JP5896434B1 - Inspection device - Google Patents

Abstract

For each specification of a device that generates sound or vibration, an optimum inspection filter is determined from a plurality of filters and an inspection method, an inspection device, and a filter to be used for the inspection method can be determined. A filter determination method, a program, and a filter determination device are provided. Furthermore, the present invention provides an inspection method and an inspection apparatus capable of inspecting a device that generates sound or vibration whose frequency changes with the passage of time using a filter whose characteristics can be changed with the passage of time. A signal relating to a sound or vibration generated by the device to be inspected and whose frequency changes over time is processed using a filter whose characteristics can be changed over time to inspect the device.

Description

  The present application relates to an inspection method for inspecting a device that generates sound or vibration, an inspection device, a filter determination method for determining a filter used in the inspection method, a program, and a filter determination device.

  There is a technique for detecting an abnormality of a device from sound or vibration generated by the device. For example, Patent Document 1 describes a detection device that detects a sticking sound generated by a speaker and determines whether the speaker is good or bad. The detection device according to Patent Literature 1 detects a cluttering sound by extracting a signal component of the cluttering sound from a signal based on the sound generated by the speaker using a filter optimized in advance.

JP-A-6-315197

However, Patent Document 1 does not describe any specific description regarding “pre-optimized filter”.
The sound or vibration generated by an abnormal device varies depending on the specifications of the device. Therefore, in order to accurately determine the quality of various devices, a filter optimized for each specification is required. However, it is practically difficult to optimize the filter in advance for each device specification.

The present application has been made in view of such circumstances. The purpose of the present application is to use an inspection method, an inspection apparatus, and an inspection method capable of inspecting the apparatus by determining an optimum inspection filter from a plurality of filters for each specification of the apparatus that generates sound or vibration. It is an object to provide a filter determination method, a program, and a filter determination device that can determine a filter.
Furthermore, an object of the present application is to provide an inspection method and an inspection apparatus capable of inspecting a device that generates sound or vibration whose frequency changes with time, using a filter whose characteristics can be changed with time. is there.

In the inspection method according to the present application, prior to inspecting a device to be inspected, a signal relating to sound or vibration generated by one or a plurality of non-defective products and one or a plurality of defective devices prepared in advance is provided for each of a plurality of filters. Based on the plurality of signals processed in step 1, a filter for inspection whose characteristics can be changed over time is determined, and a signal related to sound or vibration generated by the device to be inspected and whose frequency changes over time. Is processed using the determined filter for inspection, and the quality of the device to be inspected is determined based on the processed signal.

  The inspection method according to the present application calculates a level difference between signals related to sound or vibration generated by the non-defective product and the defective product, processes a signal corresponding to the calculated level difference with each of a plurality of filters, and each of the plurality of filters. Among the level differences corresponding to the signals processed in step 1, the filter corresponding to the maximum level difference is determined as the inspection filter.

  In the inspection method according to the present application, a level difference of a signal related to sound or vibration generated by each of the non-defective product and the plurality of defective products is calculated for each of the plurality of defective products, and a signal corresponding to the calculated level difference is a plurality of filters. The maximum level difference corresponding to the processed signal is calculated for each of the plurality of defective products, and among each maximum value of the level difference calculated for each of the plurality of defective products for each of the plurality of filters. The minimum value of the level difference is selected, and the filter corresponding to the maximum level difference among the minimum values of the level difference selected for each of the plurality of filters is determined as the inspection filter.

  The inspection method according to the present application is based on a signal based on time, frequency, and level information respectively calculated from a plurality of signals related to non-defective products prepared in advance, and the maximum for each time and frequency defined at a predetermined interval. Each level is extracted, each extracted maximum level is set to the level of the signal related to the sound or vibration generated by the non-defective product, and the maximum value of the level difference is determined based on the set level of the signal related to the non-defective product. It is calculated for each of a plurality of defective products.

  In the inspection method according to the present application, when the selected minimum value of the level difference exceeds a predetermined value, the selection of the minimum value of the level difference is finished, and the filter corresponding to the minimum value of the level difference exceeding the predetermined value Is determined as a filter for inspection.

The inspection apparatus according to the present application is based on a filter output unit that outputs a plurality of filters and a signal relating to sound or vibration generated by one or more non-defective products and one or more defective products prepared in advance. An apparatus to be inspected uses a determination unit that determines an inspection filter whose characteristics can be changed over time from a plurality of filters output by the filter output unit, and an inspection filter determined by the determination unit. A processing unit that processes a signal related to a sound or vibration that occurs with a frequency that changes over time , and a determination unit that determines the quality of the inspection target device based on the signal processed by the processing unit. It is characterized by.

  The inspection apparatus according to the present application calculates a level difference between a non-defective signal storage unit that stores a signal related to the non-defective product, a signal related to the non-defective product stored in the non-defective product signal storage unit, and a signal related to the defective product prepared in advance. A level difference calculation unit; and a signal processing unit that processes a signal corresponding to the level difference calculated by the level difference calculation unit in each of the plurality of filters output by the filter output unit, and the determination unit includes the signal Among the level differences corresponding to the signals processed by the processing unit, the filter corresponding to the maximum level difference is determined as the inspection filter, and the signal for comparing the signal processed by the processing unit and a predetermined signal A comparison unit is provided, and the determination unit is configured to determine the quality of the device to be inspected based on the result of comparison by the signal comparison unit.

  In the inspection apparatus according to the present application, the level difference calculation unit calculates a level difference of a signal related to sound or vibration generated by the non-defective product and the plurality of defective products for each of the plurality of defective products, The signal processing unit is configured to process a signal corresponding to the level difference calculated for each of the plurality of defective products by the plurality of filters output from the filter output unit, and the signal processing unit For each of the plurality of defective products, and for each of the plurality of filters output by the filter output unit, the calculation unit is provided for each of the plurality of defective products. A selection unit that selects a minimum value of the level difference among the calculated maximum values of the level difference, and the determination unit selects a maximum value among the minimum values of the level difference selected by the selection unit for each of the plurality of filters. The level It characterized that you have a filter corresponding to the difference to determine a filter for inspection.

  In the inspection apparatus according to the present application, when the minimum value of the selected level difference exceeds a predetermined value, the selection unit ends selection of the minimum value of the level difference, and the determination unit The filter corresponding to the minimum value of the level difference exceeding the value is determined as the inspection filter.

  The filter determination method according to the present application determines a test filter from a plurality of filters whose characteristics can be changed over time in order to inspect a device to be inspected that generates sound or vibration whose frequency changes over time. A method for determining a filter, wherein a level difference of a signal relating to sound or vibration generated by a non-defective product and one or a plurality of defective products prepared in advance is calculated for each of the plurality of defective products, and the calculated level difference is calculated. A corresponding signal is processed by each of a plurality of filters, a maximum level difference corresponding to the processed signal is calculated for each of the one or more defective products, and each of the one or more defective products is calculated for each of the plurality of filters. The minimum value of the level difference is selected from the maximum values of the level difference calculated in the above, and the filter corresponding to the maximum level difference among the minimum values of the level difference selected for each of the plurality of filters is selected. And determining a filter for checking the filter.

  The program according to the present application is a program that causes a computer to execute a process of determining one filter from a plurality of filters whose characteristics can be changed over time based on a signal related to sound or vibration whose frequency changes over time. Calculating a level difference between one signal related to sound or vibration and one or a plurality of other signals for each of the one or a plurality of other signals, and changing a filter capable of changing the characteristics over time, Each of the plurality of changed filters processes a signal corresponding to the calculated level difference, calculates a maximum value of the level difference corresponding to the processed signal for each of the one or more other signals, and each of the plurality of filters. The minimum level difference value is selected from the maximum level difference values calculated for each of the one or more other signals, and the maximum level difference value selected for each of the plurality of filters is selected. Among the values, characterized in that to execute a process of determining the filter corresponding to the maximum level difference between one filter to the computer.

  The filter determination device according to the present application is an inspection device for inspecting a device to be inspected that generates a sound or vibration whose frequency changes over time, from a plurality of filters whose characteristics can be changed over time. A filter output unit that outputs a plurality of filters, and a level difference between signals relating to sound or vibration generated by a non-defective product and one or more defective products prepared in advance. A level difference calculation unit that calculates for each of a plurality of defective products, a signal processing unit that processes a signal corresponding to the level difference calculated by the level difference calculation unit, and a plurality of filters that are output by the filter output unit; A calculation unit that calculates a maximum level difference corresponding to the signal processed by the signal processing unit for each of the one or a plurality of defective products, and a plurality of filters output by the filter output unit; A selection unit that selects a minimum level difference among the maximum values of the level difference calculated by the calculation unit for each one or a plurality of defective products, and the selection unit selects each of the plurality of filters. A determination unit is provided that determines a filter corresponding to the maximum level difference among the minimum values of the level differences, as an inspection filter.

