1.1. Performance-Based Design
Architectural problems generally combine a great multitude of objectives, which pertain to different fields, such as cultural, aesthetic, economic, structural and energetic ones. As these objectives often contrast with each other, it is crucial for the design team to find the most favorable solution in overall terms.
According to the classification proposed by Shi [1
], architectural objectives can be divided into three categories: structural performance, performance of the physical environment—both of which can be quantified—and aesthetic and cultural performance, which relates to unquantifiable aspects. Digital simulations and measurements on scale-models allow for the evaluation of the quantifiable performances of the design proposals prior to their construction. In common practice, such tools are generally used in late design phases to verify the adherence to the performance requirements prescribed in codes, standards and laws, and evaluate the need for late design adjustments [2
However, in the recent years, another approach has gained popularity, according to which performance simulations are used to drive the design process. This approach is known as performance-based design (PBD): pertinent information on one or more performance aspects is gathered since early design phases, and the proposals are iteratively optimized based on the performance feedbacks. The design process follows the loop of generation–evaluation–modification, until a solution that meets the performance goal is achieved. While not discouraging the inclusion of unquantifiable goals and aesthetical considerations, pertinent information on quantifiable performances can greatly support decision-making processes in the conceptual design stages.
In early design phases, designers consider a wide range of possible design solutions. Design decisions made at this stage have the greatest impact on the final performances, while late-time design adjustments can rarely compensate poor decisions made in early stages [5
]. Moreover, design modification taken in early design phases is less costly to implement than those taken in subsequent phases [4
]. Therefore, PBD approach benefits both design workflow and final outcome, combining a decrease in cost and time, and enhances design quality [10
]. Indeed, with PBD designers have a greater control over the performances since early design phases and, as a result, the need for late-time design modifications or “a posteriori” measures is prevented, enhancing the overall efficiency of the design process.
The PBD approach emerged in the l970s and has become increasingly more appealing to architects due to the technological advancement and, in particular, to the development of performance simulation and parametric modeling tools [1
]. Some educational experiences, such as those in [16
], suggest that architectural students are increasingly being encouraged to use these tools and include performance feedback to support design decisions during the early-stage design exploration.
PBD marks the paradigm shift from the traditional “form-making” to the “form-finding” approach [18
]. It allows one to displace traditional know-how, enabling designers to understand the effects of different design features of the proposal on performance, and to identify design scenarios that best fulfill the unique requirements of each project.
The PBD method can be subdivided into two subclasses according to the way the design optimization process is conducted. In “formation models” the modifications are applied manually by the operator, while in “generative models” the design proposals are directly optimized by the computer [19
]. The latter subclass is also known as “performance-driven design” [1
Following the manual procedure, the designers control the form-generation process, allowing for the introduction of unquantifiable criteria and the technical expertise of the operator. For instance, based on performance simulation feedbacks and technical knowledge, the operator generates new design alternatives and tests them until a satisfying solution is achieved. The generation of design proposals can be eased by the use of parametric modeling tools, such as Grasshopper [20
] for Rhinoceros [21
] and GenerativeComponents [22
]. These tools allow one to define complex geometries and to easily modify them by controlling their parameters, thus preventing the need for the operator to manually redraw each design iteration. Adequate technical skills are required in manual processes, since the success of the optimization greatly relies on the correct understanding of the relations between design features and performance [2
]. However, the time and manpower required in the processes may limit the number of iterations pursued and the effectiveness of the design optimization [1
On the other hand, generative models allow one to explore a wide number of design options with a limited involvement of the operator, exploiting the functionalities of optimization tools (e.g., Galapagos [24
] and Octopus [25
]. These tools, when paired to a parametric model and a performance simulation tool, allow one to automate the search of the most performing solution within the variation space defined by the operator, while also narrowing the space of possible solutions based on the estimated performance [27
]. The population of candidate solutions evolves over many generations, until a satisfactory solution is reached. This enables one to explore a wide solution space and to find potentially unconsidered design options to address the specific requirements of the project [3
]. In automated processes it is also possible to effectively combine different performance goals in multiobjective optimization procedures, which can hardly be implemented manually. In automated procedures, the involvement of the designer is generally limited to the definition of the target performance objectives and of the boundaries of the variation space within which the generative process operates, which may reflect quantifiable and unquantifiable criteria [11
]. Despite these advantages, manual procedures may be preferred by professionals to allow the design exploration to be guided by their intuition and expertise gathered by working in the field.
