CN105651952A - Method for improving quality detection precision of linseed oil - Google Patents

Abstract

The invention discloses a method for improving the detection accuracy of linseed oil quality, comprising the following steps: (1) volatilizing the linseed oil to be tested into a gas to be tested through an air intake device and inputting the gas to be tested into a detection device; 2) The detection device collects the gas to be measured through the gas sensor array to obtain sensor data and sends the sensor data to the identification system; (3) The identification system identifies and matches the collected sensor data with the pre-stored template data to obtain the flax Seed oil quality inspection results; among them, the pre-stored template data is processed through a specific training algorithm. By adopting the technical solution of the present invention, by using multiple gas sensors to detect volatilized linseed oil gas, the quality of linseed oil can be quickly and accurately identified; at the same time, by performing specific algorithm processing on standard sensor data, thereby greatly improving Accuracy of linseed oil quality testing.

Description

A method for improving the detection accuracy of linseed oil quality

technical field

The invention belongs to the technical field of oil electronic detection, and in particular relates to a method for improving the detection accuracy of linseed oil quality.

Background technique

Flaxseed oil contains a large amount of α-linolenic acid, and eating cold-pressed flaxseed oil is the most effective way to supplement linolenic acid. From the perspective of healthy eating, flaxseed oil has many functions and functions. Flaxseed oil has many functions: make skin soft and bright, lose weight, improve women's premenstrual syndrome, improve anti-stress, reduce allergic reactions, relieve asthma, improve arthritis, lower cholesterol, improve constipation, promote cell health, and promote brain flexibility etc.

The types and contents of aroma substances in linseed oil are rich, and the aroma characteristics are obvious. The aroma quality of seed oil of different flax varieties and different preparation processes is obviously different. According to national standards, linseed oil is divided into three categories: linseed crude oil, pressed linseed oil and leached linseed oil. Pressed linseed oil can be divided into primary and secondary. The leached finished linseed oil can be further divided into primary, secondary, tertiary and quaternary.

Now many families are starting to consume flaxseed oil, but there are many brands of flaxseed oil on the market and they are messy. When people buy linseed oil, the quality of linseed oil is particularly important. However, the current quality classification and evaluation of linseed oil still mainly relies on the sensory evaluation method, and the identification results are also affected by various factors, including the age, experience, and identification environment of the examiner. Even an experienced appraiser will be affected by various accidental factors, and it is even more difficult to predict the identification ability of ordinary people; in addition, if a large-scale classification of linseed oil is required, the artificial sensory identification The method appears to be inefficient and the reliability of the identification result cannot be guaranteed.

Although the prior art has a variety of means in the detection of the quality of linseed oil, it is difficult to be generally applied to ordinary consumers. At present, the most accurate quality detection method of linseed oil is the chemical analysis method, which can quantitatively and qualitatively analyze the components of linseed oil, and directly judge the content of harmful substances in the sample and whether it is of high purity through the content of the components High-quality oil, but it takes a long time to test and the cost of the instruments used is high. It can only be carried out in some special laboratories, and it is difficult to promote it to the public. There are also some commonly used detection methods such as physical and chemical detection methods, which are mainly aimed at the higher acid value, iodine value, more oxidation products, and more in some inferior oils (such as waste oil) than normal linseed oil. The quality of linseed oil can be judged by analytical methods such as chromatography and colorimetry. This method has a high accuracy rate for the detection of secondary linseed oil, and the detection time is relatively short, but the scope of application There are significant limitations. In addition, there are some detection methods that directly look at the appearance and add common reagents to separate the mixed oil, which have become inapplicable with the advancement of linseed oil adulteration and the continuous improvement of linseed oil pressing techniques.

Therefore, in view of the above-mentioned defects existing in the current prior art, it is necessary to conduct research to provide a solution to solve the defects existing in the prior art.

