CN113791727B - Edge acquisition equipment applied to industrial acoustic intelligent sensing - Google Patents

Abstract

The invention discloses edge acquisition equipment applied to industrial acoustic intelligent sensing, which comprises at least one sensor, an acquisition part, an acquisition controller, a data processor and an edge AI processor, wherein the acquisition part comprises a first acquisition part and a second acquisition part; the acquisition part comprises at least one AD acquisition chip; the acquisition part samples acoustic signals obtained by the sensor; the acquisition controller controls each AD sampling chip to synchronously sample according to the same sampling frequency, and continuously and uninterruptedly send sampled data to the data processor according to the sampling frequency; the data processor preprocesses the AD sampling data and then sends the AD sampling data to the edge terminal AI processor; the side-end AI processor inputs the preprocessed AD sampling data into various mathematical models, trained AI models or mathematical formulas, and analyzes to obtain results. The invention can analyze acoustic information of the product carrier which produces sound actively or passively, and obtain the analysis results of the product quality, fault, differentiation, defect, classification and the like, and the detection result is accurate, safe and reliable.

Description

An edge-end acquisition device applied to industrial acoustic intelligent perception

technical field

The invention relates to the technical field of acoustic detection, in particular to an edge acquisition device applied to industrial acoustic intelligent sensing.

Background technique

In the process of my country's development into a manufacturing powerhouse, there are problems such as an imperfect intelligent technology system, urgent breakthroughs in core technologies, and a low level of application of intelligent technology industries. The integration of a new generation of artificial intelligence and manufacturing technology can quickly improve the intelligence of the manufacturing industry level, and accelerate the process of my country's transformation from a manufacturing power to a manufacturing power. At present, with the rapid development of big data, chips, GPU and other technologies, the conditions and environment for the development of artificial intelligence AI have undergone tremendous changes. Intelligent system development.

Acoustic testing is a non-destructive testing method that collects and analyzes sound signals containing information such as the state, process, size, and defects of the object to be tested, and realizes fault identification. Therefore, acoustic testing has broad development prospects in industrial fault diagnosis. Some failures of refrigeration products, five-axis CNC tool grinder equipment parts, automatic assembly lines, etc., mainly rely on human hearing to detect and troubleshoot. However, the accuracy, efficiency, consistency and reliability of human hearing detection are all low. The present invention intends to rely on audio-visual intelligent perception technology to solve the above-mentioned problems.

Contents of the invention

In view of this, in order to solve the above-mentioned problems in the prior art, the present invention proposes an edge acquisition device applied to industrial acoustic intelligent sensing.

The present invention solves the above problems by the following technical means:

An edge-end acquisition device applied to industrial acoustic intelligent perception, including at least one sensor, an acquisition part, an acquisition controller, a data processor, and an edge-end AI processor; the acquisition part includes at least one AD acquisition chip;

Each sensor is connected to an AD acquisition chip;

The sensor is used to collect acoustic signals;

The acquisition part samples the acoustic signal obtained by the sensor;

The acquisition controller controls each AD sampling chip to sample synchronously at the same sampling frequency, and sends the sampled data to the data processor continuously and uninterruptedly at the sampling frequency;

After the data processor receives the AD sampling data, it preprocesses the AD sampling data and sends it to the edge-end AI processor;

The edge-end AI processor inputs the preprocessed AD sampling data into various mathematical models, trained AI models or mathematical formulas, and analyzes to obtain results.

Further, the collection part also includes a digital input port, an IO output port and a bus interface.

Further, the acquisition controller also obtains information through the bus interface, and sends it to the data processor for processing, and the edge-end AI processor obtains the result after processing; the adoption controller also continuously samples the digital data according to the sampling frequency. The digital input signal from the input port is sent to the data processor, and the edge-end AI processor processes it to obtain the result.

Further, the result information of the side-end AI processor is displayed on the display, or output to other systems through Ethernet, or transmitted to the acquisition controller, and output to other systems through the acquisition controller through the IO output port or the bus interface.

Further, the bus interface and the side-end AI processor are used for communication between the device and other systems, the bus interface accepts instructions to the device, and the side-end AI processor transmits the processing results of the device; the instructions to the device include collection start, end, transmission The data specifies the time length t1, the sampling frequency, the amplification factor of AD, the current sound object data label q, the specified time interval t2, and the specified format required for data processing.

