Electronic nose-based method and system for detecting odor of bamboo powder for feeding

Abstract

The application discloses a kind of based on electronic nose's bamboo powder smell detection method and system for feeding, it is related to feed smell detection technical field, comprising: obtaining bamboo powder for feeding to be measured and normal reference bamboo powder for feeding, collection gas-sensitive channel response and form dynamic odor characteristic vector;According to the reference characteristic center of adjacent processing state, normal response change direction and normal deviation range, construct normal smell evolution benchmark;The bamboo powder for feeding to be measured is projected along each evolution section and the target evolution section is determined, and the feature change is decomposed into processing state migration component and residual component;Compare residual component with normal deviation range, determine abnormal smell deviation and abnormal gas-sensitive channel;Combined with the deviation direction, the deviation degree, the deviation direction combination, the deviation degree relationship and the difference sensitive relationship of abnormal gas-sensitive channel, determine abnormal smell direction.The application has the advantages that normal processing smell migration and abnormal smell deviation can be distinguished, and the accuracy of abnormal smell detection is improved.

Description

Technical Field

[0001] This invention relates to the field of feed odor detection technology, specifically to a method and system for detecting the odor of bamboo powder used in feed based on an electronic nose. Background Technology

[0002] As a non-grain feed ingredient with development potential, bamboo powder for animal feed exhibits odor characteristics that change with processing processes such as enzymatic hydrolysis and fermentation, reflecting differences in volatile components and processing states. Electronic noses utilize multiple gas-sensitive channels to respond to volatile gases. By collecting and analyzing odor information from bamboo powder in different processing states, they can provide a data foundation for raw material screening, processing evaluation, quality monitoring, and abnormal odor identification, which is of great significance for improving the consistency of quality evaluation of bamboo powder for animal feed.

[0003] Existing methods and systems for detecting the odor of bamboo powder used in animal feed based on electronic noses mostly rely on the overall response difference between the test sample and the reference sample to determine the state. Odor migration caused by normal enzymatic hydrolysis and fermentation and response shifts caused by abnormal odors can easily manifest as characteristic changes at the same time. It is difficult to distinguish between normal processing state migration, normal sample dispersion and true abnormal shifts. It is also difficult to further locate abnormal gas-sensitive channels and determine the direction of abnormal odors by combining the channel shift direction and degree relationship. Therefore, it is necessary to provide a method and system for detecting the odor of bamboo powder used in animal feed based on electronic noses to solve the above-mentioned problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides a method and system for detecting the odor of bamboo powder used in animal feed based on an electronic nose. This technical solution solves the problem mentioned in the background art that the existing state judgment based on the overall gas-sensitive response difference is difficult to distinguish between normal processing odor migration, normal sample dispersion and abnormal odor shift, which leads to confusion between normal processing changes and real abnormal responses, and makes it difficult to further locate abnormal gas-sensitive channels and determine the direction of abnormal odors.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for detecting the odor of bamboo powder used in animal feed based on an electronic nose includes: The test feed bamboo powder and the normal reference feed bamboo powder that is untreated, enzymatically hydrolyzed and fermented after enzymatic hydrolysis were obtained. Volatile gases were generated under the same headspace equilibrium conditions. The dynamic response sequence of each gas-sensitive channel in the detection and cleaning stages was continuously collected using an electronic nose, and the dynamic response characteristics of each gas-sensitive channel were extracted to form a dynamic odor feature vector. According to the processing sequence of untreated, enzymatic hydrolysis and post-enzymatic hydrolysis fermentation, the starting reference feature center, ending reference feature center, normal response change direction and normal deviation range of each evolution stage are determined based on the dynamic odor feature vector of normal reference feed bamboo powder in adjacent processing states. The normal odor evolution benchmark is formed by each evolution stage. The dynamic odor feature vector of the bamboo powder to be tested is projected along the normal response change direction of each evolution segment. The target evolution segment is determined according to the positional relationship and deviation between the projection position and the corresponding starting and ending reference feature centers. The response decomposition direction of the bamboo powder to be tested is determined by the normal response change direction of the target evolution segment. Using the initial reference feature center of the target evolution segment as the base point, the feature changes corresponding to the bamboo powder to be tested are decomposed into the processing state migration component along the response decomposition direction and the residual component deviating from the response decomposition direction. The degree of deviation represented by the residual component is compared with the normal deviation range of the target evolution segment. The residual component with a deviation degree exceeding the normal deviation range is identified as abnormal odor shift, and abnormal gas-sensitive channels are identified. Based on the direction and degree of abnormal odor shift in each abnormal gas-sensitive channel, the combination of shift directions and the degree of shift between abnormal gas-sensitive channels, and the differential sensitivity of each abnormal gas-sensitive channel to different types of volatile odor substances, the abnormal odor direction of the bamboo powder to be tested is determined, and the odor detection results of the bamboo powder to be tested are formed.

[0006] Preferably, a dynamic odor feature vector is formed, specifically including: The dynamic response sequences of the detection and cleaning stages are relativized using the baseline response of each gas-sensitive channel under clean air. At least two of the following are extracted from the dynamic response sequence of the detection phase: response amplitude, response rate of change, response accumulation, and response level at the end of the detection phase, to form response establishment features; and at least one of the following is extracted from the dynamic response sequence of the cleaning phase: response recovery amount, response recovery rate, and recovery endpoint deviation, to form response recovery features. For the bamboo powder used in the test feed and the bamboo powder used in the normal reference feed, the same response was selected to establish feature items and response recovery feature items. The extracted dynamic response features were scaled in the same way and combined in a consistent order of gas-sensitive channels and features to form a dynamic odor feature vector with consistent dimensions and consistent meaning of features corresponding to each dimension.

[0007] Preferably, the starting reference feature center, ending reference feature center, and normal response change direction of each evolutionary segment are determined, specifically including: The dynamic odor feature vectors of normal reference feed bamboo powder were grouped according to the processing state. The arithmetic mean of each dynamic odor feature vector under the same processing state on the same feature dimension was calculated. The arithmetic mean was combined according to the dimensional order of the dynamic odor feature vectors to obtain the reference feature centers corresponding to the untreated state, the enzymatic hydrolysis state, and the post-enzymatic hydrolysis fermentation state. The reference feature centers in the untreated state and the reference feature centers in the enzymatic hydrolysis state are respectively determined as the starting reference feature center and the ending reference feature center of the first evolution stage. The reference feature centers in the enzymatic hydrolysis state and the reference feature centers in the post-enzymatic hydrolysis fermentation state are respectively determined as the starting reference feature center and the ending reference feature center of the second evolution stage. The characteristic difference vector between the termination reference feature center and the starting reference feature center of each evolution segment is determined, and the characteristic difference vector is normalized. The normalized characteristic difference vector is then used as the normal response change direction of the corresponding evolution segment.