Contents of the invention

In view of this, it is indeed necessary to provide a method for improving the detection accuracy of linseed oil quality, by performing specific algorithm processing on standard sensor data, thereby greatly improving the detection accuracy of linseed oil quality.

In order to overcome the defects in the prior art, the present invention provides the following technical solutions:

A method for improving the detection accuracy of linseed oil quality, comprising the following steps:

(1) volatilize the linseed oil to be tested into the gas to be tested through the air intake device and input the gas to be tested to the detection device;

(2) The detection device collects the gas to be measured through the gas sensor array to obtain sensor data and sends the sensor data to the identification system;

(3) The recognition system obtains the quality detection result of linseed oil after recognizing and matching the collected sensor data and the pre-stored template data;

Wherein, the pre-stored template data is obtained through the following steps:

S1: The recognition system collects sensor data of standard linseed oil, and initializes and normalizes part of the stable original data, namely and transform the above formula into In order to avoid the denominator being zero pairs; where x is the original data, y is the normalized data, x _max is the maximum value in the original data, and x _min is the minimum value in the original data;

S2: By calculating the input and output values of each node and calculating the error with the target output; among them, calculating the input value of the hidden layer, using the s function As an activation function, the formula is Among them, h _i is the input value of the hidden layer, W _ih is the weight value from the input layer to the hidden layer _, and Xi is the input data; after obtaining the input value of the hidden layer, calculate the output value of the hidden layer h _o =f( h _i ), similarly calculate the input and output values of the output layer; finally use the mean square error formula Obtain the error e, where Y is the target output value, and y _o is the real output of the output layer;

S3: Constantly update the weight until the error is within the acceptable range; wherein, when updating the weight, use the formula W=W _k-1 +ΔW+a[W _k-1 -W _k-2 ], the updating process is to use The error is derived from the weights of the input and output layers to obtain extreme values, including:

The error is derived from the weights of the hidden layer and the output layer to obtain the formula:

$\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ;</span>$

And the error is derived from the input and hidden layer weights to get the formula:

$\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{{<span\; class="notranslate">\; \&part;</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; h</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span><span\; class="notranslate">\; \&Sigma;</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span><span\; class="notranslate">\; \&Sigma;</span>{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; *</span><span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

S4: Finally, the weights and extremums obtained through the above steps S1 to S3 are pre-stored as template data.

Preferably, the step (1) further includes the following steps:

Put the linseed oil of the sample to be tested into an airtight container, and put the airtight container into a constant temperature water bath for heating in a constant temperature water bath;

Pass inert gas with stable flow rate into the airtight container for purging;

The gas to be tested is delivered to the detection device.

Preferably, the air intake device includes an airtight container for holding linseed oil, an air inlet pipe, a first air guide pipe and a constant temperature water bath box, and the constant temperature water bath box is used to make the linseed oil contained in the airtight container be at Constant temperature state; one end of the air inlet pipe extends into the airtight container and is immersed in the linseed oil, and the other end is fed with pure inert gas to volatilize the linseed oil; one end of the first air guide pipe extends into the airtight container The other end is connected to the detection device to deliver the gas to be tested to the detection device.

Preferably, the detection device includes a gas chamber body, a closing cover for closing the gas chamber body, and a detection circuit board arranged on the closure cover, and the detection circuit board includes a gas sensor array, a signal processing module, control module, storage module and touch display module;

The central position of the closure cover is provided with an air intake hole, and the corresponding air intake hole is also provided at the overlapping position of the detection circuit board and the central position, and the first air guide pipe is connected to the air intake hole and extends into the air intake hole. The chamber body, the first air guide tube and the air inlet hole form a fixed structure;

A plurality of gas sensors in the gas sensor array are arranged on the detection circuit board at equal intervals with the air inlet as the center, and are used to detect the composition of linseed oil in the gas to be measured and induce corresponding When the closing cover closes the air chamber body, a small cavity is formed, and the plurality of gas sensors are fully in contact with the air in the cavity;

The closure cover is also provided with an air vent for maintaining the pressure balance inside and outside the cavity, and the aperture of the air vent is smaller than the aperture of the air intake hole;

The signal processing module performs signal processing on the electrical signal and sends the processed signal to the control module;

The storage module is used to store linseed oil template data in advance;

The control module obtains sampling data through signal sampling and matches with pre-stored template data to obtain linseed oil quality detection result information;

The touch display module is used for setting system parameters and displaying information about the quality detection results of linseed oil.

