The author demonstrated a new evaluation method of construction workers' activity sequence.
Chapter 2 describes the previous works in the field of work study in Japan. Analyzes of activities at construction sites in Japan originally started in 1960s and these methods were compiled into “Recommended Practice for Method Improvement on Construction Work” (AIJ, 1990).
Chapter 3 describes the actions at construction sites. The sequence of construction can be regarded as the sum of the sequences of each parts transportation. So the author classified the worker action to nine categories, these are ‘Insert’, ‘Remove’, ‘Arts’, ‘Walk’, ‘Get’, ‘Carry’, ‘Place’, ‘Stay’, and ‘KAKUNIN’, based on the change of condition (e.g. rotation, position, shape, and finishing).
Chapter 4 describes the method to record action categories. The author describes new datasets named ‘Action Array’. It is an array of capital letters of action categories which was captured in construction activity on 1Hz. The author focused on methods in genetics (e.g. transcription and translation in the Central dogma of molecular biology) for solving the problem on translation from capturing data to sequence information.
Chapter 5 explains the flow of calculation of ‘Smoothness’ of Action Array and ‘Similarity’ between pairs of Action Array. The proposed method needs compression of action array, which is a step for shorten consecutive letters. Smoothness is a ratio of sum length of ‘Action Set’ to the length of Compressed Action Array. Action Set means some words which are repeatedly shown in compressed action array. The author classified 7 standard types and 22 derivative types. Similarity is a ratio of the Levenshtein distance to the average length of pair of compressed action array.
Rsmoothness ＝ ( ∑laction set(i) ) ⁄ ( Lcompressed action array ) ------(1)
Rsimilarity ＝ ( Llevenshtein(com.a.a.i, com.a.a.j) ) ⁄ ( (Lcom.a.a.i+Lcom.a.a.j)⁄2 ) ------(2)
Chapter 6 presents the verification of the proposed method by test calculation using sample data which was captured in the experimental construction of scaffoldings. The author got 88 action arrays from the experimental construction and calculated 88 smoothness score and 3916 similarity score. Some results of analysis were shown in chapter 6.
