Signal processing method for an electromagnetic induction test

• US 4,823,082 A
• Filed: 02/11/1987
• Issued: 04/18/1989
• Est. Priority Date: 02/18/1986
• Status: Expired due to Fees

• Current Assignee: Klaus Schmidt
• Original Assignee: Klaus Schmidt
• Inventors: Iwasaki, Masayoshi, Nasu, Shoji, Nishimoto, Yoshiro, Yasukuni, Hiroaki
• Primary Examiner(s): Strecker, Gerard R.
• Application Number:
  US07/013,424
• Time in Patent Office: 797 Days
• Field of Search: 324/225, 324/227, 324/232, 324/233, 324/238, 324/240-243
• US Class Current: 324/232
• CPC Class Codes: G01N 27/9046 by analysing electrical signals

Abstract

A signal processing method in an electromagnetic induction test. A test object is subjected to an electromagnetic induction by use of N (N≧1) kinds of test frequencies to obtain M (M≦2N) kinds of phase detection outputs corresponding to respective states of the test object. The phase detection outputs are then sampled at a plurality of measuring points so that there exists n (n<M) kinds of disturbance factors of which a portion contribute to the electromagnetic induction test. The phase detection outputs have a variance greater than a variance of a target parameter of the test object and contribute to the electromagnetic induction test. A first coordinate axis Z, is determined in a direction in which a variance of a distribution of the group of the measuring points comes to be a maximum in an M-dimensional space formed with the coordinate axes associated with the M kinds of the phase detection outputs. The remaining coordinate axes are likewise determined. Based on the coordinate of each of the measuring points viewed from the coordinate axis Z_n+1, a measuring is effected to determine a presence or an absence of a defect, a physical quantity, and a geometric quantity on a surface of the test object.

Claims

1. A signal processing method in an electromagnetic induction test for detecting defects, physical quantities and geometric quantities in a surface of a test object in which said test object is subjected to an electromagnetic induction by use of N(N≧

1) kinds of test frequencies to obtain M(M≦

2N) kinds of phase detection outputs corresponding to respective states of the test object said method comprising the steps of;

sampling the phase detection outputs at a plurality of measuring points, so that when there exists n(n<

M) kinds of disturbance factors of which a portion contribute to the electromagnetic induction test, said phase detection outputs have a variance due to said defects, physical quantities and geometric quantities greater than a variance of a target parameter of the test object and contribute to the electromagnetic induction test;

determining a first coordinate axis Z₁ in a direction in which said variance of said phase detection outputs approaches a maximum in an M-dimensional space formed from coordinate axes associated with the M kinds of the phase detection outputs;

determining a second coordinate axis Z₂ in a direction in which said variance of said phase detection outputs approaches a maximum in a space orthogonal to the coordinate axis Z₁ of the multi-dimensional space;

similarly determining a third coordinate axis Z₃, a fourth coordinate axis Z₄, . . . , and an M-th coordinate axis Z_m in directions in which said variance of said phase detection outputs approaches a maximum in a space orthogonal to the previously named axis of the multi-dimensional space; and

determining a presence or an absence of a defect, a physical quantity and a geometric quantity on a surface or in the vicinity of the test object related to the coordinate of the Z_n+1 axis for each measuring point.

2. A signal processing method according to claim 1 wherein said coordinate axes Z₁, Z₂, Z₃, Z₄, and so on are set such that a variance/covariance matrix is obtained, said matrix being configured with phase detection outputs corresponding to said measuring points sampled, and said coordinate axes Z₁, Z₂, Z₃, Z₄, and so on are set so as to correspond to eigenvectors associated with eigenvalues λ

₁, λ

₂, λ

₃, λ

₄ of the variance/covariance matrix.

3. A signal processing method according to claim 1 wherein said coordinate axes Z₁, Z₂, Z₃, Z₄, and so on are set such that a variance of a distribution of the measuring points is sequentially obtained in all directions of said multidimensional space and a coordinate axis is set in a direction in which the variance becomes to be maximum.

4. A signal processing method according to claim 1, 2, or 3 wherein a defect is detected on a surface or in the neighborhood of a surface of a steel piece produced through a continuous casting by use of two kinds of frequencies including a low frequency and a high frequency.

5. A signal processing method according to claim 1, 2, or 3 wherein a defect is detected on a surface or in the neighborhood of a surface of a steel piece produced through a continuous casting by use of a kind of frequency.