الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 188 |  |  |
| | | from | 188 | | |
المستخلص
The following conclusions can be drawn from the study:
e prototype Case-Based Reasoning system is proceeding in retrieving the structural number of an existing asphalt pavement.
e constructed knowledge base includes two thousands and four hundreds cases (structural asphalt pavement sections).
ase description consists of three parts: title, description, and questions. The description includes the deflection parameters 0” D2, D3, D4, Ds, 06, and D7 while questions include 0, and D2•
e knowledge base shows that case coverage is very well because e standard deviation for all parameters is very low and the eatures means are so closely to their medians.
en tests were done to define features (description parameters) eights. They indicate that the suitable weights were 35% for the yen description parameters and 65% for the two question• rameters.’ Then, the suitable weight for each description meter is equal 5% and equal 32.5% for each question arameter.
undred tests were done to check system accuracy and rformance. They indicate that the system is capable to retrieve the ost matching case to the queried case with higher distance easure. The distance measure for the hundred tests ranges from 1% to 98%.
ests results indicate that 47% of tests achieve distance measure ater than 95%, while 40% of them indicate distance mesure
ranges from 93% to 95% and 13% of them indicate distance measure ranges from 90% to 92%.
The test cases show that 910/0 of cases achieve distance measure greater than 90%, that is conforming to the methodology of using such systems in that at least 90% of cases should achieve distance ’measure greater than 90%.
e retrieval time is very small for all test cases. It can not measured by CBR Express. As well as the case is queried, the retrieved case is brought immediately by the system.
Global system verification indicates that the prototype CBR system is proceeding in retrieving the most matching case to the queried case. The performance and accuracy of the retrieval process is very cceptable.
ecommendations
put the prototype CBR system into actual application, a real ld measuring knowledge base should be constructed using the ling weight detlectometer in measuring deflection basin.
constructed knowledge base should be as greater as possible to
ver all probabilities of existing of different layer types and ckness.