  According to the inspection method, inspection apparatus, filter determination method, program, and filter determination apparatus according to the present invention, the optimal inspection filter is determined from a plurality of filters for each specification of the apparatus that generates sound or vibration. The device can be inspected. Furthermore, according to the inspection method and the inspection apparatus according to the present invention, since the filter whose characteristics can be changed with the passage of time is used, a device that generates a sound or vibration whose frequency changes with the passage of time is applied to the change of the sound or vibration. Can be inspected.

It is a block diagram which shows the hardware structural example of an inspection system provided with the inspection apparatus which concerns on this application. It is explanatory drawing which shows the outline | summary of operation | movement of a test | inspection system. It is a flowchart which shows an example of the outline | summary procedure of a filter determination process. It is a flowchart which shows an example of the procedure of an inspection process. It is explanatory drawing which shows an example of the filter from which the aspect of the change from which a cutoff frequency becomes high changes as time passes. It is explanatory drawing which shows an example of the filter which has a part from which a cutoff frequency becomes low with progress of time. FIG. 10 is an explanatory diagram illustrating an outline of operation of an inspection system according to a second embodiment. It is an example of the spectrogram calculated about a good quality speaker. It is explanatory drawing which shows the frequency distribution example of the level in a certain time. It is explanatory drawing which shows the frequency distribution example of the integrated value of the level in a certain time. It is an example of a three-dimensional graph of integrated value g (i) (t, f). It is explanatory drawing which shows notionally calculation of ratio over (ii, t, f). It is explanatory drawing which shows the curve which shows the time change of the cutoff frequency of the filter which discriminate | determines inferior goods from a non-defective product group on the graph of over (ii, t, f) regarding the ii-th inferior goods in FIG. It is sectional drawing which cut | disconnected FIG. 13 along the cut-off curve C. FIG. It is explanatory drawing which shows notionally the process which extracts defect detection capability value D (C) from detection value Over (ii, C) of n defective products. It is explanatory drawing which shows the example of a pattern. It is explanatory drawing which shows the example of a pattern. It is explanatory drawing which shows the example of a pattern. It is explanatory drawing which shows the image of the process which calculates | requires D (C) from a pattern. It is explanatory drawing which shows the restriction | limiting of the calculation range regarding a filter. It is explanatory drawing which shows the image of the process which changes the shape of b pattern. It is a flowchart which shows an example of the procedure of the process which changes each part of b pattern sequentially. It is explanatory drawing which shows three parts C0, C1, and C01 in b pattern. It is explanatory drawing which shows the difference value on a pattern. It is explanatory drawing which shows the process which calculates Over0 (ii, f, mi) corresponding to C0. It is explanatory drawing which shows the process which calculates Over1 (ii, f, mi) corresponding to C1. It is explanatory drawing which shows the image which sets the maximum value to Over (ii, C) from Over (ii, Cx) corresponding to C0, C1, and C01. It is explanatory drawing which shows the image of the process which changes the shape of c pattern. It is a flowchart which shows an example of the procedure of a filter determination process. It is a flowchart which shows an example of the procedure of a filter determination process.

  An inspection apparatus according to the present embodiment will be described with reference to the drawings. The inspection apparatus according to the present embodiment inspects an apparatus that generates sound or vibration, and performs pass / fail determination of the apparatus. Hereinafter, the inspection apparatus according to the present embodiment will be described using a speaker as an example of an apparatus to be inspected.

Embodiment 1
FIG. 1 is a block diagram illustrating a hardware configuration example of an inspection system including an inspection apparatus according to the present application. The inspection system includes a filter determination device 20, a measurement unit 30, a microphone 40, and an anechoic box 50. An inspection device for inspecting the speaker 60 to be inspected includes a filter determination device 20 and a measurement unit 30. The inspection apparatus uses the filter determination device 20 as a main part.

  The filter determination device 20 is a personal computer. The filter determination device 20 outputs data corresponding to a sine wave sweep signal changing from a low frequency to a high frequency to the measurement unit 30. The filter determination device 20 inputs a signal corresponding to the sound generated by the speaker 60 from the measurement unit 30. The filter determination device 20 executes various processes based on the signal input from the measurement unit 30.

  The filter determination apparatus 20 includes a CPU 21, ROM 22, RAM 23, hard disk 24, disk drive 25, display unit 26, operation unit 27, communication unit 28, and I / O unit 29.

  The CPU 21 is a processor that executes an instruction set described in the program 2P. The ROM 22 stores a BIOS and the like executed by the CPU 21 when the filter determination device 20 is activated.

  The RAM 23 is a main storage device. The hard disk 24 is an auxiliary storage device. The hard disk 24 stores a sound signal from the speaker 60 amplified by the measurement unit 30.

  The disk drive 25 reads information from the optical disk 2d of the external storage medium and records information on the optical disk 2d. The display unit 26 is a display device that displays an image. The operation unit 27 is an input device that accepts various inputs from the user. The communication unit 28 is an interface with the LAN 100. The I / O unit 29 is an interface with the measurement unit 30.

  Note that the CPU 21 may read the program 2P from a portable storage medium such as the optical disk 2d via the disk drive 25. The CPU 21 may read the program 2P from another information processing apparatus or storage device via the communication unit 28. Furthermore, a semiconductor memory 2m such as a flash memory storing the program 2P may be mounted in the filter determination device 20.

The measurement unit 30 inputs a sweep signal to the speaker 60.
The measurement unit 30 includes a D / A converter 31, a power amplifier 32, a measurement amplifier 33, and an A / D converter 34. The D / A converter 31 converts the digital sweep signal input from the filter determination device 20 into an analog signal, and outputs the converted analog signal to the power amplifier 32. The power amplifier 32 amplifies the analog signal converted from the D / A converter 31. The power amplifier 32 outputs a sweep signal to the speaker 60.

  The anechoic box 50 has high sound insulation and sound absorption characteristics. The upper wall of the anechoic box 50 facing the speaker 60 is provided with an opening for taking in sound generated by the speaker 60. The microphone 40 is installed inside the anechoic box 50 so as to face the speaker 60.

  The measurement amplifier 33 amplifies the level of the analog signal output from the microphone 40. The A / D converter 34 converts the signal amplified by the measurement amplifier 33 into a digital signal. The A / D converter 34 outputs the converted digital signal to the filter determination device 20.

  FIG. 2 is an explanatory diagram showing an outline of the operation of the inspection system. An auditory test is performed on a plurality of speakers 60 in advance, and one good product and a plurality of defective products are prepared as samples to be processed for determining a filter for inspection. One prepared speaker 60 is attached to the anechoic box 50. The filter determination apparatus 20 inputs data corresponding to a sweep signal of a sine wave that changes at a constant speed from a low frequency (for example, 10 Hz) to a high frequency (for example, 20 kHz) to the D / A converter 31 of the measurement unit 30. The sweep time is, for example, 1.0 second. The speaker 60 generates a sound corresponding to the sweep signal corresponding to the data input from the filter determination device 20 via the D / A converter 31 and the power amplifier 32.

  The microphone 40 receives the sound generated by the sample speaker 60. The microphone 40 outputs a signal to the filter determination device 20 via the measurement amplifier 33 and the A / D converter 34. The CPU 21 inputs a signal from the measurement unit 30. The CPU 21 stores the input signal in the hard disk 24 in association with the non-defective / defective product identification information related to the sample speaker 60. The inspection system repeats the same storage process for all the prepared speakers 60 for the sample.

  The CPU 21 reads the signal stored in the hard disk 24 together with the non-defective / defective product identification information related to the sample speaker 60 corresponding to the signal. The CPU 21 calculates a three-dimensional spectrogram of time, frequency, and level from the read signal, and integrates the level at the time from the maximum frequency to the frequency for each frequency at each time. FIG. 8 shows an example of a spectrogram, FIG. 9 shows a cross-section of a spectrogram at a certain time, that is, an example of a frequency distribution of levels, and FIG. 10 shows an accumulation of levels from the maximum frequency to that frequency at that time. . The integrated graph at all times is as shown in FIG. Details of FIGS. 8 to 11 will be described in a second embodiment. The CPU 21 stores the calculated integrated graph in the hard disk 24 along with the good / bad classification. The CPU 21 repeats this integrated graph calculation process for all the signals of the sample speaker 60 stored in the hard disk 24.

In general, a defective speaker 60 generates an abnormal sound mainly at a high frequency with respect to an input signal. Abnormal noise is often lower in level than the fundamental, second, third, etc. harmonics of the input signal, so it is necessary to remove lower harmonics. A high-pass filter that passes through is used. A level obtained by passing a signal of a speaker 60 for a sample through a high-pass filter having a frequency at a certain time as a cutoff frequency becomes a level at the frequency of the integrated graph of the speaker 60 at the time.
In the present invention, the CPU 21 determines the filter for processing the sound signal so that the level difference between the non-defective product and the defective product becomes large in order to clearly distinguish the speaker 60 to be inspected from the good product and the defective product. .

