und Rückbaumaßnahmen: Erstellung einer praxisnahen Datenbank für die Nachhaltigkeitsbeurteilung: Abschlussbericht, Forschungsinitiative Zukunft Bau, vol 2837. Fraunhofer-IRB-Verl, StuttgartGoogle Scholar
- Lublasser E, Iturralde K, Linner T, Brell-Cokcan S, Bock T (2016) Automated refurbishment and end-of-life processes—research approaches in German and Japanese construction. In: Bock T (ed) Proceedings of the CIB*IAARC W119 CIC 2016 Workshop, pp 20–27Google Scholar