ѧѧ
ACTA SCIENTIAE CIRCUMSTANTIAE
1999ꡡ19 1 Vol.19No.1 1999



˹ԪˮϵͳģӦԵо*
12ܡ1.
ժҪˮϵͳصоеѵ㣬˹Ԫۺ˼룬ˮϵͳ.ͨBP˹Ԫĳ̿ϵͳģͣؿ˹ԪˮϵͳģӦԣʹˮϵͳоܻͿƻ.
ؼʡ˹Ԫ磻̿ϵͳģͱʶ.
Adjustability of artificial neural network in model-building of water purification system
TIAN Yu1, ZHOU Ding1 
1.Department of Applied Chemistry,Harbin Institute of Technology,Harbin 150001  WANG Baozhen2
,2.Department of Municipal and Environmental Engineering,Harbin Architecture Institute,Harbin 150001
AbstractA theory of artificial neutral network was proposed and applied to characterize ozonation and biological activated carbon system.Through setting up a model of ozonation/BAC system basing on BP artificial neutral network, the factors affecting ozonation/BAC system and the adjustability of the network to water purification system were discussed.
Keywordsartificial neutral network,ozonation and biological activated carbon system,model-building.
1ͳˮϵͳģľ
ˮеĽģҪУģ͡ۺηԼ鼰뾭鹫ʽ.ģǸϵͳѧȷ֪ʶ֪Ķɡԭѧģ.ģһɿԸߣṹӣϵͳԭ֪ʶҪϸ.ȫƵѧģ޵ģʹͨƵѧģͱʽеһЩҲҪʵ鷽ȷ.ηԲ΢̶ֱַ׼ˮ̵ʵҪϸߣ뿼ǲȫֵĳЩ£ηܻᵼ²ȷĺ.þ鼰뾭鹫ʽģķֽʶĳЩ֪ʶģͽṹͨȷģͲ.ַʹʱ֪ܵʶ׼ȷԺ͹㷺ԵԼӦ.ͨǿ3ֽģʵӦжһЩ⣬ڸԵľԣ1ݣģ;ͨԺʵЧģҪһ㣬ڿؽģ͵Ļѧۺͷ.
2˹ԪڽģеӦ
Ϊһ庯ƽڽģֳܴԽ.ҪŵУ(1) 籾һģԵĶ̬ϵͳ˷ǳڷϵͳĽģ.(2) ʵʱʶ̶Աʶ㷨ĿҪܸߣʶ㷨ӡʵӦ.ಢдԪɵģʹÿٲд㷨˱ʶٶ.³ֵоƬΪʵӦṩʻ.(3) ϵͳģ򵥣ֻͨϵͳݼɣҪʶĽ״νṹ֪ʶһձõıʶ23ݣ
ڻĶǿϡԵص㣬ͳıʶŻԵΪĳӦһҪǿıƽΪˮĽģһոµ;.ѡչȽϳBPΪо󣬰Ӧõ̿ϵͳĽģоУʵؿ˹ԪˮϵͳģӦ.

ͼ1 ϵͳˮʴͼ
Fig.1The water purificaton circulation of ozonation biological activated carbon system
3BP˹Ԫĳ̿ϵͳķʵо
ڳ̿ϵͳУͻ̿ˮ໥໥Ӱ죬ˮʴͼͼ1ʾ.ˮˮʡͶʱ䡢̿ʱ䡢΢ȳΪϵͳӰأȷƽ׶ΪҪȷֵʱҪƵķΧֱӹϵϵͳˮˮʵĺûЧܵķ.ܵǷԹϵϵͳȷԼΪصӰ죬̿ϵͳѧģǿհ.ĿǰҲûһѧģܹƽͳУˮˮʡͶʱصķԹϵ4.еʵݣBPԪ罨ģ.
ѡһڵ㡢һڵ3BPڳ̿ϵͳķʵ.ʵõڴʯܳˮ̿ϵͳȾʵо1ʾ.õķͼ2ʾ.ͼ2(a)ͼбƽSSE(Sum Squared Error)ѵEpochӶٵģ177ıƽС趨0.01.ͼ2(b)ѧϰ֮ͬĹϵ.ͼ2(c)˳̿ϵͳгͶˮCODMn\,ˮCODMn֮Ĺϵ.
ó̿ϵͳԤģͿԼˮCODMnֵһʱˮCODMnͶ֮ĶӦϵߣźҪʵ.Ƿֱѡȡ6CODMnֵԤģͣͼ3.
1ͬͶµCODMnȥ
Table 1 The removal rate of CODMn with various ozone dosages

