Research On Prediction Algorithm Of Concrete Slump

Concrete refers to the general name of engineering composite materials that integrate water,cement,mineral powder,fly ash,etc.,and is the main engineering material that affects the safety of buildings.It is widely used in the irrigation link of building construction.The slump of concrete embodies the workability of concrete,including the water retention,cohesiveness,and compressive properties of concrete,which is specifically expressed as the value of concrete collapse under natural conditions.Once the slump of the concrete does not reach the index,it will affect its workability.The specific performance is that too large will cause the concrete to fall off,and if it is too small,the poor fluidity is not conducive to irrigation.In turn,it will affect the stability of the building and cause great harm.At present,the industry still adopts manual measurement of the slump of concrete,which can no longer meet the requirements of precision,adjustment,and predictability.With the support of the school-enterprise cooperation project,based on the actual and research needs in the current working conditions,this thesis proposes a fast and accurate slump prediction algorithm model that incorporates artificial intelligence methods.By studying the research status of a concrete slump,this thesis understands several of the most important factors that affect slump.A prediction model based on an improved particle swarm optimization algorithm combined with a neural network is designed,and the neural network is trained by using the ratio of materials as the input of the neural network.The paper mainly proposes two improved particle swarm optimization algorithms and combines them separately with neural networks to obtain two prediction models.(1)By improving the particle initialization method of the traditional particle swarm optimization algorithm,the particle swarm initialization fitness evaluation function is further improved.On the basis of the initialization of the traditional particle swarm fitness function,the small batch of particles is divided,the fitness function of each small batch of particle swarms is calculated separately,and then the fitness function of the subpopulation is obtained using the min function.(2)Due to the introduction of segment k division in the improvement process,the improved algorithm 2 is further studied,and a self-adjusting k-valued particle swarm optimization neural network algorithm is proposed.The difference between this self-adjusting particle swarm optimization algorithm and the improved algorithm 1 is that once the number of populations and particles is determined,the k value of the improved algorithm 1 is fixed.The improved algorithm two is to combine the particle searchability.Since the particle search ability will change according to the position of each generation of particles and the optimal update of the individual,k will change with each iteration to achieve the self-adjustment purposes.Experimental simulation results: show that the model proposed in this thesis has achieved good prediction results on public data sets and concrete slump data sets;the improved algorithm proposed has higher prediction accuracy than the decision tree algorithm(DT),BP Neural network algorithm,support vector machine algorithm(SVM)and traditional particle swarm optimization neural network algorithm on concrete slump data sets.