This thesis presents the application of data science techniques, especially machine learning, for the development of seismic damage and loss prediction models for residential buildings. Current post-earthquake building damage evaluation forms are developed for a particular country in mind. The lack of consistency hinders the comparison of building damage between different regions. A new paper form has been developed to address the need for a global universal methodology for post-earthquake building damage assessment. The form was successfully trialled in the street ‘La Morena’ in Mexico City following the 2017 Puebla earthquake. Aside from developing a framework for better input data for performance based earthquake engineering, this project also extended current techniques to derive insights from post-earthquake observations. Machine learning (ML) was applied to seismic damage data of residential buildings in Mexico City following the 2017 Puebla earthquake and in Christchurch following the 2010-2011 Canterbury earthquake sequence (CES). The experience showcased that it is readily possible to develop empirical data only driven models that can successfully identify key damage drivers and hidden underlying correlations without prior engineering knowledge. With adequate maintenance, such models have the potential to be rapidly and easily updated to allow improved damage and loss prediction accuracy and greater ability for models to be generalised. For ML models developed for the key events of the CES, the model trained using data from the 22 February 2011 event generalised the best for loss prediction. This is thought to be because of the large number of instances available for this event and the relatively limited class imbalance between the categories of the target attribute. For the CES, ML highlighted the importance of peak ground acceleration (PGA), building age, building size, liquefaction occurrence, and soil conditions as main factors which affected the losses in residential buildings in Christchurch. ML also highlighted the influence of liquefaction on the buildings losses related to the 22 February 2011 event. Further to the ML model development, the application of post-hoc methodologies was shown to be an effective way to derive insights for ML algorithms that are not intrinsically interpretable. Overall, these provide a basis for the development of ‘greybox’ ML models.
This study analyses the Earthquake Commission’s (EQC) insurance claims database to investigate the influence of seismic intensity and property damage resulting from the Canterbury Earthquake Sequence (CES) on the repair costs and claim settlement duration for residential buildings. Firstly, the ratio of building repair cost to its replacement cost was expressed as a Building Loss Ratio (BLR), which was further extended to Regional Loss Ratio (RLR) for greater Christchurch by multiplying the average of all building loss ratios with the proportion of building stock that lodged an insurance claim. Secondly, the total time required to settle the claim and the time taken to complete each phase of the claim settlement process were obtained. Based on the database, the regional loss ratio for greater Christchurch for three events producing shakings of intensities 6, 7, and 8 on the modified Mercalli intensity scale were 0.013, 0.066, and 0.171, respectively. Furthermore, small (less than NZD15,000), medium (between NZD15,000 and NZD100,000), and large (more than NZD100,000) claims took 0.35-0.55, 1.95-2.45, and 3.35-3.85 years to settle regardless of the building’s construction period and earthquake intensities. The number of claims was also disaggregated by various building characteristics to evaluate their relative contribution to the damage and repair costs.