Ͷ
mg/LCODMn,mg/LCODMnȥʣ%
ԭˮˮˮˮˮ
1.004.494.221.936.0157.02
2.024.063.941.782.9656.16
3.034.844.211.6113.0266.74
3.724.274.151.652.8161.35
4.064.964.842.022.4259.27
5.014.564.521.980.8756.58
5.694.163.681.8414.8155.77
7.284.323.681.8414.8157.41
9.294.213.011.5428.5063.42
10.284.563.801.9616.6757.02
15.624.363.921.4410.0966.97


ͼ2ϵͳˮCODMnͶˮCODMnάͼ
Fig.2 The three-dimensional diagram of influent CODMn,ozone dosages and effluent CODMn

ͼ3̿ϵͳˮCODMnһʱͶԳˮCODMnӰ
(ͼнˮCODMn(mg/L)ֱΪ(a) 4.06(b) 4.24 (c) 4.42 (d) 4.60 (e) 4.78 (f) 4.86)
Fig.3 The influence of ozone dosages on effluent CODMn when influent CODMn is stable 
4ʵ
ʵ飬ǵó½ۣ
BP˹Ԫĳ̿ϵͳģ׼ȷϵͳӰصĹϵ(1) ͶԳˮCODMnֵӰˮCODMnֵĲͬͬģӳͶһʹˮCODMnֵͣʱʹ.(2) Գ̿ϵͳнˮCODMnͶˮCODMn֮ϵԤģͣñȽϾõĳͶ.ͼ3ʾˮCODMnֵΪ4.24ʱõĳͶΪ3 mg/L,ʱ̿ϵͳˮCODMnֵΪ1.75 mg/LͶ10 mg/LʱȻʹϵͳˮCODMnֵ1.5 mg/L,ɻʹͶʹ󣬲.(3) ͶˮCODMnֵĹϵڵͳͶʱֻԭˮкһлӳͶʹˮCODMnֵ٣ųͶӣڳὫԭˮеĴлֽΪСмнϸߵֵҲױ̿ʹˮCODMnֵ.
ܽɵĻBP˹Ԫ̿ϵͳʵ飬ԿBPҪȡĲ(㻹Ƕ)㡢м㡢ĽڵĿѧϰٶȼʼȨֵ.ڵȽȷֱĸڱηʵΪͶͽˮCODMnֵΪˮCODMnֵڵֱΪ21мڵѧϰٶȼʼȨֵɶȱȽϴĲĿǰͨȷĽʽȷҪͨߵľʵѡ.ˣ֤˹ԪˮϵͳģЧԵĻϣһˮϵͳص㣬ĽͳBPĽṹ㷨ܹݲͬıƽѵݣŻƣԱʶŶȽ۷⽫ǽо.
5
1.о˹Ԫ뷽뵽ˮϵͳоУʹˮоܻͿƻǶԻѧһؿ.
2.˹ԪеķֲʽϢ洢д֯ѧϰص㣬˻˹ԪˮϵͳԤģͣ˷ôͳģĲ㣺(a) ø÷ͨϵͳݼɽΪ׼ȷģͣģ͵ͨԼʱЧֻȡʵݵĹ㷺׼ȷ(b) ø÷ģͶڲͬˮ׼ȷԤ⣬ܹΪԱṩƲΪʵˮ߿ṩһе;.
3.BP˹Ԫ罨ˮϵͳģ;õķˮϵͳģͱʶ.
* ʡȻѧص֧Ŀ
ߵλܡҵѧӦûѧϵ150001
   ѧ뻷ϵ150001
ο
[1]Ҷ.ˮѧģ͵оչ.ѧչ19931(1)7478
[2]Hopfield J J,Tank D W.Neural computation of decision in optimization problems.Bio Cybern,1985,(52):9297
[3] Hush D R,Horne B G.Progress in supervised neural networks:what's new since lippmann.IEEE Signal Progress Magazine,1993,10(1):839
[4]Wang B Z.The efficacy and mechanismof removal of organic substances from water by ozone and activated cabon.Water Science and Technology,1989,(1):6367
1997-11-28յԭ塡1998-03-25յ޸ĸ
