Hybrid Neural – Based Model for Predicting the Construction Project Performance of High – Rise Building Projects in Metro Manila, Philippines

Assessing construction project performance through adapting an innovative approach can multiply the production of high-quality project outputs which is a catalyst to the socio-economic progress of a country. Preliminary data collection was employed through a meta-cognitive analysis of reviewing related literature which directs to the backbone of qualitative information that is relevant to the periodically experienced construction performance-influencing factors and to develop an assessment questionnaire about the influencing factors affecting the project performance. The IBM SPSS program was used to verify the reliability and consistency of the fundamental statistics of the questionnaire responses of cost, time and quality performance ratings. A predictive mathematical model was developed for forecasting cost, time, and quality performance rating employing Levenberg-Marquardt training algorithm with Hyperbolic Tangent Sigmoid function. The prediction model result shows a highly satisfying performance on its variance from the substantive values and suggests a high correlation between these values. The relative importance of the factors affecting the cost, time, and quality performance rating was calculated via sensitivity analysis through connection weights using Garson’s Algorithm to view the order of influence of the parameters that have great significance to the success of a project.

Relative Contribution of Process Control Parameters on the Raveling Resistance of Microsurfacing Mix

Raveling is one of the key performance parameters of microsurfacing treatment. During the material handling and mix production, process control parameters including aggregate gradation, emulsion content, and water content vary inevitably and might increase the risk of raveling. The objective of this study was to quantify the relative contribution of these process control parameters on the raveling resistance of the microsurfacing mix. For this purpose, a total of 30 combinations of aggregate gradation, emulsion content, and water content were subjected to raveling using wet track abrasion test. The investigations showed that the raveling increased for coarser gradation and lower emulsion content, whereas the variation in raveling was minimal with water content. Further, the test results were modeled using an artificial neural network (ANN). The ANN model was able to capture the influence of process control parameters on the raveling resistance of the microsurfacing mix. Garson’s algorithm was used to quantify the relative contribution of each process control parameter on raveling. It was found that the relative contributions of aggregate gradation, emulsion content, and water content were 40%, 28%, and 32%, respectively. Because of their substantial contribution, it is critical to ensure proper quality control of process control parameters during material handling and production of microsurfacing mix. In particular, coarser aggregate gradation in conjunction with lower emulsion content should be avoided to minimize the risk of raveling.

Interpreting the neural networkfor prediction of fermentation of thick juice from sugar beet processing

Methods that can provide adequate accuracy in the estimation of variables from incomplete information are desirable for the prediction of fermentation processes. A feed-forward back-propagation artificial neural network was used for modelling of thick juice fermentation. Fermentation time and starting sugar content were usedas input variables, i.e. nodes. Neural network had one output node (ethanol content, yeast cell number or sugar content). The hidden layer had nine neurons. Garson's algorithm and connection weights were used for interpreting neural network. The inadequacy of Garson's algorithm can be seen by comparing with the results of regression analysis, which indicates that the influence of the fermentation time is higher. A better agreement of the results was obtained using network connection weights, a method that can be used to determine the relative importance of input variables.