  The CPU 21 dynamically changes the cutoff frequency of the high-pass filter in order to make the cutoff frequency of the high-pass filter follow the signal from the speaker 60 corresponding to the sweep signal. That is, the filter used by the CPU 21 can be said to be a dynamic filter 1f. The dynamic filter 1f is represented by a time change of the cutoff frequency. In FIG. 2, the horizontal axis of the graph showing the dynamic filter 1f is time, and the vertical axis is frequency. That is, the dynamic filter 1f can be expressed by a two-dimensional curve having time and frequency as elements. Further, the level graph obtained as a result of processing the signal of the speaker 60 for a sample by the dynamic filter 1f is a cut surface obtained by cutting the integrated graph of the speaker 60 by the curve of the dynamic filter 1f.

  The CPU 21 uses the dynamic filter 1f to process components that are common to both good and defective samples. CPU21 determines the filter for test | inspecting the speaker 60 to be test | inspected based on the level difference after the process concerning the quality goods for samples, and the inferior goods. The CPU 21 inspects the speaker 60 to be inspected using the determined dynamic filter 1f. Since the dynamic filter 1f is a digital filter created by the CPU 21, the filter characteristics of the dynamic filter 1f are determined by filter parameters.

A filter determination process for inspecting the speaker 60 to be inspected will be described.
If any characteristic change filter is allowed, it is sufficient to create a filter with a curve that passes through all the points of the maximum value of the difference graph of each defective product. It is difficult to realize such a filter. The shape of the filter needs to be limited to a certain pattern that can be used in an actual inspection line. A predetermined pattern of filters is stored in a plurality of hard disks 24. In the pattern in this embodiment, the cut-off frequency continuously changes from a low frequency to a high frequency as time elapses. In addition, the pattern includes a plurality of patterns that change following the frequency change of the fundamental wave or a plurality of harmonics from a low order to a high order.
Note that the CPU 21 may generate a plurality of constant pattern filters in real time, and output each of the generated constant pattern filters for the following processing.

  The CPU 21 selects one filter from the prepared filters. CPU21 acquires the cut surface cut | disconnected with the filter curve with respect to the integration graph of the quality goods for a sample with the selected filter. The cut surface becomes a level graph obtained by filtering non-defective products with the filter.

CPU21 acquires the cut surface which cut | disconnected the integration graph of the some inferior goods for samples with the selected filter with the filter curve. The cut surface becomes a level graph in which the defective product is processed by the filter. FIG. 2 schematically shows level graphs 1g and 2g of the non-defective product and the defective product after processing, with the horizontal axis representing time and the vertical axis representing level.
The CPU 21 may store a non-defective level graph 1g that has been subjected to the filter processing in the hard disk 24.

  The CPU 21 calculates the maximum value of the difference graph between the level graph of the processed non-defective product and the level graph of one defective product as the detected value of the defective product by the filter. The CPU 21 stores the calculated detection value in the hard disk 24 in association with the identification information of the defective product. The CPU 21 repeats the process of calculating the detection value for all defective samples. FIG. 2 shows a list 200 of detection values stored in the hard disk 24. The larger the detected value, the farther the defective product is from the non-defective product.

  There is a variation in sound output characteristics among a plurality of defective products. The filter for discriminating whether the speaker 60 to be inspected is a non-defective product or a defective product needs to be a filter in which all defective products prepared as samples are determined as defective products. For that purpose, it is preferable from the viewpoint of safety to adopt a level difference corresponding to a defective product having the sound output characteristic closest to the non-defective product.

  Therefore, the CPU 21 extracts the minimum value as the defect detection capability value of the filter from the detection values calculated for all the defective products in the list 200. The CPU 21 stores the extracted defect detection capability value in the hard disk 24 as a list 300 in association with the selected filter. The CPU 21 repeatedly stores the defect detection capability value in the list 300 for each filter.

  When the CPU 21 has completely calculated the defect detection capability value using all of the filters prepared in advance, the CPU 21 selects the filter corresponding to the maximum value from the plurality of defect detection capability values stored as the list 300 in the hard disk 24 for inspection. Determine the filter. Alternatively, when the CPU 21 changes the filter and finds a filter having a defect detection capability value greater than or equal to a defect detection capability value that the user thinks is sufficient for separating a non-defective product and a defective product, for example, May be determined as a filter for inspection. As a result, the processing time for determining the inspection filter can be shortened.

FIG. 3 is a flowchart illustrating an example of an outline procedure of the filter determination process. In the procedure of FIG. 3, it is assumed that the calculation of the integrated graph relating to a plurality of defective products has been completed in advance.
The CPU 21 selects an unprocessed filter (step S101). The CPU 21 selects an unprocessed defective product integration graph (step S102). The CPU 21 processes the integrated graph relating to the non-defective product and the defective product with the selected filter (step S103). The CPU 21 calculates the maximum value of the difference graph between the processed non-defective product level graph and the defective product level graph as the detected value of the defective product by the filter (step S104).

  The CPU 21 stores the calculated detection value in association with the defective product identification information in the list 200 in the hard disk 24 (step S105). CPU21 determines whether the process was complete | finished about the integration graph of all the inferior goods (step S106). CPU21 returns a process to step S102, when it determines with the process not being complete | finished about the integration graph of all the inferior goods (step S106: NO). When the CPU 21 determines that the processing has been completed for the integration graph of all defective products (step S106: YES), the CPU 21 determines the minimum value from the plurality of detection values stored in the list 200 in the hard disk 24 as the defect detection capability value of the filter. (Step S107). The CPU 21 stores the extracted defect detection capability value in the list 300 in the hard disk 24 in association with the selected filter (step S108).

  The CPU 21 determines whether or not processing has been completed for all filters (step S109). If the CPU 21 determines that the processing has not been completed for all the filters (step S109: NO), the CPU 21 returns the processing to step S101. If the CPU 21 determines that the process has been completed for all filters (step S109: YES), the CPU 21 extracts the maximum value from the defect detection capability values stored in the list 300 in the hard disk 24 (step S110). The CPU 21 determines a filter corresponding to the extracted maximum value as an inspection filter (step S111), and ends the process.

Next, processing for inspecting the speaker 60 to be inspected using the determined filter will be described.
The inspector attaches the speaker 60 to be inspected to the upper wall of the anechoic box 50. The inspector operates the filter determination device 20 to generate a sound corresponding to the sweep signal on the speaker 60 to be inspected. The microphone 40 receives sound from the speaker 60 to be inspected, and outputs the received sound signal to the filter determination device 20 via the measurement unit 30. The filter determination device 20 inspects the speaker 60 to be inspected based on the input signal and the inspection filter. The filter determination apparatus 20 displays the result of inspecting the inspection target speaker 60 on the display unit 26, for example. The inspector removes the speaker 60 to be inspected from the upper wall of the anechoic box 50. The inspector sorts the speaker 60 to be inspected into a non-defective product / defective product according to the result displayed on the display unit 26. Thereafter, the inspector repeats the above work.
Of course, the work performed by the inspector may be automated.

FIG. 4 is a flowchart illustrating an example of the procedure of the inspection process.
The CPU 21 outputs data corresponding to the sweep signal to the measurement unit 30 (step S1). The CPU 21 inputs a sound signal generated by the speaker 60 to be inspected from the measurement unit 30 from the measurement unit 30 (step S2). The CPU 21 filters the input signal with the inspection filter determined in the filter determination process (step S3). The CPU 21 compares the filtered signal with inspection reference data prepared in advance (step S4). The CPU 21 determines whether or not the speaker 60 to be inspected is a non-defective product based on the comparison result (step S5).

  CPU21 outputs a determination result to the display part 26 (step S6). CPU21 determines whether the instruction | indication of whether to repeat said test | inspection process was received from the operation part 27 (step S7). CPU21 returns a process to step S1, when it determines with the instruction | indication which repeats said test | inspection process having been received from the operation part 27 (step S7: YES). CPU21 complete | finishes a process, when the instruction | indication which repeats said test | inspection process is not received from the operation part 27 (step S7: NO). The case where the CPU 21 has not received an instruction to repeat the inspection process from the operation unit 27 is a case where the CPU 21 has received an instruction to stop the inspection process.

  In the present embodiment, the filter determination device 20 inputs a sweep signal that changes from a low frequency to a high frequency to the speaker 60. However, the filter determination device 20 may input a reverse sweep signal that changes from a high frequency to a low frequency to the speaker 60.

  In the present embodiment, the filter determination device 20 inputs a sweep signal to the speaker 60. However, the filter determination device 20 may input a step sine wave whose frequency is switched and a sine wave having a fixed frequency to the speaker 60. In such a case, the time change of the frequency may not be monotonously increasing or monotonically decreasing. Alternatively, the filter determination device 20 may input a synthesized wave having a plurality of frequencies to the speaker 60.

  The apparatus to be inspected according to the present embodiment is a speaker 60 that generates sound based on a signal input from the outside. However, the device to be inspected may be a device that itself generates sound or vibration whose frequency changes over time. The devices to be inspected are rotating members such as motors, gears and bearings, elevator cars, compressors, and the like.

