Abnormality detecting device for construction machine

An abnormality detecting device for a construction machine equipped with a plurality of sensor means placed at a plurality of parts of the construction machine, the sensor means detecting a plurality of related physical states and outputting a plurality of pieces of sensor information, comprising:

• correlation coefficient calculation means which receives the plurality of pieces of sensor information outputted from the plurality of sensor means, generates time-series physical state information in a predetermined period on each of the plurality of pieces of sensor information corresponding to the plurality of sensor means, and calculates a plurality of correlation coefficients between separated pieces of time-series physical state information on each of the plurality of pieces of sensor information;
• correlation coefficient comparison means which compares the correlation coefficients calculated by the correlation coefficient calculation means and calculates the degree of difference of each correlation coefficient from other correlation coefficients;
• and
• abnormality judgment means which judges that when the degree of difference calculated by the correlation coefficient comparison means exceeds a preset value, an abnormality has occurred to a part related to the corresponding sensor means,
• wherein assuming that Xi and Xj (i,j=1, . . . , n) (n:
• the number of the sensor means supplying the plurality of pieces of sensor information) represent two pieces of time-series physical state information arbitrarily selected from the time-series physical state information generated by the correlation coefficient calculation means, Xi(t) and Xj(t) represent the measurement values of Xi and Xj at time t and ρ(i, j) represents the correlation coefficient between the input values Xi and Xj between time t=0 and time t=Δt(T−1) (Δt:
• the time between measurements, T:
• the number of measurements), the correlation coefficien ρ(i, j) is calculated using averages μi and μj and standard deviations σi and σj according to the following equation:
• ρ(i, j)=Σ(Xi(t)−μi)(Xj(t)−μj)/(T·σi·σj) where
• μi=ΣXi(t)/T, μj=ΣXj(t)/T
• σi=(nΣXi(t)2−(ΣXi(t))2)/(T·(T−1))
• σj=(nΣXj(t)2−(ΣXj(t))2)/(T·(T−1)).