Development of an Intelligent System for Optimal Design of High Strength and Durable Concrete Mix Proportion | |
| Author | Rattapoohm Parichatprecha |
| Call Number | AIT Diss. no.ST-08-01 |
| Subject(s) | Intelligent System Optimal designs (Statistics) Concrete mixers |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Engineering in Structural Engineering, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. ST-08-01 |
| Abstract | Nowadays, concrete possessing both high strength and durability, hereinafter called high strength and durable concrete (HSDC), is utili zed globally. However, little attention has been given to HSDC mix design methods. Traditional mix design methods has only been based on the personal experience of each civil engineer. These methods require a large number of trial mixes to achieve a combination of ingredients wh ich meet specific performance requirements. Moreover, the cost of concrete and new types of admixture have not been considered in the traditional method. In this study, a pr ototype of an intelligent system is developed to improve the process of selection and proportioning of constituen ts as well as to make the design process of HSDC readily available to the concrete industry. The work program can be divided into three main parts. The first part deals with the development of HSDC mixtures which improve the performance of concrete by applying high strength concrete producing techniques, novel type s of concrete admixture and concepts of environmental friendliness. In addition to the re sults obtained from these experimental programs, the influence of relevant parameters on workability, compressive strength, and durability were investigated. The second part of this study involve s the application of Artif icial Neural Networks (ANNs) which are used to construct predicti on models for workability, compressive strength, and durability of HSDC. These models are devel oped using the experimental results from the first part combined with results from previ ous research. The accuracy and reliability of the proposed models is evaluated through comparis on with the multiple regression technique. The final part emphasizes on the development of an intelligent system to design optimum HSDC mixtures. When designing HSDC mixtures, engineers have to consider not only the properties of the concrete but also the cost of concrete. To achieve the goal of optimum mix proportions, a prototype of the proposed system is devel oped. ANNs models from the second part are incorporated with genetic algorithm (GA) optimi zation techniques to search for near optimum mixtures while minimizing the cost of the concrete . Further, the proposed system can be used as a simulation tool to analyze or investigate th e relationship between relevant parameters and various properties of HSDC. The first part deals with the development of HSDC mixtures. Rigorous experimental programs were conducted in the laboratory to inve stigate the influence of different pozzolanic materials, cement contents, and water-to-binde r ratios on the workability, compressive strength, and durability of HSDC. The target 28-day compre ssive strength of HSDC mixtures is designed in the range of 40-150 MPa, and the workability of concrete expressed in terms of slump is kept constant at 10-15 cm by varying the dosage of superp lasticizer. In this experiment, a new type of superplasticizer called polycarboxylic ether polymer is used in order to achieve the required workability, compressive strength, and durability. Tw o types of pozzolanic materials were used; namely, pulverized fly ash and condensed silica fu me. Cementitous materials were varied from 400-600 kg/m 3 with W/B ranging from 0.2 to 0.4. C ontrol specimens without pozzolanic materials were also cast and tested to find the optimum cement content in terms of compressive strength and durability. In total, 65 mixtures were made and the specimens were tested for their properties. The durability of each mix is experi mentally investigated by measuring the total charge passed of concrete in accordance with ASTM C1202-97. The results indicated that HSDC mixtures can be satisfactorily developed by employing HSC producing techniques combined with novel types of concrete admixture and c oncepts of environmental friendliness. The influence of pozzolanic materials and relevant parameters on workability, compressive strength and durability were also determined. All experime ntal results were compiled and stored in a database for use in the development of prediction models of concrete properties. The second part deals with the developmen t of ANNs for predicting initial slump, 28-day compressive strength, and durability. The accuracy of the prediction models is dependent on the type and proportion of ingredients, namely ordinary Portland cement, pulverized fly ash, condensed silica fume, ordinary tap water, na phthalene formaldehyde condensates, modified polycarboxylic ether polymers, coarse aggregat e, and fine aggregate. To improve the performance of the prediction models, several hundr ed mixtures from previous research were culled and combined with the data obtained from the testing program from the first part to produce the HSD database. Using the resulting da ta, ANN models were developed, trained, and tested using 511, 179, and 91 records for compressive strength, initial slump, and durability, respectively. The results showed that the ANN m odels are capable of accurately predicting the properties of HSDC. Furthermore, the performa nce of developed ANN models can be improved by updating the database from additional test results so that they can efficiently predict concrete performance. Finally, the system prototype was deve loped by incorporating genetic algorithm optimization technique with artificial neural ne tworks to determine the optimum mixture of HSDC based on its cost. In this study, penalty function technique is incorporated into the objective function to transform the constrained objective function into an unconstrained one. The trained ANNs described in the second part are used to evaluate the fitness function in the optimization process and incorporated with GAs using MATLAB programming. Explanation facilities and user interface were also built by means of knowledge based expert systems (KBES). In the application stage, the user activates the system by providing the required performance and cost of constituent material s. In this way, the optimum proportions and predicted performances of HSDC are achieved. In addition, experimental investigations were carried out to validate the proposed method by comp aring the predicted performance with tested results from trial batches. The results indicate th at the proposed system can be used to design an HSDC mix which corresponds to its required perf ormance. Furthermore, the proposed system takes into account the influence of the fluctuating unit price of materials in order to achieve the lowest cost for the concrete, which cannot easily be obtained by traditional methods or trial-and- error techniques . |
| Year | 2008 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. ST-08-01 |
| Type | Dissertation |
| School | School of Engineering and Technology |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Structural Engineering (STE) /Former Name = Structural Engineering and Construction (ST) |
| Chairperson(s) | Worsak Kanok-Nukulchai ;Pichai Nimityongskul (Co-Chairperson); |
| Examination Committee(s) | Bergado, Dennes T. ;Kunnawee Kanitpong; |
| Scholarship Donor(s) | Royal Thai Government Fellowship; |
| Degree | Thesis (Ph. D.) - Asian Institute of Technology, 2008 |