In this thesis, focus is given to develop methodologies for rapidly estimating specific components of loss and downtime functions. The thesis proposes methodologies for deriving loss functions by (i) considering individual component performance; (ii) grouping them as per their performance characteristics; and (iii) applying them to similar building usage categories. The degree of variation in building stock and understanding their characteristics are important factors to be considered in the loss estimation methodology and the field surveys carried out to collect data add value to the study. To facilitate developing ‘downtime’ functions, this study investigates two key components of downtime: (i) time delay from post-event damage assessment of properties; and (ii) time delay in settling the insurance claims lodged. In these two areas, this research enables understanding of critical factors that influence certain aspects of downtime and suggests approaches to quantify those factors. By scrutinising the residential damage insurance claims data provided by the Earthquake Commission (EQC) for the 2010- 2011 Canterbury Earthquake Sequence (CES), this work provides insights into various processes of claims settlement, the time taken to complete them and the EQC loss contributions to building stock in Christchurch city and Canterbury region. The study has shown diligence in investigating the EQC insurance claim data obtained from the CES to get new insights and build confidence in the models developed and the results generated. The first stage of this research develops contribution functions (probabilistic relationships between the expected losses for a wide range of building components and the building’s maximum response) for common types of claddings used in New Zealand buildings combining the probabilistic density functions (developed using the quantity of claddings measured from Christchurch buildings), fragility functions (obtained from the published literature) and cost functions (developed based on inputs from builders) through Monte Carlo simulations. From the developed contribution functions, glazing, masonry veneer, monolithic and precast concrete cladding systems are found to incur 50% loss at inter-storey drift levels equal to 0.027, 0.003, 0.005 and 0.011, respectively. Further, the maximum expected cladding loss for glazing, masonry veneer, monolithic, precast concrete cladding systems are found to be 368.2, 331.9, 365.0, and 136.2 NZD per square meter of floor area, respectively. In the second stage of this research, a detailed cost breakdown of typical buildings designed and built for different purposes is conducted. The contributions of structural and non- structural components to the total building cost are compared for buildings of different usages, and based on the similar ratios of non-structural performance group costs to the structural performance group cost, four-building groups are identified; (i) Structural components dominant group: outdoor sports, stadiums, parkings and long-span warehouses, (ii) non- structural drift-sensitive components dominant group: houses, single-storey suburban buildings (all usages), theatres/halls, workshops and clubhouses, (iii) non-structural acceleration- sensitive components dominant group: hospitals, research labs, museums and retail/cold stores, and (iv) apartments, hotels, offices, industrials, indoor sports, classrooms, devotionals and aquariums. By statistically analysing the cost breakdowns, performance group weighting factors are proposed for structural, and acceleration-sensitive and drift-sensitive non-structural components for all four building groups. Thus proposed building usage groupings and corresponding weighting factors facilitate rapid seismic loss estimation of any type of building given the EDPs at storey levels are known. A model for the quantification of post-earthquake inspection duration is developed in the third stage of this research. Herein, phase durations for the three assessment phases (one rapid impact and two rapid building) are computed using the number of buildings needing inspections, the number of engineers involved in inspections and a phase duration coefficient (which considers the median building inspection time, efficiency of engineer and the number of engineers involved in each assessment teams). The proposed model can be used: (i) by national/regional authorities to decide the length of the emergency period following a major earthquake, and estimate the number of engineers required to conduct a post-earthquake inspection within the desired emergency period, and (ii) to quantify the delay due to inspection for the downtime modelling framework. The final stage of this research investigates the repair costs and insurance claim settlement time for damaged residential buildings in the 2010-2011 Canterbury earthquake sequence. Based on the EQC claim settlement process, claims are categorized into three groups; (i) Small Claims: claims less than NZD15,000 which were settled through cash payment, (ii) Medium Claims: claims less than NZD100,000 which were managed through Canterbury Home Repair Programme (CHRP), and (iii) Large Claims: claims above NZD100,000 which were managed by an insurance provider. The regional loss ratio (RLR) for greater Christchurch for three events inducing shakings of approximate seismic intensities 6, 7, and 8 are found to be 0.013, 0.066, and 0.171, respectively. Furthermore, the claim duration (time between an event and the claim lodgement date), assessment duration (time between the claim lodgement day and the most recent assessment day), and repair duration (time between the most recent assessment day and the repair completion day) for the insured residential buildings in the region affected by the Canterbury earthquake sequence is found to be in the range of 0.5-4 weeks, 1.5- 5 months, and 1-3 years, respectively. The results of this phase will provide useful information to earthquake engineering researchers working on seismic risk/loss and insurance modelling.