Further, the acquisition controller controls each AD acquisition chip to control the amplification factor of the AD sampling chip, so that the gain of each AD sampling chip can be adjusted independently.

Further, the data processor starts from the digital input trigger time point of the digital input port, and sends the data of the specified time length t1 to the edge-end AI processor; starts from the time point when the bus interface receives the command, and sends the data of the specified time length The data of t1 is sent to the edge AI processor.

Further, in the preprocessing of the data processor, the AD data is marked according to the digital input signal of the digital input port or the command information received by the bus interface, so as to realize the automatic marking of the AI model data. The digital input signal or command information is the current voice Object data label q.

Further, when the acquisition controller receives a digital input trigger from the digital input port, it outputs a signal to the digital analog IO port according to the specified time interval t2, if the analysis result received from the side-end AI processor within the specified time interval t2 is not output signal, the digital analog IO port output is cancelled, so as to realize precise digital analog IO output control.

Further, the edge acquisition device applied to industrial acoustic intelligent sensing also includes a reference source, the acquisition controller turns off the sensor input, turns on the reference source, and the reference source inputs a reference voltage to each AD acquisition chip, and then the data processor collects The voltage value obtained by the controller is compared with the reference voltage value to obtain the deviation value of each AD acquisition chip; in the subsequent data preprocessing, the deviation value is used to correct the data collected by the AD acquisition chip, so as to realize the automatic calibration function of the acquisition device.

Compared with the prior art, the beneficial effects of the present invention at least include:

The edge acquisition device of the present invention can perform acoustic information collection, data processing and analysis on active or passive sounding product carriers, and then obtain analysis results on product quality, faults, differentiation, defects, classification, etc., and output them in real time Analysis results, detection results are accurate, safe and reliable.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained based on these drawings without creative effort.

Fig. 1 is a schematic structural diagram of an edge acquisition device applied to industrial acoustic intelligent sensing according to the present invention.

Detailed ways

In order to make the above objects, features and advantages of the present invention more comprehensible, the technical solutions of the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be pointed out that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all those skilled in the art can obtain without creative work. Other embodiments all belong to the protection scope of the present invention.

As shown in Figure 1, the present invention provides an edge-end acquisition device applied to industrial acoustic intelligent perception, including at least one sensor, an acquisition part, an acquisition controller, a data processor, and an edge-end AI processor; the acquisition part includes at least one AD acquisition chip and digital input port can have IO output port and bus interface.

The sensor is used to collect the acoustic signal. The AD acquisition chip samples the acoustic signal obtained by the sensor, and the acquisition controller sends it to the data processor. The data processor processes the data and sends it to the side-end AI processor. The side-end AI processor After the data is analyzed, the result is obtained; in addition, the acquisition controller will also obtain information through the digital input port or the bus interface, and send it to the data processor for processing; the edge AI processor will obtain the result after processing.

The result information of the edge-end AI processor can be displayed on the display, output to other systems through Ethernet, or transmitted to the acquisition controller, and output to other systems through the acquisition controller through the IO output port or bus interface.

The bus interface and the side-end AI processor can be used for communication between the device and other systems. The bus interface can accept instructions to the device, and the side-end AI processor can transmit the processing results of the device; the instructions to the device can include the start, end, and transmission of data The specified time length t1, sampling frequency, AD amplification factor, current sound object data label q, specified time interval t2, specified format for data processing, etc.

The acquisition device has an automatic calibration function; the acquisition controller turns off the sensor input, turns on the reference source, and the reference source inputs the reference voltage to each AD sampling chip, and then the data processor compares the voltage value obtained from the acquisition controller with the reference voltage value, Obtain the deviation value of each AD sampling chip; use the deviation value to correct the data collected by the AD chip during data preprocessing.

Each AD sampling gain is independently adjustable: each sensor is connected to an AD chip; the acquisition controller can control each AD sampling chip and control the amplification factor (gain) of AD.

The acquisition controller controls each AD sampling chip to sample synchronously at the same sampling frequency, and continuously and uninterruptedly send the sampled data to the data processor according to the sampling frequency; continuously and uninterruptedly sample the digital input signal of the digital input port according to the sampling frequency and send it to the data processing device.