[0008] Preferably, the normal deviation range of each evolutionary stage is determined, specifically including: Using the initial reference feature center of each evolution segment as the base point, a reference evolution axis is established along the normal response change direction of the corresponding evolution segment; Project the dynamic odor feature vectors of each normal reference feed bamboo powder under the processing state at both ends of the corresponding evolution segment onto the reference evolution axis, and determine the orthogonal deviation distance between each dynamic odor feature vector and the corresponding projection position; Multiple orthogonal deviation distances corresponding to the same evolutionary segment are arranged in ascending order. The corresponding orthogonal deviation distance quantiles are determined according to the preset coverage ratio. The orthogonal deviation distance quantiles are determined as the upper limit of normal deviation for the corresponding evolutionary segment. The region where the orthogonal deviation distance is not greater than the upper limit of normal deviation is determined as the normal deviation range for the corresponding evolutionary segment.

[0009] Preferably, the target evolutionary segment is determined, specifically including: The dynamic odor feature vector of the bamboo powder to be tested is projected along the normal response change direction of each evolution segment. The relative position of the projection position in the corresponding evolution segment is determined based on the directed projection length of the projection position relative to the starting reference feature center of the corresponding evolution segment, and the distance between the starting reference feature center and the ending reference feature center of the corresponding evolution segment. Evolutionary segments with projected positions within the positional interval defined by the starting reference feature center and the ending reference feature center are identified as candidate evolutionary segments, and the degree of deviation between the dynamic odor feature vector of the bamboo powder to be tested and the corresponding projected positions of each candidate evolutionary segment is determined. If there is one candidate evolutionary segment, that candidate evolutionary segment is determined as the target evolutionary segment; if there are multiple candidate evolutionary segments, the candidate evolutionary segment with the smallest deviation is determined as the target evolutionary segment. When no candidate evolutionary segment exists, the degree of boundary deviation between each projection position and the nearest boundary of the corresponding position interval is determined, and the evolutionary segment with the smallest boundary deviation is determined as the initial evolutionary segment. When there are multiple preliminary evolutionary segments with the same degree of boundary deviation, the preliminary evolutionary segment with the smallest degree of deviation between the dynamic odor feature vector of the bamboo powder to be tested and the corresponding projection position is determined as the target evolutionary segment.

[0010] Preferably, the determination of the processing state transition component and the remaining component specifically includes: Determine the feature difference vector between the dynamic odor feature vector of the bamboo powder to be tested and the initial reference feature center of the target evolution segment; The projection vector of the feature difference vector along the response decomposition direction of the target evolution segment is determined as the processing state migration component. The processing state migration component is used to characterize the odor feature changes that can be explained by the normal processing state migration of the target evolution segment. The vector difference between the feature difference vector and the processing state migration component is determined as the residual component, and the degree of deviation represented by the residual component is determined according to the vector length of the residual component. The residual component is used to represent the odor feature changes that cannot be explained by the normal processing state migration of the target evolution segment.

[0011] Preferably, the abnormal gas-sensitive channel is identified, specifically including: According to the feature dimension corresponding to each gas-sensitive channel in the dynamic odor feature vector, the remaining components constituting the abnormal odor shift are divided into channel remaining sub-vectors corresponding to each gas-sensitive channel, and the channel shift degree of the corresponding gas-sensitive channel is determined according to the vector length of each channel remaining sub-vector. Using the starting reference feature center of the target evolution segment as the base point, the feature changes of each normal reference feed bamboo powder corresponding to the starting and ending processing states of the target evolution segment are decomposed into normal processing state migration components along the response decomposition direction and normal residual components deviating from the response decomposition direction. The normal channel offset degree corresponding to each normal residual component is determined according to the feature dimension corresponding to each gas-sensitive channel. Arrange the offset degrees of multiple normal channels corresponding to the same gas-sensitive channel in ascending order, determine the corresponding channel offset degree quantile according to the preset channel coverage ratio, determine the channel offset degree quantile as the upper limit of the normal channel offset of the corresponding gas-sensitive channel, and determine the range not greater than the upper limit of the normal channel offset as the normal channel offset range. Gas-sensitive channels whose channel deviation exceeds the upper limit of the corresponding normal channel deviation are identified as abnormal gas-sensitive channels.

[0012] Preferably, the determination of the deviation direction, deviation degree, and the relationship between the deviation direction and deviation degree among abnormal odor shifting channels in each abnormal gas-sensitive channel specifically includes: The channel residual sub-vectors corresponding to each abnormal gas-sensitive channel are normalized in direction. The channel residual sub-vectors after direction normalization are determined as the offset direction of the corresponding abnormal gas-sensitive channel, and the vector length of the channel residual sub-vectors before direction normalization is determined as the offset degree of the corresponding abnormal gas-sensitive channel. The offset directions of the abnormal gas-sensitive channels are combined according to their arrangement order in the dynamic odor feature vector to form a combination of offset directions between the abnormal gas-sensitive channels. The degree of deviation of different abnormal gas-sensitive channels is compared, and the relationship between the degree of deviation of different abnormal gas-sensitive channels is determined according to the ratio of the degree of deviation and the order of the degree of deviation.

[0013] Preferably, determining the direction of the abnormal odor of the bamboo powder to be tested specifically includes: Based on the sensitivity and response change direction of each gas-sensitive channel to different categories of volatile odor substances, a differential sensitivity relationship between gas-sensitive channels and categories of volatile odor substances is established. Based on the odor properties of different categories of volatile odor substances, a correlation relationship between categories of volatile odor substances and odor direction is established. Based on the differential sensitivity relationship between each abnormal gas-sensitive channel and the category of volatile odor substances, candidate volatile odor substance categories associated with abnormal odor shifts are screened, and the candidate abnormal odor directions are determined based on the correlation between each candidate volatile odor substance category and the odor direction. For each candidate abnormal odor direction, the category of volatile odor substances associated with the candidate abnormal odor direction is determined, and gas-sensitive channels that have a sensitive response to the determined volatile odor substance category are screened. Based on the sensitivity and response change direction of the screened gas-sensitive channels to the determined volatile odor substance category, the sensitive gas-sensitive channels, channel response directions, and relative response degrees between channels for the corresponding candidate abnormal odor direction are determined, forming a sensitive gas-sensitive channel response pattern. The abnormal gas-sensitive channels, the combination of the offset directions of the abnormal gas-sensitive channels, and the relationship of the offset degree are matched with the sensitive gas-sensitive channels, the channel response directions, and the relative response degree between channels corresponding to each candidate abnormal odor direction. The matching degree of each candidate abnormal odor direction is determined based on the degree of overlap between the abnormal gas-sensitive channels and the sensitive gas-sensitive channels, the degree of consistency between the offset direction and the channel response direction, and the degree of consistency between the offset degree relationship and the relative response degree between channels. The candidate abnormal odor direction with the highest matching degree is determined as the abnormal odor direction of the feed bamboo powder to be tested.