The pattern prepared in advance for determining the filter for inspection may include a pattern in which the aspect of change in which the cutoff frequency increases changes with time.
FIG. 5 is an explanatory diagram showing an example of a filter in which the mode of change in which the cutoff frequency becomes higher changes with time. The time on the horizontal axis is a logarithm of milliseconds. The cut-off frequency first changes following the frequency change of the higher-order harmonics, and maintains a constant frequency from the middle. Thereafter, the cut-off frequency transitions to the lower-order harmonic side, but deviates from the frequency change of the harmonic. A plurality of filters having patterns as shown in FIG. 5 are prepared.

Patterns prepared in advance for determining a filter for inspection include not only the cutoff frequency that increases monotonously with time but also a portion that has a portion that changes to have a maximum value and a minimum value. But you can.
FIG. 6 is an explanatory diagram showing an example of a filter having a portion where the cutoff frequency becomes lower as time elapses. The time on the horizontal axis is a logarithm of milliseconds. In the case of the filter of FIG. 6, the cut-off frequency first changes following the frequency change of the higher order harmonics, becomes lower than the harmonic frequency from the middle, and changes so as to have a maximum value and a minimum value. . Therefore, the filter of FIG. 6 has a portion where the cutoff frequency becomes lower as time elapses. A plurality of filters having patterns as shown in FIG. 6 are prepared.

  In the present embodiment, the dynamic filter 1f employs a high-pass filter. However, the dynamic filter 1f may be a band pass filter, a comb filter, a low pass filter, or the like. The dynamic filter 1f may be a notch filter that cuts a specific frequency, a parametric equalizer that increases the level of a specific band, or the like. The output extracted by the dynamic filter 1f may be an envelope output, a crest fact output, or the like. The integration method for creating the integration graph varies depending on the type of filter.

Embodiment 2
In the second embodiment, after calculating the level difference between the signals related to the non-defective product and the defective speaker 60, the signal corresponding to the level difference is filtered by a plurality of dynamic filters 1f to determine the inspection filter. It relates to form.
In the second embodiment, the same reference numerals are assigned to the same components as those in the first embodiment, and detailed description thereof is omitted.

FIG. 7 is an explanatory diagram showing an outline of the operation of the inspection system according to the second embodiment.
A log sweep signal of a sine wave is input to a plurality of non-defective products and a plurality of defective products related to the sample speaker 60. A sine wave log sweep signal is a sine wave signal whose frequency changes exponentially and continuously over time. A signal related to the sound generated by the sample speaker 60 is input to the filter determination device 20. The CPU 21 stores the input signal in the hard disk 24.

  The CPU 21 reads the signal stored in the hard disk 24 together with the non-defective / defective product identification information related to the sample speaker 60 corresponding to the signal. The CPU 21 calculates a three-dimensional spectrogram of time, frequency, and level from the read signal, and integrates the level at the time from the maximum frequency to the frequency for each frequency at each time. FIG. 7 shows integrated graphs 1m and 2m calculated for non-defective products and defective products, with the horizontal axis representing time, the vertical axis representing frequency, and the height axis representing level. The CPU 21 stores the calculated integrated graphs 1m and 2m in the hard disk 24 together with the good / bad classification. The CPU 21 repeats this integrated graph calculation process for all the signals of the sample speaker 60 stored in the hard disk 24.

In the first embodiment, the integrated graph 1m in one non-defective speaker 60 selected by the hearing test is used as non-defective reference data. However, in the second embodiment, reference data representing a non-defective product is calculated from the integrated graph 1m of a plurality of non-defective speakers.
FIG. 8 is an example of a spectrogram 1 s calculated for a non-defective speaker 60. In the spectrogram 1s of FIG. 8, the horizontal axis indicates time t, the vertical axis indicates the frequency f, and the vertical axis with respect to the paper surface indicates the level. Note that the level in FIG. 8 is expressed by the difference in shading. The signal waveform input from the measurement unit 30 to the filter determination device 20 is shown at the bottom of FIG.
FIG. 9 is an explanatory diagram showing a frequency distribution example of the level at a certain point in the spectrogram. The horizontal axis in FIG. 9 is the level, and the vertical axis is the frequency f.

  When the index for distinguishing a plurality of non-defective products is i, the level at the point (t, f) at each time point / frequency distributed discretely on the spectrogram 1s is expressed by the following equation (1). .

Level P (i) (t, f) is data of a two-dimensional array.
In the spectrogram 1s of each good product, the CPU 21 calculates an integrated value of a level from the maximum frequency f _max to each of a plurality of frequency points at each time point, and creates an integrated graph. The integrated value of the level integrated up to a certain frequency point corresponds to the level of the output signal of the high-pass filter whose cutoff frequency is that frequency.
The integrated value g (i) (t, f) is expressed by the following equation (2).

FIG. 10 is an explanatory diagram showing a frequency distribution example of the integrated value of the level at a certain time point. The horizontal axis in FIG. 10 is the level, and the vertical axis is the frequency. In FIG. 10, the dotted line indicates the level, and the solid line indicates the integrated value of the level. That is, the black circle on the solid line in FIG. 10 indicates the integrated value of the level from the highest frequency f _max to a certain frequency point at a certain time. The integrated value g (i) (t, f) is two-dimensional array data.
FIG. 11 is an example of a three-dimensional graph of integrated values g (i) (t, f), that is, an integrated graph. The vertical axis in FIG. 11 is the integrated value, and the axis in the plane direction is time t and frequency f, respectively.

The maximum value is extracted for each time point / frequency point (t, f) from the integrated value g (i) (t, f) of the level calculated for each of the non-defective speakers 60. The extracted maximum value G (t, f) is the maximum value of the level integrated value for all the non-defective products, and is the level integrated value 1 m representing the non-defective product.
A filter that discriminates the speaker 60 to be inspected into a non-defective product and a defective product needs to be a filter that determines that all the non-defective products prepared as samples are determined to be good products. For that purpose, it is preferable from the viewpoint of safety to adopt the maximum value of the level corresponding to the non-defective product having the sound output characteristic closest to the defective product.
The maximum value G (t, f) is expressed by the following equation (3).

  From the spectrograms 2s of a plurality of defective products, the level integrated value 2m is calculated in the same manner as the non-defective products. When the index for distinguishing a plurality of defective products is ii, the level at the point (t, f) at each time point / frequency on the spectrogram 2s is expressed by the following equation (4).

  From the spectrogram 2s of each defective product, the integrated value of the level up to each frequency at a certain time is calculated. The integrated value g (ii) (t, f) is expressed by the following equation (5).

Next, the ratio over (ii, t, f) of the level integrated value g (ii) (t, f) acquired from a plurality of defective products to the maximum value G (t, f) of the level integrated value acquired from a plurality of non-defective products f) is calculated for each of a plurality of defective products. over (ii, t, f) is the decibel difference. The ratio over (ii, t, f) corresponds to the level difference so that the logarithmic subtraction corresponds to the true division.
Over (ii, t, f) is expressed by the following equation (6).

  FIG. 12 is an explanatory diagram conceptually showing the calculation of the ratio over (ii, t, f). A graph of G (t, f) related to non-defective products is shown in the upper left of FIG. In the upper right of FIG. 12, g (ii) (t, f) related to the ii-th defective product is shown. Calculating the level difference between the non-defective product and the defective product corresponds to calculating the ratio over (ii, t, f). The graph of over (ii, t, f) regarding the ii-th defective product is shown in the lower part of FIG.

  The CPU 21 calculates over (ii, t, f) corresponding to the difference 1d between the level integrated value 1m of the non-defective product and the level integrated value 2m of the plurality of defective products for each of the plurality of defective products. As illustrated in FIG. 7, the CPU 21 sequentially cuts the difference graph 1d of the integrated values calculated for each of the plurality of defective products using a plurality of curves of the dynamic filters 1f prepared in advance. The cut surface is a graph of a level difference between the maximum value of the level when a plurality of non-defective products are processed by the dynamic filter 1f and the level of processing the defective products.

  FIG. 13 is an explanatory diagram showing a curve showing a change over time of a cutoff frequency of a filter for discriminating a defective product from a non-defective product group on the graph of over (ii, t, f) relating to the ii-th defective product in FIG. It is. The cutoff frequency of the filter changes with time t. The change of the cut-off frequency in FIG. 13 is indicated by a bold line on the (t, f) plane of the graph of over (ii, t, f) in the lower part of FIG. Here, a curve indicating a temporal change in the cutoff frequency of the filter for discriminating defective products from the good product group is referred to as a cutoff curve C. In FIG. 13, the cut-off curve C is indicated by a bold line.

