18
Fault Diagnosis

Fault Diagnosis

  • Upload
    feo

  • View
    155

  • Download
    8

Embed Size (px)

DESCRIPTION

Fault Diagnosis. The Diagnosis Problem. General: Given the symptoms, determine the likely cause(s) Applies to many domains: Human illness, trouble-shooting of any kind - PowerPoint PPT Presentation

Citation preview

Page 1: Fault Diagnosis

Fault Diagnosis

Page 2: Fault Diagnosis

SAT-Based Diagnosis 2

The Diagnosis ProblemGeneral: Given the symptoms, determine the likely cause(s)

Applies to many domains: Human illness, trouble-shooting of any kind

In testing: Given test results on a failing device, {<failing test, failing outputs>}, find the likely defects/faults as the causes.

Often a preprocess to failure-mode analysis (FMA), an expensive process.

Page 3: Fault Diagnosis

SAT-Based Diagnosis 3

Model-based vs. Model-free Diagnosis?

Model Based: Assumes the error-causing defect can be modeled by one or more of the standard fault models: SAF, Bridging, Opens, ...Model-Free: Assumes the defect can excite any number of lines in any time frame, some of whom are observed in the same or later time frames according to the test results.

Both, however, must limit the multiplicity of causes for practical reasons.

Page 4: Fault Diagnosis

SAT-Based Diagnosis 4

Cause-effect vs. Effect-cause Analysis?

Cause-effect: For the given test, pre-compute the effects of all possible causes, according to an assumed fault model, and build a dictionary with effects as index and causes as contentsDiagnosis reduces to simple dictionary lookup

Effect-causeAnalyze the circuit elements connected to the outputs for observed test results and determine likely causes at the inputs of the circuit elements.Repeat the process recursively until all possible causes are determined.

Page 5: Fault Diagnosis

SAT-Based Diagnosis 5

Motivating ExampleConsider test results from c17:

Test Vector(A,B,C,D,E)

Failing Output(s)

V1: <1,1,1,0,0>

Z

V2: <0,1,1,0,1>

Y

We will work through the example in class.

Page 6: Fault Diagnosis

SAT-Based Diagnosis 6

Motivating Example (Contd.)

Now consider the results of a third test vector applied to c17:

Test Vector(A,B,C,D,E)

Failing Output(s)

V1: <1,1,1,0,0>

Z

V2: <0,1,1,0,1>

Y

V3: <1,0,1,1,1>

Z

Page 7: Fault Diagnosis

Fault Diagnosis Using Boolean Satisfiability*

* A. Smith, A. Veneris, M. F. Ali, and A. Viglas, “Fault Diagnosis and Logic Debugging Using Boolean Satisfiability,” IEEE TCAD, 24(10), Oct 2005, 1606-1621.

Page 8: Fault Diagnosis

SAT-Based Diagnosis 8

Modeling Candidate Fault Location

Page 9: Fault Diagnosis

SAT-Based Diagnosis 9

Counter for Multiple Faults

Page 10: Fault Diagnosis

SAT-Based Diagnosis 10

Complete Construction

Page 11: Fault Diagnosis

SAT-Based Diagnosis 11

Example Sequential Circuit and its ILA Representation

Page 12: Fault Diagnosis

SAT-Based Diagnosis 12

Diagnosis in Two Cycles

Page 13: Fault Diagnosis

SAT-Based Diagnosis 13

Sequential SAT Based Diagnosis-1

Page 14: Fault Diagnosis

SAT-Based Diagnosis 14

Sequential SAT Based Diagnosis-2

Page 15: Fault Diagnosis

SAT-Based Diagnosis 15

Two-Pass Diagnosis Using Structure Information

Page 16: Fault Diagnosis

SAT-Based Diagnosis 16

Model-Based Diagnosis - SAFs

Page 17: Fault Diagnosis

SAT-Based Diagnosis 17

Experimental Results - 1

Page 18: Fault Diagnosis

SAT-Based Diagnosis 18

Experimental Results - 2