[0014] Furthermore, an electronic nose-based odor detection system for feed bamboo powder is proposed for performing the detection method described in any of the preceding claims, specifically including: The odor feature acquisition unit is used to acquire the volatile gases formed by the test feed bamboo powder and the untreated, enzymatically hydrolyzed and post-enzymatically fermented normal reference feed bamboo powder under the same headspace equilibrium conditions. The electronic nose continuously collects the dynamic response sequence of each gas-sensitive channel during the detection and cleaning stages, and extracts the dynamic response features of each gas-sensitive channel to form a dynamic odor feature vector. The evolutionary benchmark construction unit is used to determine the starting reference feature center, ending reference feature center, normal response change direction and normal deviation range of each evolutionary segment according to the processing sequence of untreated, enzymatic hydrolysis and post-enzymatic hydrolysis fermentation, based on the dynamic odor feature vector of normal reference feed bamboo powder in adjacent processing states. The normal odor evolution benchmark is formed by each evolutionary segment. The target evolution segment determination unit is used to project the dynamic odor feature vector of the bamboo powder to be tested along the normal response change direction of each evolution segment, determine the target evolution segment based on the positional relationship and deviation between the projection position and the corresponding starting and ending reference feature centers, and determine the response decomposition direction of the bamboo powder to be tested based on the normal response change direction of the target evolution segment. The response decomposition unit is used to decompose the characteristic changes corresponding to the tested feed bamboo powder into processing state migration components along the response decomposition direction and residual components deviating from the response decomposition direction, with the starting reference feature center of the target evolution segment as the base point. The degree of deviation represented by the residual components is compared with the normal deviation range of the target evolution segment, and the residual components with a degree of deviation exceeding the normal deviation range are identified as abnormal odor shifts. An abnormal channel determination unit is used to determine abnormal gas-sensitive channels based on abnormal odor deviation, and to determine the deviation direction, deviation degree, combination of deviation directions and deviation degree relationships between abnormal gas-sensitive channels in each abnormal gas-sensitive channel. The odor direction determination unit is used to determine the abnormal odor direction of the feed bamboo powder to be tested based on the offset direction and degree of the abnormal gas-sensitive channel, the combination of offset directions and degree of offset between abnormal gas-sensitive channels, and the differential sensitivity of abnormal gas-sensitive channels to different types of volatile odor substances, and to form the odor detection result of feed bamboo powder.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. The proposed method for detecting the odor of bamboo powder for feed based on an electronic nose obtains normal reference bamboo powder for feed that is untreated, enzymatically hydrolyzed, and fermented after enzymatic hydrolysis. According to the processing sequence, the starting reference feature center, ending reference feature center, normal response change direction, and normal deviation range of the evolution segment corresponding to adjacent processing states are determined to form a normal odor evolution benchmark. This realizes the continuous characterization of the odor response migration law during the normal processing of bamboo powder for feed, and enables the normal odor changes caused by different processing states to be compared in a unified dynamic odor feature space. 2. The proposed method for detecting the odor of bamboo powder for feed based on an electronic nose determines the target evolution segment by projecting the dynamic odor feature vector of the bamboo powder to be tested onto each evolution segment. The feature change is decomposed into a processing state migration component and a residual component based on the normal response change direction of the target evolution segment. The deviation degree represented by the residual component is then compared with the normal deviation range, which realizes the distinction between normal processing state migration, normal sample dispersion and abnormal odor shift. This reduces the impact of odor migration generated during normal enzymatic hydrolysis and fermentation on the detection results. 3. The proposed method for detecting the odor of bamboo powder for feed based on an electronic nose determines the abnormal gas-sensitive channel by identifying abnormal odor shifts. It then combines the shift direction, degree, combination of shift directions, and degree relationship of the abnormal gas-sensitive channel with the differences in sensitivity of different gas-sensitive channels to volatile odor substances to determine the abnormal odor direction of the bamboo powder to be tested. This achieves a step-by-step localization from the overall abnormal shift to the specific abnormal gas-sensitive channel and abnormal odor direction, enabling the odor detection results to further reflect the source characteristics and change direction of the abnormal response. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of a method for detecting the odor of bamboo powder for animal feed based on an electronic nose, as proposed in this invention. Figure 2 This is a schematic diagram of the dynamic odor feature vector formation process proposed in this invention; Figure 3 This is a schematic diagram of the process for constructing the normal odor evolution benchmark proposed in this invention; Figure 4 This is a schematic diagram of the target evolution segment determination and abnormal odor judgment process proposed in this invention; Figure 5 This is a schematic diagram of the framework of an electronic nose-based odor detection system for bamboo powder used in animal feed, as proposed in this invention. Detailed Implementation

[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. This embodiment is used to illustrate the technical principles and implementation process of the present invention and is not intended to limit the scope of protection of the present invention. Without changing the technical concept of constructing the normal odor evolution benchmark, determining the target evolution segment, separating the processing state migration component from the remaining component, determining the abnormal gas-sensitive channel, and determining the abnormal odor direction, the specific implementation parameters can be adjusted according to the type of electronic nose, the number of gas-sensitive channels, and the actual detection conditions of the feed bamboo powder.

[0018] Reference Figure 1As shown, in this embodiment, the bamboo powder to be tested, as well as untreated, enzymatically hydrolyzed, and post-enzymatically hydrolyzed fermented normal reference bamboo powder were obtained. The normal reference bamboo powder was a sample of bamboo powder with a clearly defined processing state and no abnormal odor to be detected. Multiple normal reference bamboo powders were selected for each processing state to reflect the allowable differences in odor response between normal samples under the corresponding processing state.

[0019] Following the processing sequence of bamboo powder for animal feed, the untreated state, enzymatically hydrolyzed state, and post-enzymatic fermentation state are sequentially considered as normal processing states. The odor change between the untreated and enzymatically hydrolyzed states forms the first evolutionary stage, and the odor change between the enzymatically hydrolyzed and post-enzymatic fermentation states forms the second evolutionary stage. Each evolutionary stage characterizes the normal electronic nose response migration between adjacent processing states, establishing a normal odor change process with a processing sequence relationship among the three processing states: untreated, enzymatically hydrolyzed, and post-enzymatic fermentation.

