Research On Inverse Problems Of Air Quality In Human Living Environment With Genetic Algorithm And Adjoint Method

Optimizing an indoor flow pattern according to specific design goals requires systematic evaluation and prediction of the influences of critical flow control conditions such as flow inlet temperature and velocity.In order to identify the best flow control conditions,conventional approach simulates a large number of flow scenarios with different boundary conditions.This research proposes a method that combines the genetic algorithm(GA)with computational fluid dynamics(CFD)technique,which can efficiently predict and optimize the flow inlet conditions with various objective functions.A mixing convection case in a confined space was used to evaluate the performance of the developed method.The study shows that the method can predict accurately the inlet boundary conditions,with given controlling variable values in the space,with fewer CFD cases.The results reveal that the accuracy of inverse prediction is influenced by the error of CFD simulation that need be controlled within 20%.The study further used predicted mean vote(PMV),the percentage dissatisfied of draft(PD)and age of air around occupants as the design goals to optimize the inlet boundary conditions(e.g.,supply velocity,temperature,and angle)of a simple 2-D office and an aircraft cabin.Then this study proposes two new design methods: the constraint method and the optimization method.Both of the two methods success in providing varies design that satisfying all of the design goals with much less calculation than traditional method.The optimization method provides more accurate results while the constraint method needs less computation efforts.Air pollution is becoming more and more seriously in large cities.Accurate and rapid identification of outdoor pollutant sources can facilitate proper and effective air quality management in urban environments.Traditional “trial-error” process is time consuming and is incapacity in distinguishing multiple potential sources,which is common in urban pollution.Inverse prediction methods such as probability based adjoint modelling method have shown viability for locating indoor contaminant sources.This paper advances the adjoint probability method to track outdoor pollutant sources of constant release.The study develops an inverse modelling algorithm that can promptly locate multiple outdoor pollutant sources with limited pollution information detected by a movable sensor or location fixed sensor.Three numerical field experiments are conducted to illustrate and verify the predictions: the first one in an open space,the second one in a virtual urban environment and the third one in Beijing,China.The developed algorithm promptly identifies the source locations in all cases.The accuracy of the algorithms depends on the accuracy of urban building models offered.Fast airflow and heat transfer simulation is necessary in building design,real-time system control,pollution source identification and emergent environment management.Conventional algorithms such as SIMPLE or PISO needs great calculation effort and thus cannot meet the requirement of computing speed.Some developed fast calculation methods such as FFD are lack of prediction accuracy due to the inherent simplifications.This paper developed a new fast computational method named semi-Lagrangian PISO(SLPISO)that integrates the semi-Lagrangian scheme with a PISO solver.The developed SLPISO solver can significantly reduce the computational cost compared to SIMPLE and PISO while maintain the required accuracy.The accuracy and stability of the developed algorithm were analysed.A lid-driven cavity flow case and a mixing convection case in a confined space were used to evaluate and illustrate the performance of the method.Results reveal that SLPISO possesses similar accuracy with PISO but can tolerate much larger time steps in the transient simulation.A larger time step will increase the false diffusion,and thus smooth the gradients of the flow field.A proper diffusion coefficient corresponding with a given time step is,hence,necessary to obtain accurate results.