Preferably, the gas sensor array includes 10 gas sensors, and the 10 gas sensors are arranged on the detection circuit board at equal intervals with the air inlet as the center, and the 10 gas sensors The sensor models are MQ138, MQ135, MQ-5, MQ136, TGS2602, TGS2611, TGS825, TGS822, TGS813, TGS826.

Preferably, the air chamber body, the closing cover and the detection circuit board are all circular, the air inlet hole is arranged at the center of the closure cover, and the gas sensors are equally spaced along the outer circumference of the detection circuit board arrangement;

Taking the detection circuit board as the base surface, the height of the gas sensor is between 1/3 and 2/3 of the small cavity.

Preferably, a step of adjusting the flow rate of the gas to be tested is also included; the detection device further includes a first control gas valve, and the flow rate of the gas to be tested is adjusted by controlling the first control gas valve.

Preferably, the step of cleaning the detection device is also included;

The detection device also includes a cleaning device. The cleaning device includes a second control air valve and a second air guide tube. One end of the second air guide tube is connected to the detection device, and the other end is placed in the air. By controlling the The second control air valve inputs air to the detection device to remove the gas to be tested in the detection device.

Preferably, the temperature range of the constant temperature water bath in the constant temperature water bath box is 30-60°C.

Preferably, the flow rate of the inert gas in the inlet pipe is 30ml/min.

Compared with the prior art, since the technical scheme of the present invention uses a plurality of gas sensors to detect the volatilized linseed oil gas, the quality of the linseed oil can be quickly and accurately identified; at the same time, the standard sensor data is processed by a specific algorithm Processing, thereby greatly improving the accuracy of quality detection of linseed oil.

Description of drawings

Fig. 1 is a system block diagram of a method for improving the quality detection accuracy of linseed oil of the present invention;

Fig. 2 is a system architecture diagram for realizing a method for improving the quality detection accuracy of linseed oil of the present invention;

Fig. 3 realizes the data flow diagram of identification system in a kind of method for improving the quality detection accuracy of linseed oil of the present invention;

Fig. 4 is the structural diagram of detection device in the method for realizing a kind of improving linseed oil quality detection precision of the present invention;

Fig. 5 is a system architecture diagram of another embodiment of a method for improving the detection accuracy of linseed oil quality in the present invention.

The present invention will be further described in conjunction with the accompanying drawings and specific embodiments as follows.

detailed description

The present invention will be further described below in conjunction with embodiment.

Nowadays, the use of gas sensor is more and more common. The same gas sensor can produce different responses to many different gases. After pattern recognition and data analysis, the gas type can be identified, and the response of these sensors The time is short, the recovery is relatively fast, and it can be used repeatedly.

On the basis of in-depth research on the prior art, the applicant proposes a linseed oil detection method with a gas sensor array as the core and supplemented by pattern recognition and data analysis. The "aroma" of linseed oil contains various gas components, and the gas components or amounts volatilized by different types of linseed oil are different to a certain extent. According to these characteristics, the gas sensor array can judge the quality of linseed oil by distinguishing the "aroma" of linseed oil.

Referring to Fig. 1 and Fig. 2, it is shown that a kind of realization system block diagram of the method for improving the quality detection accuracy of linseed oil of the present invention, it comprises the air inlet device 1 that is used to volatilize linseed oil into gaseous state and is used to detect linseed oil Oil component detection device 2.