[0020] Specifically, untreated bamboo powder for animal feed mainly exhibits a distinct earthy and grassy odor; after enzymatic hydrolysis, the earthy odor is reduced and a fresh aroma emerges; further fermentation can develop odor characteristics such as a sweet and sour fragrance. This process of different processing states results in normal odor migration that can be characterized by changes in the response of multiple gas-sensitive channels in an electronic nose.

[0021] During electronic nose testing, the bamboo powder to be tested and each normal reference bamboo powder were placed in a headspace container, where volatile gases were formed under the same headspace equilibrium conditions. To ensure the comparability of electronic nose responses from different samples, the same sample mass, headspace volume, equilibrium temperature, and equilibrium time were used for each sample during the same normal odor evolution baseline construction and bamboo powder testing process.

[0022] In one specific embodiment, 2.00±0.05g of feed bamboo powder was weighed and placed in a 20mL headspace vial. After sealing, the vial was equilibrated at 40±2℃ for 30min to allow volatile odor substances in the feed bamboo powder to enter the headspace gas phase. Clean air was used as the carrier gas, with a flow rate set to 400mL / min. The electronic nose detection phase lasted for 120s, followed by a 60s cleaning with clean air after detection. Each feed bamboo powder sample was tested in triplicate, and dynamic response characteristics were extracted from each sample. The arithmetic mean of the three dynamic response characteristics of the same type was used as the dynamic response characteristic of the corresponding sample.

[0023] The electronic nose continuously records the response values ​​of each gas-sensitive channel over time during the detection and cleaning phases. Let the actual response of the i-th gas-sensitive channel at time t be... The baseline response of the i-th gas-sensitive channel under clean air conditions is: Then the relative response of the i-th gas-sensitive channel at time t for: ; In the formula, This is used to characterize the response change of the i-th gas-sensitive channel relative to a clean air reference state. Relativizing the response to a clean air reference reduces the impact of differences in initial response levels between different gas-sensitive channels on subsequent feature comparisons. For electronic nose gas-sensitive sensors with different output formats, corresponding reference calibration methods can be used based on the actual response mechanism; however, the tested bamboo powder and all normal reference bamboo powders are subjected to the same relativization method.

[0024] Reference Figure 2 As shown, when forming a dynamic odor feature vector, at least two of the following are extracted from the dynamic response sequence of the detection stage: response amplitude, response change rate, response accumulation, and response level at the end of the detection stage; and at least one of the following is extracted from the dynamic response sequence of the cleaning stage: response recovery amount, response recovery rate, and recovery endpoint deviation.

[0025] In this embodiment, the simultaneous extraction of response amplitude, response rate of change, cumulative response, end-of-detection response level, response recovery amount, response recovery rate, and recovery endpoint deviation is used as an example. The response amplitude is determined by the difference between the maximum relative response value and the initial relative response value of the detection phase; the response rate of change is determined by the ratio of the relative response change within a set time interval of the detection phase to the corresponding time interval; the cumulative response amount is determined by accumulating the change in response relative to the initial detection response at each sampling time point of the detection phase according to the sampling time interval, and is used to reflect the overall degree of change in the gas-sensitive channel response over time during the detection phase; the end-of-detection response level is determined by the relative response value at the end of the detection phase.

[0026] For the cleaning phase, the response recovery amount is determined by the change between the initial response level and the final response level of the cleaning phase; the response recovery rate is determined by the ratio of the change in the gas-sensitive channel response recovering to the clean air reference within a set time interval of the cleaning phase to the corresponding time interval; the recovery endpoint deviation is determined by the difference between the gas-sensitive channel response at the end of the cleaning phase and the clean air reference response. By simultaneously utilizing the dynamic response information from both the detection and cleaning phases, the response establishment process after volatile gases enter the electronic nose and the response recovery process after switching to clean air can be used together for the characterization of the odor of feed bamboo powder.

[0027] To ensure that normal reference bamboo powder and test bamboo powder are in the same feature space, during the same normal odor evolution benchmark construction and test sample detection process, the test bamboo powder, untreated normal reference bamboo powder, enzymatically hydrolyzed normal reference bamboo powder, and enzymatically hydrolyzed and fermented normal reference bamboo powder all use the same response to establish feature items, response to recover feature items, scale unification method, and feature arrangement order.

[0028] In this embodiment, a standardized approach is used for scale unification. Let the arithmetic mean of all normal reference bamboo powder for feed on the p-th feature dimension be... The standard deviation is The original feature value of any sample in the p-th feature dimension is Then the eigenvalues ​​after scaling for: ; In the formula, and The reference bamboo powder used to establish the normal odor evolution baseline was predetermined. The same standard was used to standardize the bamboo powder to be tested. and The standardized parameters are not re-determined based on the bamboo powder to be tested. When the standard deviation of a certain feature dimension is zero, that feature dimension can be uniformly deleted from the dynamic odor feature vector of all samples, or replaced by a preset non-zero minimum scale parameter, thereby maintaining the consistency of the feature dimensions of all dynamic odor feature vectors.

[0029] After scaling is standardized, the dynamic odor feature vectors of the corresponding samples are formed by combining them according to a fixed order of gas-sensitive channels and the order of dynamic response features within each gas-sensitive channel. For example, the dynamic response features of the first gas-sensitive channel are arranged first, then the dynamic response features of the second gas-sensitive channel are arranged, and so on, until the features of all gas-sensitive channels are arranged. By standardizing the feature construction method, the dynamic odor feature vectors formed by the tested bamboo powder and all normal reference bamboo powder have the same dimension, and the same feature dimension corresponds to the same dynamic response feature of the same gas-sensitive channel.

[0030] Reference Figure 3 As shown, when constructing a normal odor evolution baseline, normal reference feed bamboo powder was grouped according to its processing states: untreated, enzymatically hydrolyzed, and post-enzymatically hydrolyzed fermented. Let the processing state s have... There are j-th normal reference bamboo powder for feed, and the dynamic odor feature vector of the j-th normal reference bamboo powder is: Reference feature center corresponding to the processing state for: ; By calculating the arithmetic mean of the dynamic odor characteristics of multiple normal reference bamboo powder for feed under the same processing state in a dimension-by-dimensional manner, the reference feature centers corresponding to the untreated state, enzymatic hydrolysis state, and post-enzymatic hydrolysis fermentation state are obtained respectively.

[0031] The reference feature center of the untreated state is used as the starting reference feature center of the first evolution stage, and the reference feature center of the enzymatic hydrolysis state is used as the ending reference feature center of the first evolution stage; the reference feature center of the enzymatic hydrolysis state is used as the starting reference feature center of the second evolution stage, and the reference feature center of the fermentation state after enzymatic hydrolysis is used as the ending reference feature center of the second evolution stage.