After the data processor receives the data, it preprocesses the data and then sends it to the edge-end AI processor; it can continuously send each AD sampling data to the edge-end AI processor; it can start at the trigger time point according to the digital input of the digital input port , sending data with a specified time length t1 to the edge-end AI processor; starting from the time point in the command received by the bus interface, sending data with a specified time length t1 to the edge-end AI processor.

After the data processor receives the data, it preprocesses the data and sends it to the edge AI processor; during the preprocessing, it can receive command information (current sound object data label q) according to the digital input signal of the digital input port or the bus interface Label AD data to realize automatic labeling of AI model data.

Accurate IO output control can be realized. When the acquisition controller receives the digital input trigger of the digital input port, it will output the signal to the digital analog IO port according to the specified time interval t2. If it receives the analysis of the side-end AI processor within the specified time interval t2 If the result is no signal output, the digital analog IO port output will be canceled.

The edge-end AI processor analyzes the data to get the results, which can be input into various mathematical models, trained AI models or mathematical formulas; the AI model can be machine learning such as SVM, CNN, MLP, RNN, DNN, etc. Model.

The sensor can be composed of a group of microphones N>6, and the microphones can be arranged in an array, matrix, queue, or deployed in a certain geometric pattern in space or on a plane.

The acquisition controller can be FPGA, ARM, single-chip microcomputer and the like.

The data processor can be FPGA, ARM, single-chip microcomputer, PC, etc.

The edge AI processor can be FPGA, ARM, ASIC, GPU, PC, etc.

The bus interface can be RS232, RS485, Ethernet, etc.

The connection between the data processor and the edge-end AI processor can be bus, USB, or Ethernet.

The connection between the acquisition controller and the data processor can be bus, USB, Ethernet.

The communication between the edge AI processor and other systems can be bus, USB, Ethernet, wifi, etc.

The sampling frequency may be 40-192K.

The data processor preprocesses the data, which refers to processing the received data into a specified format.

The acquisition controller and the data processor are processed by two processors, which can ensure real-time data.

1. Industry application of refrigeration product quality inspection

After the final assembly of air-conditioning products (refrigeration products), it is necessary to turn on the power and start up for noise testing of the whole machine to troubleshoot operation failures. At present, during this process, air-conditioning products may have cross-flow fan blades touching the bottom shell, evaporator piping, pipeline collisions, axial-flow fan blades touching the guide ring, cross-flow fan blade steel shaft coming out of the bearing, and centrifugal fan blades hitting the volute. More than 20 kinds of faults, such as capillary parts touching the rear side plate, and suction pipe touching the front side plate, caused fault sounds. The invention aims at the general assembly detection of air-conditioning products, applies the audio-visual perception technology, and realizes the intelligent diagnosis of air-conditioning product failures.

2. Industry application of condition monitoring of equipment parts

Real-time monitoring of the status of parts and components of processing equipment helps to detect and troubleshoot parts or equipment failures in a timely manner, ensuring high processing accuracy and stable processing quality. Select the five-axis CNC tool grinder as the object to carry out the application of audio-visual perception technology in the equipment parts condition monitoring industry. The five-axis CNC tool grinder is used for grinding various end knives, round nose knives, ball knives, drill bits and forming knives, etc. It has the advantages of high efficiency, high precision and stable performance. The knives produced by it have been widely used in industries such as mobile phones and steel rails. In tool processing, core components such as the main shaft, feed shaft, grinding wheel, milling fluid supply system, and machine table of the grinding machine will experience poor operation, screw wear, wear/damage, excessive or insufficient milling fluid, poor lubrication/aging And more than 10 kinds of faults.

3. Industry application of automatic assembly status monitoring

The automated assembly line mainly includes conveying lines, sensor detection devices and various industrial robots, which are used in the assembly, testing, labeling, packaging and other processes of product manufacturing, and are widely used in automobile manufacturing, hardware, equipment manufacturing, food/drug packaging and other industries. In assembly production, the chains of the conveyor line are loose and rusted, the rolling elements of various bearings are broken, worn and damaged, impurities invade the bearing grease, the idler rollers do not turn, freeze, and the gears of industrial robots are worn and cracked, etc., which will cause more than 10 accidents. A fault accompanied by a fault sound. The invention aims at monitoring the state of the automatic assembly line, and implements the application of an industrial multi-modal intelligent diagnosis system to realize real-time perception and intelligent diagnosis of automatic assembly line faults.