Wherein, the air intake device 1 further includes an airtight container 11 for holding linseed oil, an air inlet pipe 12, a first air guide pipe 13 and a constant temperature water bath 14, and the constant temperature water bath 14 is used to make the linseed oil contained in the airtight container be at The state of constant temperature helps the volatilization of linseed oil and greatly increases the stripping ability of the purge gas. In a preferred embodiment, the temperature range of the constant temperature water bath in the constant temperature water bath box is 30-60°C, and 50°C is the most preferred.

One end of the air inlet pipe 12 extends into the airtight container 11 and is immersed in the sample linseed oil, and the other end is fed with pure inert gas to volatilize the linseed oil; the inert gas is continuously fed in at a certain flow rate after being dried and purified by silica gel, molecular sieve and activated carbon The air inlet pipe 12 enters the sample linseed oil contained in the airtight container 11, so that the sample linseed oil volatilizes its odor.

One end of the first air guide tube 13 extends into the air of the airtight container 11 and the other end thereof is connected to the detection device 2 to deliver the gas to be tested to the detection device 2 . Since the inlet pipe 12 is continuously fed with inert gas, preferably, the flow rate of the inert gas is 30ml/min. At this time, the air in the airtight container 11 is full of volatilized linseed oil to form the gas to be tested.

The detection circuit board 23 set in the detection device 2 is connected to the identification system 3 through the communication module 236. The identification system 3 is a computer loaded with a specific algorithm program, its computing performance and storage capacity are extremely superior, and it can store a very complete standard database. , by processing the collected data through a specific algorithm, the measurement accuracy is greatly improved.

Referring to FIG. 3 , it shows the flow chart of data processing in the recognition system 3. First, it is judged whether it is in the training mode. Only in the non-training mode, it enters the normal data detection process, otherwise it enters the data training process.

The data detection process mainly includes the following steps:

(1) Obtain the collected sensor data;

(2) Read the template data that has been trained;

(3) Identify and match the collected data with the template data;

(4) Get the quality testing results of linseed oil.

The data training process mainly includes the following steps:

S1: The recognition system collects sensor data of standard linseed oil, and initializes and normalizes part of the stable original data, namely and transform the above formula into In order to avoid the denominator being zero pairs; where x is the original data, y is the normalized data, x _max is the maximum value in the original data, and x _min is the minimum value in the original data;

S2: By calculating the input and output values of each node and calculating the error with the target output; among them, calculating the input value of the hidden layer, using the s function As an activation function, the formula is Among them, h _i is the input value of the hidden layer, W _ih is the weight value from the input layer to the hidden layer _, and Xi is the input data; after obtaining the input value of the hidden layer, calculate the output value of the hidden layer h _o =f( h _i ), similarly calculate the input and output values of the output layer; finally use the mean square error formula Obtain the error e, where Y is the target output value, and y _o is the real output of the output layer;

S3: Constantly update the weight until the error is within the acceptable range; wherein, when updating the weight, use the formula W=W _k-1 +ΔW+a[W _k-1 -W _k-2 ], the updating process is to use The error is derived from the weights of the input and output layers to obtain extreme values, including:

The error is derived from the weights of the hidden layer and the output layer to obtain the formula:

$\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ;</span>$

And the error is derived from the input and hidden layer weights to get the formula:

$\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; e</span>}}{{<span\; class="notranslate">\; \&part;</span>}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; h</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span><span\; class="notranslate">\; \&lsqb;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; *</span><span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

S4: Finally, pre-store the weights and extremums obtained through the above steps S1 to S3 as template data.