[0032] For any evolutionary segment, the initial reference feature center is denoted as... The termination reference feature center is denoted as The direction of normal response change in the corresponding evolutionary stage for: ; The normal response change direction is used to characterize the overall change direction of the dynamic response characteristics of the multi-gas-sensitive channels during the process of bamboo powder for feed changing from the initial processing state to the final processing state of the corresponding evolutionary segment. If an effective feature difference cannot be formed between the initial reference feature center and the final reference feature center, the processing state of the normal reference bamboo powder for feed and the electronic nose detection data are verified, and the corresponding evolutionary segment is re-established.

[0033] Furthermore, a reference evolution axis is established along the normal response change direction, using the initial reference feature center of each evolutionary segment as the base point. For any normal reference feed bamboo powder under the processing states at both ends of the k-th evolutionary segment, its dynamic odor feature vector is denoted as... Corresponding projection position for: ; Orthogonal deviation distance between dynamic odor feature vector and corresponding projection position for: ; Orthogonal deviation distance is used to characterize the degree of characteristic dispersion of normal reference feed bamboo powder relative to the normal response change direction of the corresponding evolutionary stage.

[0034] The orthogonal deviation distances corresponding to the normal reference bamboo powder at both ends of the same evolutionary segment are arranged in ascending order. The quantiles of the orthogonal deviation distances are determined according to the preset coverage ratio, and the corresponding quantiles are used as the upper limit of normal deviation for the evolutionary segment. The feature space region where the orthogonal deviation distance is not greater than the upper limit of normal deviation constitutes the normal deviation range of the corresponding evolutionary segment, and the boundary corresponding to the upper limit of normal deviation is included in the normal deviation range.

[0035] In one specific embodiment, the preset coverage ratio can be set to 95%, that is, the 95th percentile corresponding to the orthogonal deviation distance set is used as the upper limit of normal deviation. The preset coverage ratio can also be predetermined based on the number of normal reference samples, the normal fluctuation degree of different production batches, and the stability of repeated testing by the electronic nose, and remain unchanged during the same normal odor evolution benchmark construction and the testing of the bamboo powder to be tested.

[0036] By establishing a normal deviation range based on the actual electronic nose response dispersion of normal reference bamboo powder for feed, the batch odor variation that is allowed under normal processing conditions can be preserved, avoiding direct judgment of abnormality based solely on the difference relative to the reference feature center of the processing state.

[0037] After establishing the normal odor evolution baseline, the dynamic odor feature vector of the bamboo powder to be tested is denoted as... The projections are made along the normal response change directions of the first and second evolutionary segments, respectively. For the k-th evolutionary segment, the relative position of the projected position of the tested bamboo powder in the corresponding evolutionary segment can be determined according to the following formula: ; when When the value is between 0 and 1, it indicates that the projection position of the bamboo powder to be tested in the corresponding normal response change direction is between the starting reference feature center and the ending reference feature center of the evolution segment, and the corresponding evolution segment is determined as a candidate evolution segment.

[0038] Further, the projection position of the bamboo powder to be tested on the reference evolution axis of each candidate evolution segment is determined according to the same projection method as the normal reference bamboo powder. The orthogonal distance between the dynamic odor feature vector of the bamboo powder to be tested and the corresponding projection position is calculated. This orthogonal distance is used to characterize the degree of deviation of the bamboo powder to be tested from the candidate evolution segment.

[0039] If there is one candidate evolutionary segment, that candidate evolutionary segment is determined as the target evolutionary segment; if there are multiple candidate evolutionary segments, the candidate evolutionary segment with the smallest deviation is determined as the target evolutionary segment.

[0040] When the projected position of the bamboo powder to be tested in all evolutionary stages is not within the position interval defined by the initial reference feature center and the final reference feature center, the distance from each projected position to the nearest boundary of the corresponding position interval is calculated, and this distance is used as the degree of boundary deviation. When the projected position is outside the initial reference feature center, the degree of boundary deviation is determined by the distance between the projected position and the initial reference feature center; when the projected position exceeds the final reference feature center, the degree of boundary deviation is determined by the distance between the projected position and the final reference feature center.

[0041] The evolutionary segment with the smallest boundary deviation is selected as the initial evolutionary segment. When there are multiple initial evolutionary segments with the same boundary deviation, the deviation of the dynamic odor feature vector of the bamboo powder to be tested from the reference evolutionary axis of each initial evolutionary segment is further compared, and the initial evolutionary segment with the smallest deviation is selected as the target evolutionary segment.

[0042] Through the above processing, the target evolution segment is determined based on the actual projection position and deviation of the dynamic odor characteristics of the electronic nose of the bamboo powder to be tested in the normal odor evolution benchmark, so that the subsequent abnormal odor judgment is based on the background of normal processing odor changes that match the actual response state of the bamboo powder to be tested.

[0043] After determining the target evolutionary segment, the initial reference feature center of the target evolutionary segment is used as the response decomposition base point, and the normal response change direction of the target evolutionary segment is used as the response decomposition direction. The difference between the dynamic odor feature vector of the tested bamboo powder and the initial reference feature center of the target evolutionary segment is determined as the feature difference vector. .

[0044] The projection vector of the feature difference vector along the direction of normal response change in the target evolution segment is determined as the processing state transition component M: ; In the formula, This indicates the direction of normal response change in the target evolutionary stage. The processing state migration component is used to characterize the portion of the odor characteristic change of the tested bamboo powder relative to the initial processing state of the target evolutionary stage that can be explained by the normal processing state migration.

[0045] The vector difference between the feature difference vector and the processing state transition component is determined as the residual component R: ; The residual component is used to characterize the portion of the odor characteristics change of the tested bamboo powder used for feed that deviates from the normal response change direction of the target evolutionary segment. The vector length of the residual component is used as the degree of deviation represented by the residual component, and this degree of deviation is compared with the upper limit of the normal deviation of the target evolutionary segment.

[0046] When the deviation represented by the remaining component is not greater than the upper limit of the normal deviation of the target evolution segment, it indicates that the part of the bamboo powder to be tested that deviates from the migration direction of the normal processing state is still within the discrete range allowed by the normal reference bamboo powder to be used, and the remaining component is not identified as an abnormal odor shift.

[0047] When the deviation represented by the residual component is greater than the upper limit of the normal deviation of the target evolutionary segment, it indicates that the bamboo powder to be tested has odor characteristic changes that cannot be explained by the migration of the normal processing state of the target evolutionary segment and the dispersion of the normal reference sample. The residual component is then identified as an abnormal odor shift.

