The vertical hydraulic cylinder has the similar theory with the horizontal ones。

    The  above  analysis  describes  the  relationship  between characteristic  signals,  fault phenomena  and fault  causes。During the operation, each characteristic signal is related to many phenomena and causes of fault while each phenomenon or cause of faults may be indicated by many characteristic signals。

     As mentioned above, we can diagnosis some failure causes by fuzzy neural network based on the sensor we have。 The failure causes are as fellows: shortage of oil, hydraulic pump failure, relief valve failure, electromagnetic reversing valve failure,  bi-directional  hydraulic  lock  failure,  leakage  of horizontal hydraulic cylinder of legs, leakage of vertical hydraulic cylinder of legs, and obstruction of back pressure valve and oil filters, etc。

     According to the related design and tuning of the parameters of the hydraulic system, the normal range of characteristic signal parameters and the severity of the possible deviation are obtained (as shown in table 1)。

         Table 1 Normal range of characteristic signal of the hydraulic system

2。2 Fuzzification process and selection on membership functions of characteristic signal of the hydraulic system

      According to the measurement of each characteristic signal parameter of hydraulic system, we can know if the parameter is normal, slants small or slants big。 As for the membership degree in the range, namely the membership degree between fault causes  and  fault phenomena,  it  is defined by the corresponding membership functions。

      The relevance between  fuzzy membership  functions and actual  situation  affects  the  diagnosis  results  directly。 Therefore, to determine the membership function is the key to

the whole fault diagnosis。 In many cases, according to the actual situation, the most simple and effective method is to use some common membership function to approximately express some fuzzy variables。 According to past experience and  actual change  of parameters,  this paper selects the commonly-used bell membership functions as a normal state of  membership  functions,  the  down-Z-type  membership functions as slants small state of membership functions and up-Z-type  membership  functions  as  slants  big  state  of membership functions。

    Considering that there is no obvious boundary of these fuzzy concepts of slants small, normal and slants small, overlapping part must be set for these membership functions reflected in the membership function curve of fuzzy sets。 Choosing the right overlap rate is an important factor to guarantee the reliability of the diagnosis。 With reference to past experience, the overlap rate of the membership functions of this paper was selected between 0。2 and 0。6。

     After a comprehensive consideration of the number, shape,position  distribution,  overlapping  rate  and  so  on,  we determined membership  functions  of characteristic  signal parameters of the hydraulic system of outriggers (Figs。 3, 4, 5,6, 7 and 8)。

     According to actual situation of the operation, we have adjusted the parameters for the membership functions。

Fig。 3 Membership functions of the temperature of hydraulic oil

Fig。 4 Membership functions of the oil level

Fig。 5 Membership functions of the oil relief pressure

Fig。 6 Membership functions of the work pressure of  hydrocylinder

Fig。 7 Membership functions of the work flow of hydrocylinder 起重机液压系统支腿的智能故障诊断英文文献和中文翻译(3):http://www.chuibin.com/fanyi/lunwen_82719.html