Wherein, the step of obtaining the collected sensor data further includes the following steps:

The linseed oil of the sample to be tested is put into an airtight container 11, and the airtight container 11 is put into a constant temperature water bath box 14 for heating in a constant temperature water bath;

Pass into the inert gas with stable flow rate in the airtight container 11 to carry out purging;

Transport the gas to be tested to the detection device 2;

The detection device 2 collects the gas to be measured through the gas sensor array to obtain sensor data and sends the sensor data to the identification system 3;

Wherein, the pre-stored template data is obtained through the following steps:

Referring to Fig. 4, shown is the structural block diagram of detecting device 2 in the present invention, and detecting device 2 further comprises air chamber body 21, is used for sealing the sealing cover 22 of air chamber body and is arranged on the sealing cover detection circuit board 23, detection circuit The board 23 includes a gas sensor array 231, a signal processing module 232, a control module 233, a storage module 234 and a touch display module 235;

The central position of the closing cover 22 is provided with an air intake hole 24, and the detection circuit board 23 is also provided with a corresponding air intake hole 24 at the coincident position of the central position. The first air guide pipe 13 is inserted into the air intake hole 24 and extends into the air chamber body 21. The first air guide tube 13 and the air intake hole 24 form an engaging and fixed structure;

A plurality of gas sensors in the gas sensor array 231 are arranged on the detection circuit board 23 at equal intervals with the air inlet 24 as the center, and are used to detect the composition of linseed oil in the gas to be measured and induce corresponding electrical signals , when the sealing cover 22 seals the air chamber body, a small cavity is formed, and a plurality of gas sensors are in full contact with the air in the cavity; the concentration of the gas to be measured is often very small. Being in sufficient contact with the sensor array 231 will affect the detection accuracy. Therefore, by adopting a small cavity design, the gas to be measured can immediately fill the entire cavity space, so that it can quickly and fully contact with the sensor array 231 . At the same time, since multiple gas sensors are arranged on the detection circuit board at equal intervals with the air inlet as the center, the contact time and probability of each gas sensor with the gas to be measured are almost exactly the same, thereby greatly improving the detection accuracy.

In a preferred embodiment, the air chamber body, the sealing cover and the detection circuit board are all circular, the air inlet is arranged at the center of the closure cover, and the gas sensors are arranged at equal intervals along the outer circumference of the detection circuit board. And the arrangement can further save space and form a smaller airtight chamber. At the same time, the gas to be measured is sent in from the central hole, which can make the gas sensor and the gas to be measured contact more uniformly, further reducing the measurement error.

In a preferred embodiment, with the detection circuit board 23 as the base surface, the height of the gas sensor is between 1/3 and 2/3 of the small cavity, and the gas to be measured can quickly fill the airtight chamber and connect with the gas sensor. The sensor is fully engaged.

Vent holes 25 are also provided on the closure cover 22 for maintaining pressure balance inside and outside the cavity. The airtight chamber is not a real airtight space, otherwise the pressure in the airtight chamber will continue to increase with the continuous introduction of the gas to be measured. By setting the vent hole 25, when the pressure in the airtight chamber reaches a certain level, the pressure in the airtight chamber can 25 to discharge the gas to be measured, so as to maintain the air pressure balance inside and outside the cavity. In order to allow the airtight chamber to be filled with the gas to be measured quickly while maintaining the pressure balance inside and outside the airtight chamber, the aperture of the vent hole 25 is smaller than the aperture of the air inlet hole 24, because if the aperture of the vent hole 25 is too large, the gas to be measured will Quickly discharge the airtight chamber, affecting the measurement accuracy. Preferably, the diameter of the vent hole 25 is generally between 1/3 and 1/5 of the diameter of the air inlet hole 24 .

In the detection circuit board 23, the signal processing module 232 performs signal processing on the electrical signal and sends the processed signal to the control module 233;

The storage module 234 is used to store the linseed oil template data in advance;

The control module 233 obtains the sampling data through signal sampling and matches with the pre-stored template data to obtain the quality detection result information of linseed oil.

The touch display module 235 is used for setting system parameters and displaying information about the quality inspection results of linseed oil.