[0048] By separating the processing state migration component from the remaining component, the odor response migration caused by the normal enzymatic hydrolysis and normal fermentation process itself is first separated as normal processing background. The response differences that are allowed between normal reference feed bamboo powder are further accommodated by the normal deviation range, thereby allowing odor feature changes that exceed the normal processing evolution background to enter the abnormal odor analysis process.

[0049] Reference Figure 4 As shown, after determining the abnormal odor shift, based on the correspondence between each feature dimension in the dynamic odor feature vector and the gas-sensitive channel, the remaining components constituting the abnormal odor shift are divided into channel remaining sub-vectors corresponding to each gas-sensitive channel.

[0050] For example, when a gas-sensitive channel corresponds to seven dynamic response features—response amplitude, response rate of change, response accumulation, response level at the end of the detection phase, response recovery amount, response recovery rate, and recovery endpoint deviation—the components in the remaining components corresponding to the above seven feature dimensions together form the channel remaining sub-vector of the gas-sensitive channel.

[0051] The length of the remaining subvector of the channel is used as the channel offset degree of the corresponding gas-sensitive channel. In order to determine whether the channel offset degree exceeds the channel fluctuation range allowed under normal processing conditions, the normal reference feed bamboo powder corresponding to the initial and final processing states of the target evolution segment is decomposed using the same response decomposition method as the feed bamboo powder to be tested, and the normal residual components are obtained respectively.

[0052] The normal residual components are further divided according to the characteristic dimensions corresponding to each gas-sensitive channel to obtain the normal channel residual sub-vectors corresponding to each normal reference feed bamboo powder, and the normal channel offset degree is determined by the vector length of the normal channel residual sub-vectors.

[0053] For the same gas-sensitive channel, the normal channel offset degrees corresponding to multiple normal reference bamboo powder feeds are arranged in ascending order. A quantile for the channel offset degree is determined based on a preset channel coverage ratio, and this quantile is used as the upper limit of the normal channel offset for the corresponding gas-sensitive channel. The range where the normal channel offset degree does not exceed the upper limit of the normal channel offset constitutes the normal channel offset range.

[0054] In one specific embodiment, the preset channel coverage ratio can be set to 95%. When the channel offset of a certain gas-sensitive channel of the bamboo powder to be tested is greater than the upper limit of the corresponding normal channel offset, the gas-sensitive channel is determined to be an abnormal gas-sensitive channel; when the channel offset is not greater than the upper limit of the corresponding normal channel offset, the offset of the gas-sensitive channel is determined to be within the normal channel offset range.

[0055] A two-level judgment process is formed by the overall normal deviation range and the normal deviation range of the gas-sensitive channel. First, abnormal odor deviations that cannot be explained by normal processing and evolution are identified from the overall dynamic odor feature space. Then, the abnormal gas-sensitive channels involved in the abnormal odor deviations are further determined.

[0056] After identifying the abnormal gas-sensitive channels, the direction of the channel residual sub-vector corresponding to each abnormal gas-sensitive channel is normalized. The direction-normalized channel residual sub-vector is determined as the offset direction of the corresponding abnormal gas-sensitive channel, and the vector length of the channel residual sub-vector before direction normalization is taken as the offset degree of the corresponding abnormal gas-sensitive channel.

[0057] Based on the arrangement order of the abnormal gas-sensitive channels in the dynamic odor feature vector, the offset directions of each abnormal gas-sensitive channel are combined to form a combination of offset directions among the abnormal gas-sensitive channels. The combination of offset directions is used to reflect the abnormal change direction of multiple abnormal gas-sensitive channels in their respective dynamic response feature spaces. Therefore, it can reflect not only the changes in the strength of the gas-sensitive channel response, but also the differences in dynamic changes during the response establishment and response recovery processes.

[0058] Further comparisons were made of the offset degrees of different abnormal gas-sensitive channels, and the ratios between the offset degrees of different abnormal gas-sensitive channels were calculated. These channels were then sorted according to the magnitude of their offset to establish the offset degree relationships among them. These offset degree relationships characterize the relative response strength of different abnormal gas-sensitive channels to changes in the current abnormal odor, providing information on the response relationships between multiple gas-sensitive channels for determining the direction of abnormal odors.

[0059] When determining the direction of abnormal odor in the bamboo powder to be tested, the differential sensitivity relationship between each gas-sensitive channel and different categories of volatile odor substances is established in advance. This embodiment uses an electronic nose containing multiple metal oxide semiconductor sensors. Different gas-sensitive channels can respond to alcohols, aldehydes, aromatic compounds, esters, ketones, nitrogen oxides, sulfur-containing organic compounds, alkenes, and alkanes to different degrees.

[0060] Specifically, representative volatile odor substances with distinct odor attributes can be selected to pre-calibrate each gas-sensitive channel. For example, linalool can be used to characterize volatile odor substances related to sweet or floral aromas, leaf alcohol can be used to characterize volatile odor substances related to grassy aromas, hexanal can be used to characterize volatile odor substances related to rancidity and oxidation, phenylethanol can be used to characterize volatile odor substances related to floral or honey aromas, ethyl acetate can be used to characterize volatile odor substances related to fruity aromas, and 2-heptanone can be used to characterize volatile odor substances related to fruity or fresh aromas.

[0061] During the calibration process, multiple concentration levels were set for the same representative volatile odor substance, and the dynamic response sequences of the gas-sensitive channels corresponding to each concentration level were continuously collected using an electronic nose. Based on the degree of change in the dynamic response characteristics of each gas-sensitive channel caused by changes in the concentration of the representative volatile odor substance, the sensitivity of each gas-sensitive channel to the corresponding volatile odor substance category was determined. Based on the trend of change in the dynamic response characteristics of each gas-sensitive channel with increasing concentration of the representative volatile odor substance, the direction of change in the response of each gas-sensitive channel to the corresponding volatile odor substance category was determined, thereby establishing the differential sensitivity relationship between the gas-sensitive channels and the volatile odor substance categories.

[0062] Simultaneously, a correlation between volatile odor substance categories and odor directions is established based on the odor attributes of different categories of volatile odor substances. For example, linalool-related volatile odor substances are associated with sweet or floral aromas, leaf alcohol-related volatile odor substances with grassy aromas, hexanal-related volatile odor substances with rancidity and oxidation, phenylethyl alcohol-related volatile odor substances with floral or honey aromas, ethyl acetate-related volatile odor substances with fruity aromas, and 2-heptanone-related volatile odor substances with fruity or fresh aromas.

