INSPECTION DATA

In this analysis, we studied the small and medium span reinforced concrete slab bridges with span of around 2m to 20m. In planning the priority maintenance scheme, the bridge’s integrity and degree of deterioration should be known in order to make engineering decisions with their own implicit knowledge. According to the current standard, if the health diagnosis of a bridge is judged as Ia or Ib, then detailed inspection A (no damage recorded) will be applied. Additionally, if the health diagnosis of the bridge is judged as II, III, or VI, then detailed inspection B (with damage recorded) will be applied [3, 4]. However, if the health diagnosis of a bridge shows a low amount of deterioration, the degree of health is II. Therefore, even if damage occurred, depending on the specifications of the bridge and the surrounding environment, its health degree could be evaluated as Ib. Hence, the Ib reserve forces of II at the time of the next inspection must be distinguished from the health degree of Ia.

We used the bridge inspection data on 971 small and medium span bridges that are managed by the municipal government (K city, Gifu Prefecture); 75% (725 bridges) are reinforced concrete slab bridges. Figure 4.1 shows the distribution of the numbers and

types of bridges. According to the records of periodic road bridge inspection for small and medium span bridges, 13 types of deterioration of bridge superstructure were included in the bridge inspection records. Figure 4.2 shows the distribution of the numbers of bridge and bridge health index, 97% (944 bridges) bridges’ health index is I and II. Figure 4.3 shows the distribution of the numbers of bridges and bridge length, 88% (850 bridges) bridges’ bridge length is less than 10m. Table 4.1, Table 4.2 and Table 4.3 shows the detail of bridge type, health index and bridge length. From the K city’s bridge inspection data, the reinforced concrete slab bridges give high proportion.

According to this reason, we want to focus evaluate the reinforced concrete slab bridges’ health integrity.

Figure 4.4 shows the distribution of reinforced concrete slab bridges’ number and soundness. Figure 4.5 shows the distribution of reinforced concrete slab bridges’

number and bridge length. According to the records of periodic road bridge inspection for small and medium bridges, 13 types of deterioration of bridge superstructure were included in the bridge inspection records. Figure 4.6 and Figure 4.7 shows the distribution of the number of bridges and the type of deterioration. Figure 4.7 shows that the deterioration of peeling and exposed steel bars had a high proportion compared to other types of deterioration. In this case study, we examined the deteriorating peeling and exposed steel bars in reinforced concrete slab bridges. Because of the limited amount of bridge inspection data, only three types of bridge inspection data were used, the corresponding health rating of which I, II, and III were. Of the 725 reinforced concrete slab bridges, 90% had a health rating of I.Furthermore, the deterioration of peeling and exposed steel bars was not recorded in the inspection records when the health rating of bridge member was assessed at Ib (almost healthy) [1, 3, 4, 7].

Therefore, we focused on the health degree of Ib. We also examined the inspection photos that were not included in the records of periodic road bridge inspections of small and medium span bridges. In this study, the 83 pieces of inspection data pertaining to the deterioration of peeling and exposed steel bars on reinforced concrete slab bridges in Gifu Prefecture were used (see Table 4.4).

In this study, the statistical analyses of two levels of health were conducted.

According our previous study [2], the different experienced inspectors had the same assessment for classifying the health rating of “Good” (health rating: I) and “Poor”

(health rating: II, III, IV). Thus, the inspectors’ influences on the inspection data were not considered in this current study. These levels (“Good” and “Poor”) were used to evaluate the reinforced concrete slab bridges based on actual bridge inspection records in Gifu Prefecture. We examined additional features (additional inspection items) to ass-

Figure 4.1 Distribution of bridge numbers and bridge type

Figure 4.2 Distribution of bridge numbers and health index

Figure 4.3 Distribution of bridge numbers and bridge length

Figure 4.4 Distribution of reinforced concrete slab bridges’ number and soundness

Figure 4.5 Distribution of reinforced concrete slab bridges’ number and length

Figure 4.6 Distribution of bridge number and deterioration type (reinforced concrete slab bridge)

Figure 4.7 Distribution of bridge number and deterioration type (reinforced concrete slab bridge)

Table 4.1 Bridge type and number

Bridge type Bridge number

Steel I-girder bridge 19

Steel girder (H-shaped) bridge 48

Reinforced concrete slab bridge 725

Reinforced concrete T-beam bridge 5 Reinforced concrete box girder bridge 19 Pre-stressed concrete slab bridge 108 Pre-stressed concrete T-beam bridge 23

Unknown 24

Total number 971

Table 4.2 Bridge type and health index

Bridge type I II III IV Total number

Steel I-girder bridge 6 12 1 0 19

Steel girder (H-shaped) bridge 12 26 10 0 48

Reinforced concrete slab bridge 638 79 8 0 725

Reinforced concrete T-beam bridge 0 2 3 0 5

Reinforced concrete box girder bridge 14 4 1 0 19

Pre-stressed concrete slab bridge 98 7 3 0 108

Pre-stressed concrete T-beam bridge 19 4 0 0 23

Unknown 10 13 1 0 24

Total number 797 147 27 0 971

ess the health degree of the reinforced concrete slab by utilizing the SVM. These analyses were conducted using the SVM analysis software Waikato Environment for Knowledge Analysis (WEKA) version 3.9. We conducted this experiment by using the classificationsolver in a Library for Support Vector Machine (LIBSVM) [5] with the RBF kernel (Gaussian kernel) [6].

As pointed out in the pre-processing of the inspection data, all practical implementations of SVMs adhered to strict requirements for training and testing. The first requirement was that all data were numerical. Therefore, if categorical features were present, they were converted to numerical values using variable transformation techniques. Because the SVM model implementations in WEKA do not support missing values, we needed to either remove data with missing values or use some form of data

Table 4.3 Bridge length and bridge number

Bridge length (m) Number of bridge

0-5 623

5-10 227

10-15 30

15-20 20

20-25 24

25-30 25

30-35 7

35-40 5

40-45 3

45-50 1

50-55 0

55-60 1

60-65 0

65-70 0

70-75 1

75-80 0

80-85 0

85-90 0

90-95 1

95-100 0

100-105 0

105-110 1

110-115 2

imputation. Furthermore, SVMs assume that data are in a standard range. To ensure that all feature variables were treated equally, it was best to use the feature “normalization”

before training the model. The input data were the degree of deterioration of each reinforced concrete slab’s inspection item, and the output data were the reinforced concrete slab’s degree of health.