In a preferred embodiment, the control module adopts a single-chip microcomputer modeled as STM32F103x. The 12-bit ADC of the STM32F103x is a successive approximation analog-to-digital converter. It has 18 channels and can measure 16 external and 2 internal signal sources. A/D conversion of each channel can be performed in one-shot, continuous, scan or intermittent mode. The present invention uses 10 channels in ADC1 as signal channels, opens the continuous conversion mode and scanning mode, opens up 10 memory spaces in the DMA as data storage, and reads the data from the DMA after each conversion is completed to operate.

In a preferred embodiment, the memory used by the storage module is an SD card, and the SD card is read and written through the SDIO interface of the STM32, and the data is saved in the SD card. Initialize the SD card, and synchronize the SDIO clock SDIO_SCK with the SD card. The operation on the SD card is often a large amount of data throughput, so DMA can be used to improve efficiency. All SDIO commands and command responses are transmitted through the SDIO-CMD pin, and the commands can only be issued by the main control chip. The remaining 4 pins are data ports for reading and writing data.

In a preferred embodiment, the display screen used in the present invention is a 3.2-inch TFT touch color screen. The principle of the touch screen is actually very simple, that is, the stylus will generate a corresponding voltage when it touches the touch screen, and a voltage will be generated after AD conversion. Data, by judging the data and then generating instructions, the system can be controlled. The touch screen control chip is ADS7853, which is a serial interface chip with built-in 12-bit analog-to-digital conversion and low on-resistance analog switch. Its reference VREF is often the same as the STM32 peripheral ADC.

The volatile substances of linseed oil are mainly composed of aldehydes, olefins, alkanes, ketones, and esters. Different varieties and different quality grades of linseed oil contain different components and contents. Therefore, the determination of the above components , the quality of linseed oil can be identified. In view of the above characteristics, the sensor array uses formaldehyde sensor MQ138, combustible gas sensor MQ-5, TGS2611, TGS813, hydrogen sulfide sensor MQ136, TGS825, alcohol sensor TGS822, ammonia sensor TGS826, air quality sensor TGS2602, and ammonia, sulfide , MQ135 sensor with high sensitivity to benzene series vapor. Among them, the sensor MQ138 has high sensitivity to aldehydes, alcohols, ketones, and aromatic compounds; MQ135 is used to detect ammonia, toluene, and hydrogen; MQ-5 is used to detect butane, propane, and methane; MQ136 is used to detect sulfide Hydrogen gas; TGS2602 is used to detect air pollutants; TGS2611 is used to detect methane and natural gas; TGS825 is used to detect hydrogen sulfide gas; TGS822 is used to detect alcohol, organic solvents, etc.; TGS813 is used to detect methane, ethane, propane and other flammability Gas; TGS826 is used to detect ammonia. Through the configuration of the gas sensor array, the quality of various linseed oils can be accurately detected.

Referring to Fig. 5, it shows a system architecture diagram of another embodiment of a method for improving the detection accuracy of linseed oil quality according to the present invention, the detection device 2 also includes a first control air valve 26, and the first control air valve 26 is controlled by Controlled by the control module 233, it is used to adjust the flow rate of the gas to be tested inputted into the first air duct, and can calibrate the detection result in combination with the flow rate information, thereby improving the measurement accuracy.

Further, the detection device 2 also includes a cleaning device 27. The cleaning device 27 includes a second control air valve 28 and a second air guide tube 29. One end of the second air guide tube 29 is connected to the detection device 2, and the other end is placed in the air. The control air valve 28 is controlled by the control module 233, and is used to control the second air guide tube to input air into the gas chamber body so as to remove the gas to be measured in the cavity. Before each measurement of the gas to be measured, the airtight gas chamber is cleaned, so that the accuracy of each measurement result can be improved.