  FIG. 14 is a cross-sectional view of FIG. 13 taken along the cut-off curve C. The vertical axis in FIG. 14 is the difference between the maximum value G (t, f) of the level integrated value of the non-defective product and the level integrated value g (ii) (t, f) of the ii-th defective product. The horizontal axis in FIG. 14 is the frequency. Since the level integrated value is an integrated value of the level from the highest frequency to the frequency point, FIG. 14 shows a level integrated value that is filtered between a non-defective product and a defective product by a high-pass filter that filters a high frequency band. The level difference concerning is shown. The arrows in FIG. 14 indicate the maximum value of the level difference, that is, the detected value of the defective product (ii). The detected value Over (ii, C) of the defective product (ii) by the cut-off curve C is expressed by the following equation (7).

  Returning to FIG. 7, the description will be continued. FIG. 7 shows the level difference 11 after the filtering process, with the vertical axis representing the level and the horizontal axis representing the frequency. The CPU 21 calculates a detection value Over (ii, C) based on the cut-off curve C for each of a plurality of defective products. The CPU 21 writes each detected value calculated for each of the plurality of defective products as a list 200 in the hard disk 24 in association with the identification information of the defective products.

The CPU 21 extracts the minimum value of each detection value written in the list 200, that is, the defect detection capability value.
FIG. 15 is an explanatory diagram conceptually showing a process of extracting the defect detection capability value D (C) from the detected value Over (ii, C) of n defective products.
The minimum detection value, that is, the defect detection capability value D (C), can be divided into an integrated graph 1m for all non-defective products and an integrated graph 2m for all defective products by a cut-off curve C corresponding to D (C). It is the amount that can be. The defect detection capability value D (C) is expressed by the following equation (8).

Returning to FIG. 7 again, the description will be continued. The CPU 21 writes the extracted defect detection capability value D (C) as a list 300 in the hard disk 24 in association with the filter identification information. The CPU 21 changes the dynamic filter 1f prepared in advance and repeats the above processing. As a result, the defect detection capability value D (C) is written in the list 300 by the number of prepared dynamic filters 1f.
When the extraction process of the defect detection capability value D (C) is completed for all the dynamic filters 1f prepared in advance, the CPU 21 detects the maximum defect detection from each defect detection capability value D (C) written in the list 300. The filter corresponding to the capability value D (C) is determined as the inspection filter.

The CPU 21 inspects the speaker 60 to be inspected with the determined inspection filter.
The CPU 21 inputs a signal related to the speaker 60 to be inspected. The CPU 21 filters the input signal with the determined filter for inspection. The CPU 21 compares the signal relating to the filtered speaker to be inspected with the inspection reference data prepared in advance. The CPU 21 determines pass / fail of the speaker 60 to be inspected based on the comparison result.

  The ratio over (ii, t, f) is used as reference data for inspection standard data for determining whether the level of the filtered speaker 60 to be inspected corresponds to a non-defective product or a defective product. Can do.

In the second embodiment, reference data representing non-defective products is obtained from level integrated values of spectrograms 1s in a plurality of non-defective speakers 60. However, as in the first embodiment, the spectrogram 1s of one good speaker 60 may be used as reference data. In this case, the level integrated value at the point (t, f) at each time point / frequency related to one good product is used as reference data.
Further, a method for obtaining reference data representing a non-defective product from level integrated values of spectrograms 1s in a plurality of non-defective speakers 60 may be applied to the first embodiment. In such a case, the level at the point (t, f) at each time point / frequency is obtained for each of a plurality of non-defective products, and the maximum level for each point (t, f) is used as reference data.

Embodiment 3
The third embodiment relates to a mode in which the filter used for the inspection of the speaker 60 is determined while suppressing the amount of calculation until the maximum defect detection capability value D (C) is obtained. In order to determine the filter that maximizes the defect detection capability value D (C), if the defect detection capability value D (C) is calculated for many patterns prepared in advance, the amount of calculation becomes enormous, and the filter determination is a reality. It ’s not right. Therefore, an inspection filter is determined from three types of patterns that are highly practical. Note that determining the filter corresponds to determining the shape of the cut-off curve C.
In the third embodiment, the same reference numerals are assigned to the same components as those in the first and second embodiments, and detailed description thereof is omitted.

  The dynamic filter 1f for determining the inspection filter is a high-pass filter. The cut-off frequency of the dynamic filter 1f changes following the frequency change of the fundamental wave or the harmonic wave of the signal based on the sound from the speaker 60 to which the sweep signal is input. One of the parameters of the dynamic filter 1f is a magnification m of the harmonic frequency with respect to the fundamental frequency.

16A, 16B, and 16C are explanatory diagrams illustrating pattern examples. In each of FIGS. 16A, 16B, and 16C, the horizontal axis represents time, and the vertical axis represents frequency. The patterns in FIGS. 16A, 16B, and 16C are referred to as a pattern, b pattern, and c pattern for convenience.
The shape of the a pattern is the same shape as the time variation of the harmonic frequency where the cut-off frequency is the order m = a constant integer value.

  The shape of the b pattern is a shape in which the cut-off frequency increases following high-order harmonics, and then the cut-off frequency increases following low-order harmonics. The shape of the b pattern is a shape in which the cut-off frequency is constant in an intermediate frequency band in which the higher order harmonics transition to the lower order harmonics. Even if the speaker 60 is determined as non-defective, for example, in the spectrogram 1s of the non-defective product shown in FIG. 8, the frequency at the boundary on the high frequency side is a plurality of harmonics in the vicinity of time 360 ms to 480 ms and frequency 10 kHz. It appears almost constant. Even in the case of a non-defective speaker 60, when there is an inspection specification that allows a plurality of harmonics in the vicinity of a time of 360 to 480 ms and a frequency of 10 kHz, the shape of the filter that blocks a signal common to the non-defective product and the defective product is b The pattern shape is reasonable.

  The cut-off frequency follows the harmonic frequency of the order m0 while the fundamental frequency changes from the start frequency to the frequency f0. The cut-off frequency maintains a constant frequency (f0 × m0) while the frequency of the fundamental wave changes from the frequency f0 to the frequency f1 (> f0) in the intermediate frequency band. The cut-off frequency follows the harmonic frequency of the order m1 (<m0) while the fundamental frequency is between the frequency f1 and the end frequency. Note that f0 × m0 = f1 × m1. For example, when f0 = 500 Hz, m0 = 7th order, f1 = 1750 Hz, and m1 = 2nd order, 500 Hz × 7 = 1750 Hz × 2.

  The c pattern is a modification of the b pattern. The shape of the c pattern is a shape in which the cutoff frequency is not constant in the frequency constant section of the b pattern (section in which the fundamental frequency changes from f0 to f1), but gradually changes. In the case of the speaker 60 having the inspection specification that some noise may be generated in the intermediate frequency band, the signal level difference between the non-defective product and the defective product is larger in the c pattern than in the b pattern. Sometimes. Therefore, the shape of the c pattern is considered.

  In the case of the c pattern, the cut-off frequency monotonously increases in a section where the fundamental frequency changes from f0 to f1. In a diagram having a time axis and a frequency axis, when time and frequency are expressed logarithmically, the shape of the c pattern is a straight line in a section where the frequency of the fundamental wave changes from f0 to f1. Note that f0 × m0 <f1 × m1. For example, when f0 = 500 Hz, m0 = 7th order, f1 = 3 kHz, and m1 = 2 order, 500 Hz × 7 <3 kHz × 2.

  Many other patterns of the cut-off curve C are conceivable. However, since the processing is complicated and the above three patterns can be optimized in many cases for inspecting the abnormal sound of the speaker 60, the shape of the cut-off curve C is set to 3 of the a, b and c patterns here. It was made into a kind.

Even if the speaker 60 is a non-defective product, low-order distortions such as second-order distortion and third-order distortion in the speaker 60 often have a high level. Therefore, it is not suitable that the harmonic order m followed by the cutoff frequency in the high-pass filter for inspecting the quality of the speaker 60 is too low. Even if the order m is too high, the cutoff frequency of the high-pass filter becomes too high, so that the high-pass filter cuts off the abnormal frequency signal generated by the defective speaker 60. Therefore, in the third embodiment, the dynamic filter 1f that is changed when the filter coefficient of the dynamic filter 1f is changed in order to reduce the amount of calculation until the maximum value of the defect detection capability value D (C) is calculated. The order m of the harmonics followed by the cut-off frequency is limited to a certain range in advance. Here, let the upper limit value and the lower limit value of the limit range of the order m in the harmonic be m _max and _mmin , respectively. Note that m _max and m _min are empirical values corresponding to the cut-off curve C that can discriminate between good and defective products.

In order to reduce the amount of calculation, in consideration of the variation between the non-defective product and the defective product in the speaker 60 to be inspected, a level difference at which one filter can distinguish the spectrogram 1s from the good product and the spectrogram 2s from the defective product is determined in advance. Limit to a certain range. Here, the defect detection capability values D (C) corresponding to the upper limit value and the lower limit value of the level difference range are defined as D _max and D _min , respectively.

D _max and D _min are empirical values of level differences for discriminating between good products and defective products. For example, it has been found that when the level difference is larger than 6 to 10 dB, it is possible to discriminate between a good product and a defective product. Therefore, for example, 15 decibels is set as _Dmax, and for example, 6 decibels is set as _Dmin .

