A Kohonen Network Model for Performance Evaluation of Asphalt Concrete Pavement

Analyses the current road pavement maintenance technical specification, put forward a new evaluation method for maintenance and overlay of old asphalt pavement. Considering the uncertainty and complexity in the process of performance evaluation, and based on the entropy and the Kohonen network theory, the performance evaluation system is established in Kohonen network of fuzzy entropy method, combine the expert’s weight and entropy method to confirm the weight, in order to reflect the subjectivity of classification and objectivity of measured data at the same time. Keyword: Road engineering; Asphalt pavement performance; Kohonen network; Fuzzy entropy. In the current "Chinese technical specifications for maintenance of highway asphalt pavement" [1], the sub-item state of road’s performance evaluation content including the road surface breakage condition, the riding quality, the intensity and the anti-slippery performance, the indicator which uses corresponds separately for pavement condition index (PCI), road quality index (RQI), surface strength index (SSI), sideway force coefficient (SFC or BPN), composite evaluation index (PQI) to be used the pavement comprehensive assessment. Based on the above various subitems indicator evaluation results to count, and corrected weighting for each evaluation rank's intermediate quantity [2-5]. This evaluation method has 3 problems [6-9]: the differences of the road condition within the same rank is neglected easily or cannot obtain manifests. When the difference of the sub-item indicator is not big, sometimes the sub-item rank's determination can enable the PPQI value to change greatly. The determined of pavement overall evaluation rank is not enough direct. Therefore, it was difficult to compare among different road sections, also difficult to determine and optimization. The pavement performance is a multifactor comprehensive evaluation problem, and will inevitably contain a synthetic judgment of various factors in the process of all kinds of judgment [10-12], so, there will have certain human difference, and the fuzzy phenomenon both between “superior” and “good” and “good” and “medium”. Therefore, this paper based on the fuzzy entropy and the Kohonen network method, establish an evaluation model of asphalt pavement performance which used various indicator performance parameters, provide a new way for evaluating and forecasting the environmental impact assessment of asphalt pavement [13-15]. 1. SAMPLE DISTINCTION According to "Chinese technical specifications for maintenance of highway asphalt pavement" [1], scope of sample value was determined as shown in Table 1. When various indicators achieve the marginal value, it will have a serious influence to the road safety operation, therefore, when there was an indicator’s actual value surpass the marginal value in the indicator system, this road section will be determined directly as the bad one, and the influence of other indicators will no longer be considered. 2. WEIGHTS DETERMINATION When the environmental impact assessment was based on the Kohonen network theory, it was extremely important to define the indicator weight which usually influenced the result’s objectivity. In view of subjective factors of experts’ knowledge, experience and value judgment, and objective information characters of actual measuring data, it defined the index weight by combining subjective weight method (analytic hierarchy processanalytical hierarchy AHP) with objective weight method (entropy method), so as to objectively and completely reflect both the important of evaluation index and actual condition of problem. 2.1. Subjective weight According to the asphalt pavement maintenance and technical specification recommended value, the subjective Table 1. Target value and marginal value of evaluation index. Evaluation index Target value Marginal value Pavement condition index PCI 100 30 Surface strength index SSI 1.4 0.4 Road quality index RQI 10 2 Sideway force coefficient SFC 0.7 0 920 The Open Civil Engineering Journal, 2015, Volume 9 Zhi-bin et al. weight is ' i w . ) , , , ( 4 3 2 1 w w w w wi = There are lots of method to determine index weight, such as expert investigation method, analytic hierarchy process (AHP), etc, and the AHP is chosen. The acquired weight set is showed: ) , , , , ( 5 4 3 2 1 w w w w w w = 2.2. The objective weight is determined by entropy method In the information theory, the entropy value which reflected the disordering degree of the information may use to measure the size of the information, one index carried the more information, indicating that the more useful this index have in making the decision, while the entropy value is smaller, namely the system's disordering degree is smaller. Therefore the order degree and effectiveness can be evaluated by the information entropy, namely the judgment matrix made up by the evaluating indicator value determined the weight of each evaluating indicator. Its main computation step is as follows: Generally all the indexes are normalized to unit interval 0-1. In order to reflecting critical effect of target state and index, the membership degree in fuzzy mathematics is introduced in standardization. Positive indicator: bigger always better, use upper semitrapezoid distribution function for standardization.