The communication module 236 can use a serial port communication module. USART two-way communication requires at least two pins, RX is connected to the TX on the other end, and TX is connected to the RX on the other end. There are multiple serial ports in the STM32. In this system design, in order to prevent the pins from For multiplexing, USART1 and USART3 are used. USART1 is used to interact with the PC side, including the transmission of upper computer instructions and lower data transmission, receiving the string sent by the upper computer, and judging whether the string meets the preset instructions of the lower computer, and if it does, control according to the instruction , if not, wait for the next reception; and USART3 communicates with the Bluetooth serial port on the mobile phone to facilitate the control of the PWM signal and data display of the microcontroller, which is roughly the same as the touch module. The serial port baud rate is set to 115200, and the sending and receiving interrupt is enabled.

The descriptions of the above embodiments are only used to help understand the method and core idea of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, some improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. one kind is improved the method for linseed oil Quality Detection precision, it is characterised in that, comprise the following steps:

(1) by diffuser, this gas to be measured for gas to be measured and is inputed to detection device by linseed oil to be measured volatilization;

(2) detect device to carry out gas to be measured gathering by gas sensor array and obtain sensing data and this sensing data is sent to recognition system;

(3) sensing data gathered and the template data prestored are carried out drawing linseed oil Quality Detection result after identification is mated by recognition system;

Wherein, obtained by following step described in the template data that prestores:

S1: recognition system gathers the sensing data of standard linseed oil, partially stabilized raw data carries out initialize and normalized, namelyAnd above-mentioned formula is deformed intoThus avoid denominator to be zero right; Wherein x is raw data, and y is the data after standardization, x_maxFor maximum value in raw data, x_minFor minimum value in raw data;

S2: by calculate each node input and output value and and target output carry out Error Calculation; Wherein, calculate hidden layer input value, adopt s functionAs activation function, formula isWherein h_iFor the input value of hidden layer, W_ihFor input layer is to the weights of hidden layer, X_iFor input data; The output value h of hidden layer is calculated after obtaining the input value of hidden layer_o=f (h_i), the input and output value of output layer is calculated with reason; Finally utilize square root of the variance formulaTrying to achieve error e, wherein, Y is target output value, y_oIt is the true output of output layer;

S3: constantly update weight until error is in tolerance interval; Wherein, formula W=W is utilized when upgrading weights_k-1+��W+a[W_k-1-W_k-2], the process of renewal is exactly ask to lead to each weights of input and output layer by error respectively to obtain extreme value, specifically comprises:

Hidden layer and output layer weights are asked to lead and are obtained formula by error:

$\frac{\&part;e}{\&part;w_{ho}}=\frac{\&part;e}{\&part;y_{io}}*\frac{\&part;y_{io}}{\&part;w_{ho}}=\frac{\&part;\&lsqb;\frac{1}{2}*\mathrm{\&Sigma;}{(Y-y_o)}^2\&rsqb;}{\&part;y_{io}}*h_o=\&lsqb;f\left(y_i\right)-Y\&rsqb;*f\left(y_i\right)*h_o;$

And input and hidden layer weights asked to lead and obtain formula by error:

$\frac{\&part;e}{{\&part;}_{ih}}=\frac{\&part;\&lsqb;\frac{1}{2}*\mathrm{\&Sigma;}{(Y-y_o)}^2\&rsqb;}{\&part;h_o}*\frac{\&part;h_o}{\&part;h_i}=\frac{\&part;\&lsqb;\frac{1}{2}*\mathrm{\&Sigma;}{\&lsqb;Y-f\left(y_i\right)\&rsqb;}^2\&rsqb;}{\&part;h_o}*\frac{\&part;f\left(h_i\right)}{\&part;h_i}=\&lsqb;f\left(y_i\right)-Y\&rsqb;*\frac{\&part;f\left(y_i\right)}{\&part;h_o}*W_{ho}*\&part;f\left(h_i\right);$

S4: finally being trained the weights obtained and extreme value to prestore as template data by above-mentioned steps S1 to S3.

2. the method for raising linseed oil Quality Detection precision according to claim 1, it is characterised in that, further comprising the steps in described step (1):

Testing sample linseed oil is put into encloses container, and this encloses container is put into constant water bath box water bath with thermostatic control heating;

In encloses container, lead to the rare gas element into flow speed stability carry out blowing sweeping;

Gas to be measured is defeated by detection device.