[0063] After identifying the abnormal gas-sensitive channels of the bamboo powder to be tested, the categories of volatile odor substances that can elicit a sensitive response from the corresponding abnormal gas-sensitive channels are screened based on the differential sensitivity relationships between each abnormal gas-sensitive channel and different categories of volatile odor substances, forming candidate volatile odor substance categories. Then, based on the correlation between each candidate volatile odor substance category and the odor direction, multiple candidate abnormal odor directions are determined.

[0064] For each candidate abnormal odor direction, the category of volatile odor substances associated with that candidate abnormal odor direction is determined, and gas-sensitive channels that have a sensitive response to the corresponding volatile odor substance category are screened. Based on the sensitivity and response change direction of the screened gas-sensitive channels to the corresponding volatile odor substance category, the sensitive gas-sensitive channels, channel response directions, and relative response degrees between channels corresponding to the candidate abnormal odor direction are further determined, forming a sensitive gas-sensitive channel response pattern.

[0065] In this embodiment, the abnormal gas-sensitive channels actually determined by the bamboo powder to be tested are compared with the sensitive gas-sensitive channels corresponding to the directions of each candidate abnormal odor. The degree of overlap of the gas-sensitive channels is determined based on the number of overlapping gas-sensitive channels. The actual offset direction of each overlapping gas-sensitive channel is compared with the response change direction of the corresponding sensitive gas-sensitive channel. The degree of consistency of direction is determined based on the number of overlapping gas-sensitive channels with the same direction.

[0066] The actual magnitude and ratio of the offset between abnormal gas-sensitive channels are further compared with the relative response between the sensitive gas-sensitive channels corresponding to the candidate abnormal odor direction. The consistency of the offset relationship is determined based on the consistency of the response strength order and the difference between the actual offset ratio and the calibrated relative response ratio.

[0067] The degree of overlap, directional consistency, and offset of the gas-sensitive channels can be converted into values ​​ranging from 0 to 1. In one specific implementation, the arithmetic mean of the three consistency levels is used to determine the degree of matching of the candidate abnormal odor direction; alternatively, the weights corresponding to different consistency levels can be pre-determined based on the calibration results of known odor samples, and the matching degree can be determined through weighted processing. The same matching degree determination method is used for the same detection task.

[0068] After obtaining the matching degree of each candidate abnormal odor direction, the candidate abnormal odor direction with the highest matching degree is determined as the abnormal odor direction of the bamboo powder to be tested.

[0069] For example, when the abnormal gas-sensitive channels generated by the bamboo powder to be tested mainly include gas-sensitive channels with high sensitivity to volatile aldehyde odor substances, and the offset direction of each abnormal gas-sensitive channel is highly consistent with the response change direction of the gas-sensitive channel formed by hexanal calibration, and the offset degree relationship between each abnormal gas-sensitive channel is highly consistent with the relative response degree between the hexanal-corresponding sensitive gas-sensitive channels, the response mode of the sensitive gas-sensitive channel corresponding to the rancidity oxidation direction has a high degree of matching, and the abnormal odor direction of the bamboo powder to be tested can be determined as the rancidity oxidation direction.

[0070] When the combination of abnormal gas-sensitive channels, offset directions, and offset degrees has the highest matching degree with the response mode of the sensitive gas-sensitive channels corresponding to leaf alcohol, the abnormal odor direction can be determined as the direction of enhanced grassy odor. For other odor directions such as floral, fruity, and fresh scents, the same matching method for gas-sensitive channel response modes can be used for determination.

[0071] The above-mentioned method for determining the direction of abnormal odors does not directly determine the category of abnormal odors based on the high response of a single gas-sensitive channel. Instead, after excluding the influence of normal processing state migration and the discreteness of normal reference samples, the direction of abnormal odors is determined by multiple abnormal gas-sensitive channels, the offset direction of each abnormal gas-sensitive channel, and the relative offset degree between different abnormal gas-sensitive channels.

[0072] When generating the odor detection results for feed bamboo powder, if the deviation represented by the remaining component of the feed bamboo powder to be tested does not exceed the normal deviation range of the target evolutionary segment, a detection result is generated indicating that no abnormal odor deviation exceeding the normal deviation range has been detected.

[0073] When the deviation of the remaining components exceeds the normal deviation range of the target evolution segment, an odor detection result is formed, which includes the target evolution segment, abnormal odor deviation, abnormal gas-sensitive channels, the deviation direction and degree of each abnormal gas-sensitive channel, and the direction of the abnormal odor. The detection result can simultaneously characterize both the normal processing odor evolution background corresponding to the tested bamboo powder for feed and the abnormal odor changes that deviate from the normal processing odor evolution background.

[0074] Reference Figure 5 As shown, this embodiment also provides an odor detection system for bamboo powder used in animal feed based on an electronic nose, including an odor feature acquisition unit, an evolutionary benchmark construction unit, a target evolutionary segment determination unit, a response decomposition unit, an abnormal channel determination unit, and an odor direction determination unit.

[0075] The odor feature acquisition unit receives the dynamic response sequences continuously collected by each gas-sensitive channel of the electronic nose during the detection and cleaning phases. The dynamic response sequences are relativized using the baseline response under clean air, and response features and response recovery features are extracted. The scale of each dynamic response feature is unified, and a dynamic odor feature vector is formed according to a fixed gas-sensitive channel order and dynamic response feature order.

[0076] The evolutionary benchmark construction unit determines the reference feature center for the corresponding processing state based on the normal reference bamboo powder of untreated, enzymatically hydrolyzed and post-enzymatically hydrolyzed fermentation, respectively, forms the first evolutionary segment and the second evolutionary segment according to the processing sequence, and determines the normal response change direction and normal deviation range of each evolutionary segment. The first evolutionary segment and the second evolutionary segment together form the normal odor evolutionary benchmark.

[0077] The target evolution segment determination unit projects the dynamic odor feature vector of the bamboo powder to be tested along the normal response change direction of each evolution segment. Based on the relative position of the projection position between the corresponding starting reference feature center and the ending reference feature center, as well as the degree of deviation of the bamboo powder to be tested from each evolution segment, the target evolution segment is determined, and the normal response change direction of the target evolution segment is determined as the response decomposition direction.

[0078] The response decomposition unit takes the initial reference feature center of the target evolution segment as the base point, decomposes the feature changes of the tested feed bamboo powder relative to the initial reference feature center into processing state migration components and residual components, and determines the abnormal odor shift based on the comparison between the deviation degree represented by the residual component and the normal deviation range of the target evolution segment.

[0079] The abnormal channel determination unit divides the remaining components constituting the abnormal odor shift into multiple channel remaining sub-vectors according to the dynamic odor feature dimension corresponding to each gas-sensitive channel. It determines the abnormal gas-sensitive channel based on the channel shift degree corresponding to each channel remaining sub-vector and the normal channel shift range, and further determines the shift direction, shift degree, shift direction combination, and shift degree relationship of the abnormal gas-sensitive channel.