Next, the filter determination process will be described.
As a large flow of the filter determination process, the CPU 21 changes the shape of the filter from the low frequency side to the high frequency side, and determines a filter whose defect detection capability value D (C) exceeds _Dmax as a filter for inspection. To do. At that time, the CPU 21 starts the filter determination process from the a pattern. CPU21 performs a filter determination process about b pattern, when D (C) calculated about a pattern does not exceed _Dmax . CPU21 performs a filter determination process about c pattern, when D (C) calculated about b pattern does not exceed _Dmax .

CPU21 receives the D _max, D _min, m _max and m _min. CPU21 substitutes _Dmin to the variable D0. The variable D0 is a work variable that holds the maximum D (C) at the time of processing each time D (C) calculated for each of the plurality of defective products and the plurality of filters is updated to a larger value. When the calculated D (C) exceeds the received _Dmax , the CPU 21 terminates the filter search at that time. Thereby, shortening of processing time can be aimed at.

FIG. 17 is an explanatory diagram showing an image of processing for obtaining D (C) from the a pattern.
The CPU 21 calculates D (C) by setting _mmin to the order m for the a pattern. That is, the CPU 21 obtains Over (ii, C) from the data on the pattern of m = m _min on the over (ii, t, f) plane for each defective product ii for sample. The CPU 21 extracts the minimum value D (C) from Over (ii, C) of all defective products.

The CPU 21 similarly extracts the minimum value D (C) for the pattern in which the order m is incremented by 1. When the extracted minimum value D (C) is larger than D0, the CPU 21 substitutes the extracted minimum value D (C) into D0. The CPU 21 repeats the same processing until the order m becomes m _max . However, the CPU 21 ends the search if the extracted minimum value D (C) exceeds D _max during the process. At that time, the CPU 21 has determined a filter for inspecting the speaker 60 from the a-pattern filter.

When the minimum value D (C) extracted for the a pattern does not exceed D _max , the CPU 21 searches for the inspection filter for the b pattern. In that case, CPU21 performs the process which restrict | limits the calculation range regarding a filter.

  FIG. 18 is an explanatory diagram showing the limitation of the calculation range related to the filter. In each figure shown in FIG. 18, the horizontal axis represents time, and the vertical axis represents frequency. In each figure of FIG. 18, the frequency time change curve of the fundamental wave related to the sound generated by the speaker 60 is displayed by a broken line for reference. In the upper part of FIG. 18, the distribution of over (ii, t, f) regarding n defective products is indicated by white and gray areas. The white area is an area where the value of over (ii, t, f) regarding each defective product is larger than D0 obtained by the filter determination in the a pattern. The gray area is an area where the value of over (ii, t, f) for each defective product is equal to or less than D0 obtained by filter determination in the a pattern. The white area corresponds to an area where the level of abnormal sound generated by the speaker 60 is high.

The lower part of FIG. 18 shows a region indicating a calculation range related to a filter on the distribution of over (ii, t, f) related to n defective products. In FIG. 18, the outside of the area indicated by the broken line is an area where the signal level difference between the non-defective product and the defective product is small, and it is not necessary to change the shape of the pattern in order to determine the filter. Therefore, when changing the shape of the pattern, the CPU 21 excludes the outside of the broken line portion from the calculation process for determining the filter. For this purpose, the CPU 21 determines the range of the order m (m _L , m _H ) that specifies the range in which the calculation is restricted based on the distribution of over (ii, t, f) regarding the defective product, and the range of the cutoff frequency. (Fmax, fmin) is determined. Thereby, the CPU 21 can improve the processing speed.

  After executing the calculation range restriction shown in FIG. 18, the CPU 21 determines a filter for inspection with respect to the b pattern filter. The shape of the b pattern is determined by each shape of the harmonic part of the order m0, the part of the constant frequency section, and the harmonic part of the order m1 (see FIG. 16B). The CPU 21 sequentially changes these three portions to dynamically change the pattern shape. The harmonic part of order m0 corresponds to the part where the frequency of the fundamental wave changes from the lowest frequency to f0. The portion of the constant frequency section corresponds to a portion where the frequency of the fundamental wave changes from f0 to f1. The harmonic part of the order m1 corresponds to a part where the frequency of the fundamental wave changes from f1 to the highest frequency.

  FIG. 19 is an explanatory diagram showing an image of the process of changing the shape of the b pattern. The shape of the harmonic part of the order m0 is determined by the order of the harmonic. The shape of the portion of the constant frequency section is determined by the positions of both ends of the line segment indicating the constant frequency. One end position of the line segment is determined by the frequency f0 and the harmonic order m0. The other end position of the line segment is determined by the frequency f1 and the harmonic order m1. The shape of the harmonic part of the order m1 is determined by the order of the harmonic. Therefore, the shape of the b pattern is determined by the order m0, the frequency f0, the frequency f1, and the order m1. In the case of the b pattern, since f0 × m0 = f1 × m1, f1 is obtained by f0 × m0 / m1. Accordingly, the shape of the b pattern is determined by the order m0, the frequency f0, and the order m1.

The range in which the order m0 is changed is from _mmin + 1 to _mmax . Range of changing the order m1 is from m _min to m0-1. The range in which f0 is changed is from the lowest frequency generated to fmax obtained as a calculation limit range based on the level difference regarding defective products. The minimum frequency here is a frequency determined in consideration of the characteristics of the speaker 60, and is, for example, 20 Hz.

The concept of changing each part of the b pattern will be described. The order of changing each part of the b pattern is the order of the harmonic part of the order m1, the part of the constant frequency section, and the harmonic part of the order m0. CPU21 is orders m0, by fixing the frequency f0, to vary the order m1 to m0-1 from m _min. Next, the CPU 21 increases the frequency f0 by a constant value Δf. CPU21 is orders m0, by fixing the frequency f0 which is higher by a predetermined value Delta] f, changing the order m1 to m0-1 from m _min. Then CPU21 is as high as more constant value Δf frequency f0, changing the order m1 to m0-1 from m _min. When the frequency f0 reaches fmax, the CPU 21 increments the order m0 by 1 and sets it to _mmin + 2. Then, CPU 21 may order m0, by fixing the frequency f0, to vary the order m1 to m0-1 from m _min. As described above, the CPU 21 changes the shape of the b pattern by changing the order m1, the frequency f0, and the order m0 within the range sequentially determined in this order.

FIG. 20 is a flowchart illustrating an example of a processing procedure for sequentially changing each part of the b pattern.
CPU21 is the m _min +1 in order m0, the lowest frequency in the frequency f0, substituting m _min in order m1 (step S201). In this case, since already m1 = m0-1, CPU 21 is a m _min +2 next degree m0, the lowest frequency in the frequency f0, substituting m _min in order m1 (step S202). The CPU 21 increments the order m1 (step S203). The CPU 21 determines whether m1 = m0-1 (step S204). When determining that m1 = m0-1 is not satisfied (step S204: NO), the CPU 21 returns the process to step S203. CPU21, when it is determined that m1 = m0-1 (step S204: YES), substituting m _min in order m1 (step S205). The CPU 21 adds Δf to f0 (step S206).

The CPU 21 determines whether or not f0 = fmax (step S207). When determining that f0 = fmax is not satisfied (step S207: NO), the CPU 21 returns the process to step S203. CPU21, when it is determined that the f0 = fmax (step S207: YES), the lowest frequency in the frequency f0, substituting m _min in order m1 (step S208). The CPU 21 increments the order m0 (step S209). CPU21 determines whether m0 = m _max (step S210). CPU21, when judged not to be m0 = m _max (Step S210: NO), the process returns to step S203. If CPU21 was determined to be m0 = m _max (Step S210: YES), the process ends.

Note that the CPU 21 limits the range in which the orders m0 and m1 are changed to the order limit range (m _L , m _H ) related to the calculation obtained based on the level difference regarding the defective product. Further, the CPU 21 limits the range in which f0 × m0 changes to the cutoff frequency limit range (fmax, fmin) related to the calculation obtained based on the level difference regarding the defective product.

A process in which the CPU 21 calculates the minimum value D (m) for the b pattern will be described. In the b-pattern filter (m0, f0, m1, f1), the CPU 21 divides the pattern into three portions C0, C1, and C01.
FIG. 21 is an explanatory diagram showing three portions C0, C1, and C01 in the b pattern. C0 is a portion corresponding to a section in which the frequency of the fundamental wave changes from the lowest frequency to f0 (the harmonic portion of the order m0). C1 is a portion (harmonic portion of the order m1) corresponding to a section where the fundamental frequency changes from f1 to the maximum frequency. C01 is a portion corresponding to a section in which the frequency of the fundamental wave changes from f0 to f1 (part of a constant frequency section).

