It　was　observed　in　the　electrostatic　sensitivity　test　of4kinds　ofboron－oxidant　mixture　that　the　minimum　energy　of50％ignition（E₅₀）became　higher，the　gap　length　seemed　longer，and　the　series　resistance　increased　with　smaller　condencer capacitance，but　the　apparent　time　constant　was　independent　on　capacitance．With　the　shorter　gap　Iength，E50 measured　in the　approachillg　electrode　method　showed　to　be　smaller　than　the　E50　in　the fixed　gap　method．The　time　constallts　were　the same　in　the　two　methods．As　already　reported　by the　authors，the　two　classes　of　characteristics　of　the　electmstatic sensitivity　of　boron－oxidant　mixtures　were　also　observed　ill　this　experiment　under　the　conditions　of　time　constant　between l20 and 150μs，of　E50 of 20mJ，and　of　the　gap　length　of　O.5mm.

The　essential　requirement　to　the　monitoring　system　of　the　bearing　damages　is　not　only　to　detect　thembut　also　classify their　properties．This　paper　concems　the　fundamental　study　corresponding　to　that　task．In　this　paper，a　trial　to　apply　neuraI networks（NN）to detect and to classify the typical bearing damages，i．e. defectonitsouter raceandoilseizure，by making the　Inost　of　its　pattem　recognition　performance．For　the　monitoring　system　using　NN，it　is　important　to　select（1）the network　architecture，（2）iteration　number　of　leaming，and（3）input　data．By　experiment　carried　out　with　a　test　rotor suspended　by　ball－bearings，the　points（1），（2）and（3）are　discussed　on　the　basis　of　correctly　diagnosed　rate，As　the results，the　remarkable　performance　of　detection　alld　classification is　obtained　in　the　collditions　shown　next：Root mean- square，skewness　and　kurtosis　of　thne　series　of　vibration　sigllal　are　used　as　inputs　to　three（3）units　in　an　input－layer　alld the　number　of　unlt　in　a　hidden 1ayer　and　an　output layer　are　four（4）and　one（1）respectively．Simplified　determining method　for　iteration　number　of　leaming　procedure　and　NN　architecture　is　considered　during　these　trial　process．Adding　to this，by　using　advanced　leaming　method（Kalman　Neum　Training　Method），the　iteration　in　leaming　procedure　can　be reduced　drastically．