We consider the problem of controlling an invasive mechanical ventilator for pressure-controlled ventilation: a controller must let air in and out of a sedated patient's lungs according to a trajectory of airway pressures specified by a clinician. Hand-tuned PID controllers and similar variants have comprised the industry standard for decades, yet can behave poorly by over-or under-shooting their target or oscillating rapidly. We consider a data-driven machine learning approach: First, we train a simulator based on data we collect from an artificial lung. Then, we train deep neural network controllers on these simulators. We show that our controllers are able to track target pressure waveforms significantly better than PID controllers. We further show that a learned controller generalizes across lungs with varying characteristics much more readily than PID controllers do.
By the knowledge transferring in different areas, analogical design has been considered as a powerful approach to promote the generation of novel ideas in product conceptual design. An efficient representation scheme for design knowledge is vital to implement analogical transferring. In this article, inspired from the structure mapping mechanism of analogical reasoning, a structure mapping–based representation was proposed to support designers to search and use design analogies. This representation can provide designers with insights into the structural information of knowledge situations, and consequently designers are able to implement the corresponding design analogy search at the level of the structural similarity, rather than the functional or superficial similarity. Based on this new representation scheme, a structure mapping–based analogical design framework was developed. In this framework, patents are used as the source of analogical knowledge, and the relational structure–based representation for the patent knowledge is created using the advanced natural language processing tools/algorithms. Next, the search of design analogies is implemented by means of the vector space model, and a new structure mapping–based concept generation model can finally guide the designers to use design analogies. An industrial case and a compared experiment were carried out to verify the feasibility and effectiveness of the proposed framework.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.