GE Data Analysis Report

1
GE Data Analysis Report

• Intelligent CAM System Laboratory University of
  Cincinnati
• June 2003

2
Dataset Parameters

• LM6000PD DF Parameters 091102
• No column data for all PT_TYPE BOOL data
  Missing or Neglected (PT_TYPE INT data) ?
  Question?
• Field name of dataset is _VAL0.
• Some fields whose PT_TYPE INT are actually BOOL
  type, such as AIMULT.
• Many fields just have one value for the whole
  dataset (Constant).
• E_100102_120102 (166 columns will be remained 72
  remained, 26 constant fields discarded 156, 60
  and 68, 9).
• K_041503_051503 (129 columns will be remained 61
  remained, 37 constant fields discarded and 68,
  7).
• L_041503_051503
• LM6000PC Parameter List 013003
• I_080102_093002 ? Further work
• J_090102_102602 ? Further work

3
Undefined Parameters

• ? Question?
• DPS3STM_VAL0, DPS3WTR_VAL0, DT3STM_VAL0,
  DT3WTR_VAL0
• FT62231_VAL0, FIRE_STRTS
• GMVSEL_VAL0, GMVVLVDMD_VAL0
• H2ODMD_VAL0
• LMVSEL_VAL0, LMVVLVDMD_VAL0
• N2SEL_VAL0, NMVSEL_VAL0, NMVVLVDMD_VAL0
• OP_HRSL_VAL0
• PGS_VAL0, PGSSEL_VAL0, PS3PS3SCH_VAL0,
  PT62239_VAL0, PT62241_VAL0, PT62269_VAL0,
  PTBSEL_VAL0
• SPRINTHRS_VAL0, STARTSL_VAL0
• T25CHN_VAL0, T25SYN_VAL0, TE62235_VAL0,
  TE62243_VAL0, TRIPSL_VAL0
• WFLBOJ_VAL0, WFPLTDMDGS_VAL0, WST36S_VAL0,
  WST36W_VAL0
• Z_LHVGAS_VAL0

4
Data Gaps

• Places where data change wildly ? Could be
  failures
• E dataset
• Has small variance between 22400-37800
• E.g., DPS3DTA_VAL0, DPS3DTB_VAL0, and
  DPS3DTSEL_VAL0
• Changes to small or high values between
  12700-13400, 15100-16000, and 22600-39200
• E.g., CDPAVAL_VAL0 and CDPBVAL_VAL0
• K dataset
• Has small variance between 4700-5500
• Changes to small or high values between 4800-5600
  and 14000-15600
• L dataset
• Changes to small or high values between
  41950-43100
• E.g., PTBA_VAL0, PTBB_VAL0, PTBDMD_VAL0,
  PTBDRES_VAL0, and PTBSEL_VAL0 T25A_VAL0,
  T25B_VAL0, T25CHN_VAL0, T25EST_VAL0, and
  T25SEL_VAL0
• I dataset
• Peaks around position such as 20700, 21400,
  25000, 34550, 41700, 45550, and 48350
• E.g., CRFBB_VAL0, CRFNHPVIB_VAL0, CRFNLPVIB_VAL0,
  DE_XVIB_VAL0, and DE_YVIB_VAL0
• J dataset
• Has small variance (or all are zero values)
  between 1887-1983 and 3769 to 20800
• E.g., TLIQMAN_PRI_VAL0, TLIQMAN_SEC_VAL0,
  TLQA_VAL0, TLQB_VAL0, TLQSEL_VAL0, TLUBA_VAL0,
  and TLUBB_VAL0

5
Hypothesis Testing

• Remove gap data
• T-test
• Same engine and same parameter
• Check if their means have significant difference
• F-test
• Same engine and same parameter
• Check if their variances have significant
  difference
• I vs. J
• E, K vs. L

6
I vs. J

• Assume that segment B and D of dataset I are due
  to failures and there was no failure in dataset I
• Three tests between I and J
• 1st I vs. J.
• 2nd I vs. A in J
• 3rd I vs. A and C from J.
• Conclusions
• If I and J differ significantly in second but not
  in third test, then the failure in segment B
  could be related to these parameters.
• If I and J do not differ significantly in second
  but do so in the third test, then it could be
  that the failure in segment D could be because of
  these parameters.

7
E, K vs. L

• T-test
• F-test

8
J Dataset

• Data splits (4 segments)
• 1 to 1886 (values specified) 1887 to 1983 (all
  zeros) 1984 to 3768 (values specified) and 3769
  to 20800(all zeros).
• T-test
• F-Test