Manual and automated procedures are often combined in different ways in hybrid methods, allowing one to exploit the advantages of both approaches, based on the requirements of the design process.
The PBD approach enhances the efficiency of the design process by enabling to optimize the architectural proposal with respect to the performance analyzed. However, the application of PBD method is still relatively limited currently. Indeed, current architectural practice often relies on experience-based know-how and performance simulations are mainly introduced in late design phases with the aim to verify the adherence to the performance requirements.
1.2. Performance-Based Design in Acoustics
The implementation of PBD in the architectural acoustics field would allow the designer to better combine acoustic performance objectives with architectural goals. Architectural design and acoustic performances are strictly linked: the emitted sound is altered by the architectural space within which it is deployed, due to sound reflection, absorption and diffusion phenomena occurring over its surfaces. In common practice, however, acoustic concerns are mainly restricted to the design of spaces intended for artistic performances, such as music venues and theatres. In such spaces, the architectural environment is meant to support the sound generated by the artists, and acoustic design is critical for both audience and performers [12
However, as the benefits of acoustic comfort on the well-being of the population are being acknowledged, acoustics concerns are introduced in a wider variety of design problems [29
]. Indeed, acoustic requirements are being increasingly extended to the design of spaces not related to artistic performances, such as classrooms, workplaces and urban environments, where an appropriate acoustic performance would benefit the hosted activities and the well-being of the users [29
The acoustic performances of architectural spaces can be described by a number of parameters (e.g., sound pressure level, sound strength, reverberation time, clarity, etc.), each accounting for different perceptual aspects [33
]. The acoustic requirements vary in accordance with the function hosted in the space. For instance, in spaces intended for speech, as classrooms and conference rooms, early sound reflections need to be adequately controlled to ensure the speech intelligibility [35
]. In music venues, i.e., concert halls, opera houses, theatres and open-theatres, a number of parameters are usually considered to account for different perceptual aspects [36
]. The proper management of early and late reflections is crucial, and it is generally obtained by opportunely treating the ambient with reflective and diffusive surfaces. Differently, the acoustic performance of sound reproduction rooms, like home theatres and recording studios, should be neutral, to prevent the space to alter the perception of recorded sounds; in this case a combination of sound diffusing and sound absorbing surfaces is preferred [37
Currently, different commercial acoustic simulation tools are available (e.g., Odeon [38
], CATT-Acoustic [39
], Pachyderm Acoustics [40
], etc.), allowing professionals to estimate the performance of design proposals using the geometrical acoustic method. The acoustic performance of a given environment is predicted based on its geometrical features and the acoustic properties of the materials applied to the surfaces. Normally, the acoustic analysis is run in an external application from the modeling environment, and a specific virtual model need to be prepared (e.g., geometrical simplification, surfaces divided into layers based on material and specific format) in order to be fed to the acoustic simulation tool. Besides geometrical acoustic simulations, some more sophisticated simulation methods, such as wave-based ones, have been applied to concert halls and other bigger environments [41
]. However, these methods still require long simulation time and are not currently supported in any commercial acoustic simulation tool.
Although architectural design should pursue aesthetical quality in parallel with acoustic performance objectives, the process of conciliating acoustic requirements and architectural quality is often difficult and time-consuming, given the different design approaches and criteria of the two disciplines [44
]. In most cases, the architectural and acoustic specialists work rather independently, with relatively few exchanges between them [28
]. For instance, the most common approach followed in the design of concert halls relies on well-known typologies (e.g., shoe-box, fan-shaped, vineyard, etc.) and integrates acoustic simulations in late phases of the design process to verify the adherence of the project to performative requirements. In the design of spaces intended for other purposes, acoustic performances are often overlooked, recurring to acoustic treatments to adjust the performances only in late design phases or after the construction. Since the projects at the final stages are already defined, normally, major form-modifications cannot be pursued anymore, and improvements can only be obtained by altering minor design features, resulting often in costly and little effective solutions [5
In this frame, the implementation of the PBD method in the acoustic field, known as acoustic performance-based design (APBD), would be able to overcome some of the main drawbacks of the traditional method and set-up an effective collaboration between architectural and acoustic specialists. The feedback of the simulations enables one to identify the dependencies between the design features and the acoustic performance, and to optimize the project accordingly. However, despite the mentioned advantages, the application of APBD in current architectural practice appears to be limited.