CPU21 at the stage of the filter determining process in a pattern, when calculating the difference in the pattern corresponding to the harmonic of order m _i, C0, C1 respectively corresponding Over0 (ii, f, mi) , Over1 (ii , F, mi) is calculated (see FIG. 12). The CPU 21 stores the calculated Over0 (ii, f, mi) and Over1 (ii, f, mi) in the RAM 23 in advance.

  FIG. 22A is an explanatory diagram showing a difference value on a pattern. FIG. 22B is an explanatory diagram illustrating a process of calculating Over0 (ii, f, mi) corresponding to C0. FIG. 22C is an explanatory diagram illustrating a process of calculating Over1 (ii, f, mi) corresponding to C1. In FIGS. 22A, 22B, and 22C, the time elapses from the left side to the right side. The CPU 21 extracts the maximum value Over0 (ii, f, mi) corresponding to C0 by comparing the difference values in order from the start to the end of the frequency change of the input signal at the stage of the filter determination process in the a pattern. Keep it. The CPU 21 extracts the maximum value Over1 (ii, f, mi) corresponding to C1 by comparing the difference values in order from the end of the frequency change of the input signal to the start at the stage of the filter determination process of the a pattern. Keep it.

  Here, the CPU 21 obtains Over (ii, Cx) for each of C0, C1, and C01 for each defective product, and sets the obtained maximum value of Over (ii, Cx) to Over (ii, C). . Cx is any one of C0, C1, and C01.

  FIG. 23 is an explanatory diagram showing an image in which the maximum value is set to Over (ii, C) from Over (ii, Cx) corresponding to C0, C1, and C01. CPU21 substitutes Over0 (ii, f, mi) memorize | stored in RAM23 for Over (ii, C0) in the area of C0. CPU21 substitutes Over1 (ii, f, mi) memorize | stored in RAM23 for Over (ii, C1) in the area of C1. In the section of C01, the CPU 21 obtains Over (ii, C01) from the calculated data of over (ii, t, F). The CPU 21 sets the maximum value to Over (ii, C) among Over (ii, C0), Over (ii, C1), and Over (ii, C01).

The CPU 21 obtains the maximum value Over (ii, C) for all defective products, extracts the minimum value from the obtained maximum value Over (ii, C), and sets it to D (C). When the extracted minimum value D (C) is greater than D0, the CPU 21 assigns the extracted minimum value D (C) to D0 and updates D0. When the extracted minimum value D (C) is equal to or less than D0, the CPU 21 changes the pattern shape, for example, by incrementing m1 by 1 for the harmonic portion of the order m1, and repeats the above processing. However, the CPU 21 ends the filter determination process when the extracted minimum value D (C) exceeds D _max during the process.

CPU21 performs a filter determination process about c pattern, when all the minimum value D (C) extracted about b pattern does not exceed _Dmax . At that time, the CPU 21 executes the filter determination process for the c pattern after executing the restriction of the calculation range shown in FIG. 18 based on D0 which is the maximum value of the defect detection capability value calculated by the processes so far. .

FIG. 24 is an explanatory diagram showing an image of the process of changing the shape of the c pattern.
The shape of the c pattern is determined by each shape of the harmonic part of order m0, the part where the frequency gradually increases, and the harmonic part of order m1 (see FIG. 16C). The harmonic part of order m0, the part where the frequency gradually increases, and the harmonic part of order m1 correspond to C0, C01 and C1 in FIG. 21, respectively. Therefore, C0, C01, and C1 are attached to the time axis of FIG. 24 for reference.

As in the case of the b pattern, the CPU 21 sequentially changes the order m0, the frequency f0, and the order m1 to change the shape of the c pattern. However, in the case of the c pattern, since the frequency f1 is not fixed to one, the CPU 21 sequentially changes the frequency f1 to change the shape of the c pattern. The range in which the frequency f1 is changed is from the lowest frequency generated to fmax. The minimum frequency here is a frequency determined in consideration of the characteristics of the speaker 60, and is, for example, 20 Hz. The CPU 21 changes the shape of the c pattern by changing the frequency f1, the order m1, the frequency f0, and the order m0 within a range sequentially determined in this order.
The change of the frequency f1 here corresponds to changing the position of the other end point of the portion where the frequency gradually increases (corresponding to C01) to the high frequency side along the harmonic of the order m1. Specifically, the CPU 21 changes the position of the other end point so that f1 × m1 changes from a value obtained by adding a constant value to f0 × m0 to fmax.

The calculation of D (C) in the case of the c pattern will be described.
As in the case of the b pattern, the CPU 21 divides the c pattern into three portions C0, C1, and C01. The CPU 21 obtains Over (ii, Cx) and sets the maximum value to Over (ii, C).

The CPU 21 obtains the maximum value Over (ii, C) for all defective products, extracts the minimum value from the obtained maximum value Over (ii, C), and sets it to D (C). When the extracted minimum value D (C) is greater than D0, the CPU 21 assigns the extracted minimum value D (C) to D0 and updates D0. When the extracted minimum value D (C) is equal to or less than D0, the CPU 21 changes the shape of the c pattern and repeats the above processing. However, the CPU 21 ends the filter determination process when the extracted minimum value D (C) exceeds D _max during the process. When the minimum value D (C) exceeding D _max is not found for all of the a pattern, b pattern, and c pattern, the CPU 21 determines that the maximum value D0 of the defect detection capability value exceeds D _min in the processing so far. The filter corresponding to the maximum value D0 is determined as the inspection filter.

25 and 26 are flowcharts showing an example of the procedure of the filter determination process.
CPU21 receives the D _max, D _min, m _max and m _min (step S301). CPU21 substitutes _Dmin to the variable D0 (step S302). CPU21 substitutes 1 to the variable P (step S303). The variable P is a flag variable for determining the pattern. The variable P is assigned 1 for the a pattern, 2 for the b pattern, and 3 for the c pattern.

The CPU 21 determines whether or not P = 1 (step S304). When determining that P is not 1 (step S304: NO), the CPU 21 restricts the calculation range related to the filter based on D0 (step S305), and advances the process to step S306. In step S305, the CPU 21 determines an order m range (m _L , m _H ) that specifies a calculation range and a cut-off frequency range (fmax, fmin). If the CPU 21 determines that P = 1 (step S304: YES), it calculates Over (ii, C) for the unprocessed defective product (ii) (step S306). The CPU 21 determines whether or not Over (ii, C) has been calculated for all defective products (step S307). If the CPU 21 determines that Over (ii, C) has not been calculated for all defective products (step S307: NO), the process returns to step S306.

  When it is determined that Over (ii, C) has been calculated for all defective products (step S307: YES), the CPU 21 calculates D (C) (step S308). The CPU 21 determines whether or not D (C)> D0 (step S309). If the CPU 21 determines that D (C)> D0 (step S309: YES), it substitutes D (C) for D0 (step S310). The pattern corresponding to D (C) assigned to D0 in step S310 is a provisional pattern corresponding to the filter candidate determined for inspection. If the CPU 21 determines that D (C)> D0 is not satisfied (step S309: NO), the process proceeds to step S313.

The CPU 21 determines whether or not D (C) ≧ _Dmax (step S311). If the CPU 21 determines that D (C) ≧ _Dmax (step S311: YES), the CPU 21 outputs the filter coefficient of the filter corresponding to D (C) (step S312), and ends the process. When determining that D (C) ≧ _Dmax is not satisfied (step S311: NO), the CPU 21 determines whether or not processing has been executed for all filters included in the filter pattern being processed (step S313). If the CPU 21 determines that the process has not been executed for all the filters (step S313: NO), the CPU 21 changes the pattern shape according to the value of P (step S314), and returns the process to step S306.

  If the CPU 21 determines that the processing has been executed for all the filters included in the filter pattern being processed (step S313: YES), it determines whether P = 1 (step S315). If the CPU 22 determines that P = 1 (step S315: YES), it substitutes 2 for P (step S316), and the process returns to step S304. When determining that P is not 1 (step S315: NO), the CPU 21 determines whether or not P = 2 (step S317). If the CPU 21 determines that P = 2 (step S317: YES), it substitutes 3 for P (step S318), and the process returns to step S304.

When determining that P = 2 is not satisfied (step S317: NO), the CPU 21 determines whether D (C) corresponding to the provisional pattern is substituted for D0 (step S319). When the CPU 21 determines that the provisional pattern is substituted for D0 (step S319: YES), the CPU 21 advances the process to step S312. If the CPU 21 determines that D (C) corresponding to the provisional pattern is not substituted for D0 (step S319: NO), the CPU 21 outputs that there is no cut-off curve C capable of separating the non-defective product / defective product (step S319). S320), the process ends.

In the present embodiment, the CPU 21 changes the filter to be determined from the a pattern to the b pattern and further from the b pattern to the c pattern when D (C) calculated for the pattern does not exceed D _max . . However, if D (C) calculated for the a pattern does not exceed D _max , the CPU 21 does not shift to the b pattern, determines the filter corresponding to D0 at that time as the inspection filter, and performs the filter determination process. May be terminated. Similarly, if D (C) calculated for the b pattern does not exceed D _max , the CPU 21 does not shift to the c pattern, determines the filter corresponding to D0 at that time as the inspection filter, and determines the filter. Processing may be terminated. Thereby, shortening of processing time can be aimed at.