Christchurch was struck by a 6.3 magnitude earthquake on the 22 February 2011. The quake devastated the city, taking lives and causing widespread damage to the inner city and suburban homes. The central city lost over half its buildings and over 7000 homes were condemned throughout Christchurch. The loss of such a great number of homes has created the requirement for new housing to replace those that were lost. Many of which were located in the eastern, less affluent, suburbs. The response to the housing shortage is the planned creation of large scale subdivisions on the outskirts of the city. Whilst this provides the required housing it creates additional sprawl to a city that does not need it. The extension of Christchurch’s existing suburban sprawl puts pressure on roading and pushes residents further out of the city, creating a disconnection between them. Christchurch’s central city had a very small residential population prior to the earthquakes with very few options for dense inner city living. The proposed rebuild of the inner city calls for a new ‘dense, vibrant and diverse central hub’. Proposing the introduction of new residential units within the central city. However the placement of the low-rise housing in a key attribute of the rebuild, the eastern green ‘Frame’, diminishes its value as open green space. The proposed housing will also be restrictive in its target market and therefore the idea of a ‘vibrant’ inner city is difficult to achieve. This thesis acts as response to the planned rebuild of inner Christchurch. Proposing the creation of a model for inner city housing which provides an alternative option to the proposed housing and existing and ongoing suburban sprawl. The design options were explored through a design-led process were the options were critiqued and developed. The ‘final’ proposal is comprises of three tall towers, aptly named the Triple Towers, which condense the proposed low-rise housing from an 11000 square metre footprint to combined footprint of 1500 square metres. The result is an expansion of the publicly available green space along the proposed eastern frame of the city. The height of the project challenges the height restrictions and is provocative in its proposal and placement. The design explores the relationships between the occupants, the building, the ‘Frame’ and the central city. The project is discussed through an exploration of the architecture of Rem Koolhaas, Renzo Piano and Oscar Niemeyer. Rather than their architecture being taken as a direct influence on which the design is based the discussion revolves around how and why each piece of comparative architecture is relevant to the designs desired outcome.
Diverse Density proposes an alternative housing strategy to the idealistic top-down process of housing development. The term ‘Top – down’ refers to a situation in which decisions are made by a few people in authority rather than by the people who are affected by the decisions (Cambridge). Problems/Position/Question: New Zealand’s urban housing is in a period of flux. Pressures of densification have permitted the intervention of medium density housing development schemes but these are not always successful. These typically top-down processes often result in internally focused design schemes that do not adhere to their specific context. The subsequent design outcomes can cause detrimental impacts to the local, urban and architectural conditions. With vast quantities of council regulations, building restrictions and design guidelines clouding over the housing sector, commonly referred to as ‘red tape’, occupant participation in the housing development sector is dwindling. A boundless separation between top-down and traditional housing processes has occurred and our existing neighbourhoods and historic architectural character are taking on the brunt of the problem. The thought-provoking, alternative housings strategies of key research theorists Alejandro Aravena and John Habraken frame positions that challenge contemporary densification methods with an alternative strategy. This position is addressed by endeavoring to answer; How can demands for denser housing achieve dynamic design responses that adhere to changes in occupancy, function and local site conditions? Aim: The aim of this thesis is to challenge New Zealand’s current housing densification methods by proposing an alternative densification strategy. Explicit devotion will be attributed to opposing top-down building developments. Secondly, this thesis aims to test a speculative site-specific housing model. The implementation of a Christchurch housing scenario will situate an investigative study to test the strategy and its ability to stimulate greater diversity, site responsiveness, functional adaptability and occupancy permutation. The post-earthquake housing conditions of Christchurch provide an appropriate scenario to test and implement design-led investigations. Objectives: The primary objectives of this design-led research investigation it to challenge the idealistic top-down method of developing density with a new method to: - Develop contextual architectural cohesion - Encourage residential diversity - Reinvigorate architectural autonomy - Respond to, and recognise, existing site conditions - Develop a housing model that: - Adapts to occupant functionality preferences - Caters to occupancy diversity - Achieves contextual responsiveness The proposition is addressed through a speculative design-led scenario study. A well-established Christchurch urban environment is adopted to implement and critique the envisioned alternative strategy. Development of the designs responsiveness, adaptability, and functionality produce a prototype housing model that actively adheres to its particular context. Implication: The implications of this research would be an alternative densification strategy to perceive the advancement of punctual assessment of building compliance. With accelerated building processes, the research may have implications for addressing New Zealand’s housing crisis whilst simultaneously providing diverse, personable and responsive architectural solutions. A more dynamic, up-to-date and responsive housing development sector would be informed.
