We present a novel architecture for an AI system that allows a priori knowledge to combine with deep learning. In traditional neural networks, all available data is pooled at the input layer. Our alternative neural network is constructed so that partial representations (invariants) are learned in the intermediate layers, which can then be combined with a priori knowledge or with other predictive analyses of the same data. This leads to smaller training datasets due to more efficient learning. In addition, because this architecture allows inclusion of a priori knowledge and interpretable predictive models, the interpretability of the entire system increases while the data can still be used in a black box neural network. Our system makes use of networks of neurons rather than single neurons to enable the representation of approximations (invariants) of the output.
LINK
This report was produced within the framework of the RAAK PRP project ‘Veiligheid op de werkvloer’. Personal protective equipment (PPE) is used on a daily basis by millions of people all over the EU, voluntarily or as a result of EU legislation. In this report we deal specifically with the textile/garment aspects of PPE. In this context we must consider the fact that PPE encompasses a huge area with hundreds of different applications of materials and systems tuned to specific needs;from a materials point of view it represents a complex area due to the large diversity of labour conditions. Textiles and clothing represent an area where PPE is an important area of attention. On a global scale it is an area of much research. Safety and comfort are becoming more and more important and these aspects must be in balance. Uncomfortable systems will not be used and put safe working at risk. Thus there is a continuous need for technological innovation to improve the effectiveness of PPE systems. Specialization and specific combinations aimed at use under well-defined conditions contributes to finding a good balance between comfort and safety. The design of products, taking into account the individual needs represent an area of intensive research: Safety directed ‘fashion design’.The ultimate goal is the development of proactive systems by which workers (but capital goods as well) are optimally protected. There is also a lot of attention for maintenance and cleaning since protective functions may deteriorate as a result of cleaning processes. Another important point is standardization because producers need directions for product development and supply of goods. In our overview we make a distinction between static and dynamic systems. Static systems provide passive protection, simply by being a part of an equipment that separates the worker from the danger zone. Dynamic systems are more ‘intelligent’ because these can react to stimuli and subsequently can take action. These dynamic systems use sensors, communication technology and actuators. From this research the following may be concluded: 1. Safety is obtained by choice of materials for a textile construction, including the use of coatings with special properties, application of specific additives and he use of special designed fibre shapes. 2. The architecture and ultimate construction and the combinations with other materials result in products that respond adequately. This is of great importance because of the balance comfort – safety. But a lot can be improved in this respect. 3. Insight in human behaviour, ambient intelligence and systems technology will lead to new routes for product development and a more active approach and higher levels of safety on the work floor. Consequently there is a lot of research going on that is aimed at improved materials and systems. Also due to the enormous research area of smart textiles a lot of development is aimed at the integration of new technology for application in PPE. This results in complex products that enhance both passive and active safety. Especially the commissioners, government and industry, must pay a lot of attention to specifying the required properties that a product should meet under the specific conditions. This has a cost aspect as well because production volumes are usually not that large if for small groups of products specific demands are defined. We expect that through the technology that is being developed in the scope of mass customization production technologies will be developed that allows production at acceptable cost, but still aimed at products that have specific properties for unique application areas. Purchasing is now being practiced through large procurements. We must than consider the fact that specification takes place on the basis of functionality. In that case we should move away from the current cost focus but the attention should shift towards the life cycle
MULTIFILE
Thermal comfort -the state of mind, which expresses satisfaction with the thermal environment- is an important aspect of the building design process as modern man spends most of the day indoors. This paper reviews the developments in indoor thermal comfort research and practice since the second half of the 1990s, and groups these developments around two main themes; (i) thermal comfort models and standards, and (ii) advances in computerization. Within the first theme, the PMV-model (Predicted Mean Vote), created by Fanger in the late 1960s is discussed in the light of the emergence of models of adaptive thermal comfort. The adaptive models are based on adaptive opportunities of occupants and are related to options of personal control of the indoor climate and psychology and performance. Both models have been considered in the latest round of thermal comfort standard revisions. The second theme focuses on the ever increasing role played by computerization in thermal comfort research and practice, including sophisticated multi-segmental modeling and building performance simulation, transient thermal conditions and interactions, thermal manikins.
Door voortgaande stedelijke verdichting staat de groene ruimte in Nederlandse steden in toenemende mate onder druk. Veel parken, tuinen en plantsoenen staan volop in de belangstelling en worden opgewaardeerd. In de luwte daarvan zijn er vele, vaak wat verborgen, groengebieden die ook waardevol zijn maar slechts terloops aandacht krijgen. Steeds meer overhoeken en kleine en grotere groengebieden raken bebouwd of verwaarloosd, met afname, versnippering en verlies van ruimtelijke samenhang van de stedelijke groenstructuur als gevolg. Wie behartigt de belangen van deze, wat meer terzijde gelegen, groengebieden? Wie agendeert hun waarde? Wat is de potentie van deze groengebieden? Een recente studie van de Nederlandse Tuinenstichting, toont aan dat in een stad als Amsterdam nog diverse schijnbaar onsamenhangende groene gebieden aanwezig zijn, die bij nadere beschouwing een aaneengesloten stedelijke ruimtestructuur blijken te kunnen vormen. Naast elkaar gelegen groengebiedjes, privétuinen, verscholen binnentuinen, pleintjes en lijnvormige groen- en waterelementen (zoals grachten of bewaard gebleven poldersloten) blijken een onvermoede samenhang te hebben. Aanwezige fiets- en wandelpaden, trottoirs en wegen rijgen deze plekken aaneen tot waardevolle informele groen- en ruimtestructuren. Veel van deze plekken hebben een historische verankering met de aanwezige bebouwing en wegenpatronen, waardoor ze kunnen worden gezien als 'ensembles' en bepalend zijn voor de identiteit of 'genius loci' van delen van de stad. Ze vertegenwoordigen een hoge waarde voor mens en natuur en hebben vaak een tuinachtig karakter. Het onderzoek richt zich op het identificeren en analyseren van dergelijke stedelijke groenstructuren in een aantal Nederlandse steden en op het zoeken naar mogelijkheden om de ruimtelijke samenhang en betekenis van deze gebieden voor de stad te vergroten. Mogelijk kunnen de onderzochte gebieden bij stedelijke inbreidingsopgaven model staan voor de ontwikkeling van nieuwe, duurzame groene ruimtestructuren voor de stad.