  In the present embodiment, the CPU 21 limited the calculation range related to the filter. However, the CPU 21 may change the target of the filter determination process from the a pattern to the b pattern and from the b pattern to the c pattern without limiting the calculation range regarding the filter.

According to the inspection apparatus which concerns on this application, the filter for test | inspecting the said apparatus can be determined for every specification of the apparatus which generate | occur | produces a sound or a vibration.
The CPU 21 can determine the filter that maximizes the signal level difference between the non-defective product and the defective product of the speaker 60 prepared in advance by repeating the process of dynamically changing the filter.

According to the filter determination device 20, it is possible to reduce the amount of calculation until the maximum defect detection capability value D (C) is obtained.
The filter determination device 20 executes filter determination within the accepted ranges of D _max and D _min . Therefore, the filter determination device 20 can reduce the number of D (C) to be calculated. Further, when the calculated D (C) exceeds D _max , the filter determination device 20 ends the determination of the filter at that time, and selects a filter corresponding to D (C) exceeding D _max as a test filter. And decide. Thereby, the filter determination apparatus 20 can aim at shortening of processing time.

The filter determination device 20 executes filter determination processing in three patterns of a pattern, b pattern, and c pattern. The filter determination device 20 changes the shape of the pattern within the limited filter pattern range and within the accepted order ranges of m _max and _mmin . Such limiting conditions also contribute to a reduction in the amount of calculation. Furthermore, the filter determination device 20 limits the calculation range based on D0 in determining the filters for the b pattern and the c pattern. The limitation of the calculation range can also reduce the calculation amount.

  The disclosed embodiments are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  The technical features (components) described in each embodiment can be combined with each other, and a new technical feature can be formed by combining them.

20 filter determination device 21 CPU
23 RAM
29 I / O part 24 Hard disk 30 Measuring part 40 Microphone 50 Anechoic box 60 Speaker 2P Program

Claims (12)

Prior to inspecting a device to be inspected, a plurality of signals obtained by processing a signal relating to sound or vibration generated by one or a plurality of non-defective products and one or a plurality of defective products prepared in advance by a plurality of filters, respectively. Based on the determination of the filter for inspection that can change the characteristics over time,
The inspected device occurs, a sound or a signal according to the vibration frequency changes as time elapsed, treated with the determined filter for inspection,
Based on the processed signal, the quality of the device to be inspected is determined.
Inspection method characterized by that . Calculating a level difference between signals related to sound or vibration generated by the non-defective product and the defective product,
Process the signal corresponding to the calculated level difference with each of multiple filters,
The inspection method according to claim 1 , wherein a filter corresponding to the maximum level difference among the level differences corresponding to signals processed by each of the plurality of filters is determined as an inspection filter. Calculating a level difference between signals related to sound or vibration generated by the non-defective product and the plurality of defective products for each of the plurality of defective products;
Process the signal corresponding to the calculated level difference with each of multiple filters,
Calculating the maximum level difference corresponding to the processed signal for each of the plurality of defective products;
For each of the plurality of filters, select the minimum value of the level difference among the maximum values of the level difference calculated for each of the plurality of defective products,
The inspection method according to claim 1 or 2 , wherein a filter corresponding to the maximum level difference among the minimum values of the level differences selected for each of the plurality of filters is determined as an inspection filter. From the signals based on the time, frequency, and level information respectively calculated from the signals related to a plurality of non-defective products prepared in advance, each maximum level is extracted for each time and frequency defined at a predetermined interval,
Set each extracted maximum level to the level of the signal related to the sound or vibration generated by the good product,
The inspection method according to claim 3 , wherein the maximum value of the level difference is calculated for each of the plurality of defective products based on the level of the signal related to the set non-defective product. If the minimum value of the selected level difference exceeds the specified value, the selection of the minimum value of the level difference is terminated,
The inspection method according to claim 3 or 4 , wherein a filter corresponding to a minimum value of the level difference exceeding the predetermined value is determined as an inspection filter. A filter output unit for outputting a plurality of filters;
Based on signals relating to sound or vibration generated by one or a plurality of non-defective products and one or a plurality of defective products prepared in advance, characteristics are changed over time from a plurality of filters output by the filter output unit. A determination unit that determines a filter for inspection that can be performed;
A processing unit for processing a signal relating to sound or vibration generated by the device to be inspected, the frequency of which changes with time , using the inspection filter determined by the determining unit ;
A determination unit that determines the quality of the inspection target device based on the signal processed by the processing unit;
With
Inspection apparatus characterized by that . A non-defective signal storage unit for storing a signal related to the non-defective product;
A level difference calculation unit that calculates a level difference between a signal related to a non-defective product stored in the non-defective product signal storage unit and a signal related to the defective product prepared in advance;
A signal processing unit that processes a signal corresponding to the level difference calculated by the level difference calculation unit in each of the plurality of filters output by the filter output unit;
The determination unit is configured to determine a filter corresponding to the maximum level difference among the level differences corresponding to the signal processed by the signal processing unit as an inspection filter.
A signal comparison unit that compares the signal processed by the processing unit and a predetermined signal;
The inspection apparatus according to claim 6 , wherein the determination unit determines whether the apparatus to be inspected is good or bad based on a result of comparison by the signal comparison unit. The level difference calculation unit is configured to calculate a level difference of a signal related to sound or vibration generated by the non-defective product and the plurality of defective products for each of the plurality of defective products,
The signal processing unit is configured to process a signal corresponding to the level difference calculated for each of the plurality of defective products by each of the plurality of filters output by the filter output unit,
A calculation unit that calculates a maximum value of a level difference corresponding to the signal processed by the signal processing unit for each of the plurality of defective products;
For each of the plurality of filters output by the filter output unit, the calculation unit includes a selection unit that selects the minimum value of the level difference among the maximum values of the level difference calculated for each of the plurality of defective products,
The determination unit is configured to determine a filter corresponding to a maximum level difference as an inspection filter among minimum values of level differences selected by the selection unit for each of a plurality of filters. Item 8. The inspection apparatus according to Item 7 . The selection unit is configured to end the selection of the minimum value of the level difference when the selected minimum value of the level difference exceeds a predetermined value,
The inspection apparatus according to claim 8 , wherein the determination unit determines a filter corresponding to a minimum value of the level difference exceeding the predetermined value as an inspection filter. A filter determination method for determining a test filter from a plurality of filters whose characteristics can be changed over time in order to inspect a device to be inspected that generates sound or vibration whose frequency changes over time,
The level difference of the signal relating to the sound or vibration generated by each of the non-defective product and the one or more defective products prepared in advance is calculated for each of the plurality of defective products,
Process the signal corresponding to the calculated level difference with each of multiple filters,
Calculating the maximum level difference corresponding to the processed signal for each of the one or more defective products,
For each of the plurality of filters, select the minimum value of the level difference among the maximum values of the level difference calculated for each of the one or more defective products,
A filter determination method comprising: determining a filter corresponding to a maximum level difference among minimum values of level differences selected for each of a plurality of filters as an inspection filter. A program for causing a computer to execute a process of determining one filter from a plurality of filters whose characteristics can be changed over time based on a signal relating to sound or vibration whose frequency changes over time,
Calculating a level difference between one signal related to sound or vibration and one or more other signals for each of the one or more other signals;
Change the filter that can change its characteristics over time,
In each of the changed filters, the signal corresponding to the calculated level difference is processed,
Calculating the maximum level difference corresponding to the processed signal for each of the one or more other signals;
For each of a plurality of filters, select the minimum value of the level difference among the maximum values of the level difference calculated for each of the one or more other signals,
A program for causing a computer to execute a process of determining a filter corresponding to a maximum level difference as a single filter among minimum values of level differences selected for each of a plurality of filters. A filter determination device for determining an inspection filter from a plurality of filters whose characteristics can be changed over time for an inspection device for inspecting an inspection target device that generates sound or vibration whose frequency changes over time. And
A filter output unit for outputting a plurality of filters;
A level difference calculation unit that calculates a level difference of a signal related to sound or vibration generated by a non-defective product and one or a plurality of defective products prepared in advance for each of the plurality of defective products;
A signal processing unit that processes a signal corresponding to the level difference calculated by the level difference calculation unit in each of the plurality of filters output by the filter output unit;
A calculation unit that calculates the maximum value of the level difference corresponding to the signal processed by the signal processing unit for each of the one or more defective products;
For each of the plurality of filters output by the filter output unit, a selection unit that selects the minimum value of the level difference among the maximum values of the level difference calculated by the calculation unit for each of the one or a plurality of defective products,
A filter determining apparatus comprising: a determining unit that determines a filter corresponding to the maximum level difference among the minimum values of the level differences selected by each of the plurality of filters by the selecting unit.