The edge acquisition device of the present invention can perform acoustic information collection, data processing and analysis on active or passive sounding product carriers, and then obtain analysis results on product quality, faults, differentiation, defects, classification, etc., and output them in real time Analysis results, detection results are accurate, safe and reliable.

The above-mentioned embodiments only express several implementation modes of the present invention, and the description thereof is relatively specific and detailed, but should not be construed as limiting the patent scope of the present invention. It should be pointed out that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention should be based on the appended claims.

Claims (5)

1. A side-end collection device applied to industrial acoustic intelligent perception, characterized in that it includes at least one sensor, a collection part, a collection controller, a data processor and a side-end AI processor; the collection part includes at least one AD Acquisition chip; Each sensor is connected to an AD acquisition chip; The sensor is used to collect acoustic signals; The acquisition part samples the acoustic signal obtained by the sensor; The acquisition controller controls each AD sampling chip to sample synchronously at the same sampling frequency, and sends the sampled data to the data processor continuously and uninterruptedly at the sampling frequency; After the data processor receives the AD sampling data, it preprocesses the AD sampling data and sends it to the edge-end AI processor; The edge-end AI processor inputs the preprocessed AD sampling data into various mathematical models, trained AI models or mathematical formulas, and analyzes to obtain results; The acquisition part also includes a digital input port, an IO output port and a bus interface; The bus interface and the side-end AI processor are used for communication between the device and other systems. The bus interface accepts instructions to the device, and the side-end AI processor transmits the processing results of the device; the instructions to the device include the start, end, and specified time of data transmission. Length t1, sampling frequency, AD magnification, current sound object data label q, specified time interval t2, and specified format required for data processing; The data processor starts at the digital input trigger time point of the digital input port, and sends the data of the specified time length t1 to the edge-end AI processor; starts at the time point when the bus interface receives the command, and sends the data of the specified time length t1 to the edge AI processor; In the preprocessing of the data processor, the AD data is marked according to the digital input signal of the digital input port or the command information received by the bus interface, so as to realize the automatic marking of the AI model data. The digital input signal or command information is the data label of the current sound object q; When the acquisition controller receives the digital input trigger of the digital input port, it outputs a signal to the digital analog IO port according to the specified time interval t2. If the analysis result received by the edge AI processor within the specified time interval t2 is no output signal, Then cancel the digital analog IO port output, so as to realize precise digital analog IO output control. 2. The edge-end acquisition device applied to industrial acoustic intelligent perception according to claim 1, wherein the acquisition controller also obtains information through the bus interface, and sends it to the data processor for processing, and the edge-end AI The processor obtains the result after processing; the acquisition controller also continuously samples the digital input signal of the digital input port according to the sampling frequency and sends it to the data processor, and the edge AI processor obtains the result after processing. 3. The edge-end acquisition device applied to industrial acoustic intelligent perception according to claim 1, characterized in that, the result information of the edge-end AI processor is displayed on a display, or output to other systems through Ethernet, or transmitted To the acquisition controller, output to other systems through the acquisition controller from the IO output port or the bus interface. 4. The edge-end acquisition device applied to industrial acoustic intelligent perception according to claim 1, wherein the acquisition controller controls each AD acquisition chip and controls the magnification of the AD sampling chip, thereby realizing each The gain of the AD sampling chip is independently adjustable. 5. The edge-end acquisition device applied to industrial acoustic intelligent sensing according to claim 1, wherein the edge-end acquisition device applied to industrial acoustic intelligent sensing also includes a reference source, and the acquisition controller closes the sensor input, Turn on the reference source, the reference source inputs the reference voltage to each AD acquisition chip, and then the voltage value obtained by the data processor from the acquisition controller is compared with the reference voltage value to obtain the deviation value of each AD acquisition chip; The deviation value corrects the data collected by the AD acquisition chip, so as to realize the automatic calibration function of the acquisition device.

Images