Predictive maintenance device for cross-modal devices combining contact vibration and non-contact audio.

TWM689498UActive Publication Date: 2026-11-11TRANZX ADVANCED TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
TW115205242
Authority / Receiving Office
TW · TW
Patent Type
Utility models
Current Assignee / Owner
Filing Date
2026-06-08
Publication Date
2026-11-11
Estimated Expiration
2036-06-07

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

This invention discloses a cross-modal equipment predictive maintenance device that combines contact vibration and non-contact audio. It includes a contact vibration sensor located in a first zone of the equipment, a corresponding non-contact audio sensor located in a second zone, and a processing unit. The processing unit obtains a first health characteristic from vibration and a second health characteristic from audio. Under overlapping conditions where both modes are measurable, it establishes a cross-modal conversion relationship. Based on measurability, it maps the audio characteristics of the unmeasurable zone to equivalent vibration level health indicators and fuses them into a single continuous trend to determine the equipment status. Furthermore, it uses intelligent noise reduction based on cross-modal harmonics to suppress background noise in the factory area. This invention is primarily used for predictive maintenance of rotating machinery such as semiconductor dry vacuum pumps, achieving near-contact diagnostic reliability even in high-temperature areas.
Need to check novelty before this filing date? Find Prior Art

Claims

1. A predictive maintenance device for rotating machinery, used to monitor a rotating machinery having a first area where a contact sensor can be installed and a second area where a contact sensor cannot be installed, the device comprising: a contact vibration sensor disposed in the first area for acquiring a structurally conducted vibration signal of the rotating machinery; a non-contact audio sensor disposed corresponding to the second area for acquiring an air-conducted audio signal of the rotating machinery; and a processing unit electrically connected to the contact vibration sensor and the non-contact audio sensor, and configured to receive a signal from the structure... A first health characteristic is obtained from the transmitted vibration signal, and a second health characteristic is obtained from the air-conducted audio signal. Under the condition that both the transmitted vibration signal and the air-conducted audio signal are measurable, a cross-modal conversion relationship between the first health characteristic and the second health characteristic is established. Based on the measurability condition of the first region and the second region, the second health characteristic is mapped to an equivalent vibration level health index for the second region using the cross-modal conversion relationship. The first health characteristic and the equivalent vibration level health index are then fused into a single and continuous health index trend to determine the state of the rotating machinery.

2. The predictive maintenance device as claimed in claim 1, wherein the measurability condition includes the surface temperature of the second zone exceeding one of the allowable temperature limits of the contact vibration sensor.

3. The predictive maintenance device as described in claim 1 further includes a temperature sensor configured to measure the temperature of one of the second zones, and the processing unit determines the measurability condition based on the temperature.

4. The predictive maintenance device as described in claim 3, wherein the processing unit is further configured to compensate for temperature-dependent drift of one of the equivalent vibration level health indicators based on the temperature.

5. The predictive maintenance device as described in claim 1, wherein the contact vibration sensor is a triaxial accelerometer, the first health feature is obtained by fast Fourier transform of its triaxial vibration signal, and includes a rotor feature and a bearing feature.

6. The predictive maintenance device as described in claim 5, wherein the processing unit is further configured to distinguish between imbalance and misalignment by means of the radial and axial components of the triaxial vibration signal.

7. The predictive maintenance device as claimed in claim 1, wherein the second health characteristic includes a multi-band sound pressure characteristic of the mid-to-high frequency band of one of the air-conducted audio signals, expressed in a 1 / 3 octave manner.

8. The predictive maintenance device as claimed in claim 1, wherein the non-contact audio sensor is positioned at a location capable of detecting an anomaly that is not detected by the first health feature due to attenuation of a mounting point of the contact vibration sensor and its structural transmission path.

9. The predictive maintenance device as claimed in claim 1, wherein the processing unit is further configured to learn a cross-modal coherence model under the overlapping conditions, the model identifying the component in the air-conducted audio signal that is coherent with the structurally conducted vibration signal, and suppressing non-coherent environmental noise in the air-conducted audio signal according to the cross-modal coherence model to obtain the noise-reduced second health characteristic; and when the structurally conducted vibration signal is unmeasurable, applying the cross-modal coherence model to denoise the air-conducted audio signal.

10. The predictive maintenance device as described in claim 1 further includes a current sensor for measuring the current of a motor driving the rotating machinery, and the processing unit obtains an operating state based on the motor current and selects or adjusts a baseline and the cross-modal transition relationship based on the operating state.

11. The predictive maintenance device as claimed in claim 10, wherein the processing unit is further configured to obtain a third health characteristic from the motor current, and to integrate the third health characteristic with the first health characteristic and the second health characteristic to determine the state and perform root cause attribution.

12. The predictive maintenance device as claimed in claim 1, wherein the processing unit is further configured to obtain a common speed or order reference and to align the analysis of the structurally conducted vibration signal and the air-conducted audio signal, the speed or order reference being obtained by estimating a motor current.

13. The predictive maintenance device as claimed in claim 1, wherein the processing unit is further configured to attribute an anomaly to a corresponding faulty part based on an elevated frequency group of one of the first health features and the second health features, and to track changes in the trend of the health indicators to predict the remaining service life of one of the faulty parts.

14. The predictive maintenance device as described in claim 1, wherein the processing unit is further configured to determine the state using at least one of a control line and a Hotelling's T2 multivariate statistics, and to select a corresponding baseline from a fingerprint database according to the model of the rotating machinery.

15. The predictive maintenance device as described in claim 1 further includes a host communication device connected to the processing unit via RS-485, USB, LoRa or Modbus.