3. the method for raising linseed oil Quality Detection precision according to claim 1, it is characterized in that, described diffuser comprises the encloses container for splendid attire linseed oil, inlet pipe, the first airway and constant water bath box, and for making, the linseed oil of described encloses container institute splendid attire is in temperature constant state to described constant water bath box; Described inlet pipe one end is stretched into described encloses container and is immersed in linseed oil and its other end leads to and into pure rare gas element, linseed oil volatilized; Described first airway one end is stretched in described encloses container air and its other end is connected with described detection device is delivered to gas to be measured in described detection device.

4. the method for raising linseed oil Quality Detection precision according to claim 1, it is characterized in that, described detection device comprises air chamber body, covers detection line plate for closing the closed lid of described air chamber body and be arranged on described closing, and described detection line plate comprises gas sensor array, signal processing module, control module, memory module and touch display module;

The central position of described closed lid is provided with air inlet port, described detection line plate also arranges corresponding air inlet port to this central position overlapping position, described first airway accesses in described air inlet port and stretches into air chamber body, and described first airway and described air inlet port form engaging fixing structure;

Multiple gas sensors in described gas sensor array centered by described air inlet port etc. spacing circular arrangement on described detection line plate, for detecting the composition of linseed oil in gas to be measured and respond to the corresponding electrical signal of generation, forming small-sized cavity when described closed lid closes described air chamber body, described multiple gas sensor fully contacts with the air in described cavity;

Described closed covering also is provided with breather hole, and for keeping the pressure equilibrium inside and outside described cavity, the aperture of described breather hole is less than the aperture of described air inlet port;

Described electrical signal is carried out signal processing and the signal after process is sent to described control module by described signal processing module;

Described memory module is used for prestoring linseed oil template data;

Described control module is obtained sampling data by signal sampling and is carried out mating with the template data prestored and draws linseed oil Quality Detection result information;

Described touch display module is for setting system parameter and display linseed oil Quality Detection result information.

5. the method for raising linseed oil Quality Detection precision according to claim 4, it is characterized in that, described gas sensor array comprises 10 gas sensors, spacing circular arrangement is on described detection line plate centered by described air inlet port etc. for described 10 gas sensors, and described 10 gas sensor models are respectively MQ138, MQ135, MQ-5, MQ136, TGS2602, TGS2611, TGS825, TGS822, TGS813, TGS826.

6. the method for raising linseed oil Quality Detection precision according to claim 3, it is characterized in that, described air chamber body, closed lid and detection line plate are circle, described air inlet port is arranged on the circle centre position of described closed lid, and described gas sensor is along spacing arrangements such as described detection line plate peripheral circumferential;

Taking described detection line plate as base face, the height of described gas sensor is between the 1/3 to 2/3 of described small-sized cavity.

7. the method for raising linseed oil Quality Detection precision according to claim 2, it is characterised in that, also comprise the step of the flow regulating gas to be measured; Described detection device also comprises the first control air valve, regulates the flow of gas to be measured by controlling described first control air valve.

8. the method for raising linseed oil Quality Detection precision according to claim 1, it is characterised in that, also comprise the step that described detection device cleans;

Described detection device also comprises washing unit, described washing unit comprises the 2nd control air valve and the 2nd airway, described detection device is accessed in described 2nd airway one end, the other end is placed in air, to described detection device input air thus removes the gas to be measured in described detection device by controlling described 2nd control air valve.

9. the method for raising linseed oil Quality Detection precision according to claim 3, it is characterised in that, in described constant water bath box, water bath with thermostatic control temperature range is 30-60 DEG C.

10. the method for raising linseed oil Quality Detection precision according to claim 3, it is characterised in that, in described inlet pipe, the flow velocity of rare gas element is 30ml/min.