[0080] The odor direction determination unit determines candidate abnormal odor directions based on the difference sensitivity between gas-sensitive channels and volatile odor substance categories, as well as the correlation between volatile odor substance categories and odor directions. It then matches the response patterns of the sensitive gas-sensitive channels corresponding to each candidate abnormal odor direction with the actual abnormal gas-sensitive channels, offset direction combinations, and offset degree relationships generated by the bamboo powder to be tested. Based on the matching results, it determines the abnormal odor direction of the bamboo powder to be tested, thus forming the odor detection results of the bamboo powder to be tested.

[0081] In the specific implementation process, each functional unit can be implemented by a data processing device connected to the electronic nose data output interface, which executes the corresponding data processing program. The electronic nose completes the actual gas-sensitive response acquisition of volatile gases from feed bamboo powder. Based on the gas-sensitive response data acquired by the electronic nose, each functional unit sequentially completes the construction of dynamic odor characteristics, characterization of normal processing odor evolution, determination of target evolution segment, separation of normal processing state migration, location of abnormal gas-sensitive channels, and determination of abnormal odor direction, thereby forming a complete odor detection result for feed bamboo powder.

[0082] The advantages of this invention are as follows: By constructing a normal odor evolution benchmark corresponding to the untreated, enzymatically hydrolyzed, and post-enzymatically hydrolyzed fermentation states, the odor migration and abnormal responses deviating from the normal evolution direction generated during the normal processing of bamboo powder for feed are separated. Furthermore, the influence of normal sample dispersion is eliminated by combining the normal deviation range, ensuring that the determination of abnormal odors is based on the combined constraints of the processing state background and the normal fluctuation range. After identifying abnormal odor shifts, abnormal gas-sensitive channels can be further located, and the direction of abnormal odors can be determined by combining the shift direction, degree of shift, and differential sensitivity relationship between each abnormal gas-sensitive channel and different volatile odor substance categories. This improves the specificity, interpretability, and stability of odor detection results.

[0083] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for detecting the odor of bamboo powder used in animal feed based on an electronic nose, characterized in that, include: The test feed bamboo powder and the normal reference feed bamboo powder that is untreated, enzymatically hydrolyzed and fermented after enzymatic hydrolysis were obtained. Volatile gases were generated under the same headspace equilibrium conditions. The dynamic response sequence of each gas-sensitive channel in the detection and cleaning stages was continuously collected using an electronic nose, and the dynamic response characteristics of each gas-sensitive channel were extracted to form a dynamic odor feature vector. According to the processing sequence of untreated, enzymatic hydrolysis and post-enzymatic hydrolysis fermentation, the starting reference feature center, ending reference feature center, normal response change direction and normal deviation range of each evolution stage are determined based on the dynamic odor feature vector of normal reference feed bamboo powder in adjacent processing states. The normal odor evolution benchmark is formed by each evolution stage. The dynamic odor feature vector of the bamboo powder to be tested is projected along the normal response change direction of each evolution segment. The target evolution segment is determined according to the positional relationship and deviation between the projection position and the corresponding starting and ending reference feature centers. The response decomposition direction of the bamboo powder to be tested is determined by the normal response change direction of the target evolution segment. Using the initial reference feature center of the target evolution segment as the base point, the feature changes corresponding to the bamboo powder to be tested are decomposed into the processing state migration component along the response decomposition direction and the residual component deviating from the response decomposition direction. The degree of deviation represented by the residual component is compared with the normal deviation range of the target evolution segment. The residual component with a deviation degree exceeding the normal deviation range is identified as abnormal odor shift, and abnormal gas-sensitive channels are identified. Based on the direction and degree of abnormal odor shift in each abnormal gas-sensitive channel, the combination of shift directions and the degree of shift between abnormal gas-sensitive channels, and the differential sensitivity of each abnormal gas-sensitive channel to different types of volatile odor substances, the abnormal odor direction of the bamboo powder to be tested is determined, and the odor detection results of the bamboo powder to be tested are formed.

2. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 1, characterized in that, The formation of a dynamic odor feature vector specifically includes: The dynamic response sequences of the detection and cleaning stages are relativized using the baseline response of each gas-sensitive channel under clean air. At least two of the following are extracted from the dynamic response sequence of the detection phase: response amplitude, response rate of change, response accumulation, and response level at the end of the detection phase, to form response establishment features; and at least one of the following is extracted from the dynamic response sequence of the cleaning phase: response recovery amount, response recovery rate, and recovery endpoint deviation, to form response recovery features. For the bamboo powder used in the test feed and the bamboo powder used in the normal reference feed, the same response was selected to establish feature items and response recovery feature items. The extracted dynamic response features were scaled in the same way and combined in a consistent order of gas-sensitive channels and features to form a dynamic odor feature vector with consistent dimensions and consistent meaning of features corresponding to each dimension.

3. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 1, characterized in that, Determine the initial reference feature center, the final reference feature center, and the direction of normal response change for each evolutionary segment, specifically including: The dynamic odor feature vectors of normal reference feed bamboo powder were grouped according to the processing state. The arithmetic mean of each dynamic odor feature vector under the same processing state on the same feature dimension was calculated. The arithmetic mean was combined according to the dimensional order of the dynamic odor feature vectors to obtain the reference feature centers corresponding to the untreated state, the enzymatic hydrolysis state, and the post-enzymatic hydrolysis fermentation state. The reference feature centers in the untreated state and the reference feature centers in the enzymatic hydrolysis state are respectively determined as the starting reference feature center and the ending reference feature center of the first evolution stage. The reference feature centers in the enzymatic hydrolysis state and the reference feature centers in the post-enzymatic hydrolysis fermentation state are respectively determined as the starting reference feature center and the ending reference feature center of the second evolution stage. The characteristic difference vector between the termination reference feature center and the starting reference feature center of each evolution segment is determined, and the characteristic difference vector is normalized. The normalized characteristic difference vector is then used as the normal response change direction of the corresponding evolution segment.

4. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 3, characterized in that, Determine the normal deviation range for each evolutionary stage, specifically including: Using the initial reference feature center of each evolution segment as the base point, a reference evolution axis is established along the normal response change direction of the corresponding evolution segment; Project the dynamic odor feature vectors of each normal reference feed bamboo powder under the processing state at both ends of the corresponding evolution segment onto the reference evolution axis, and determine the orthogonal deviation distance between each dynamic odor feature vector and the corresponding projection position; Multiple orthogonal deviation distances corresponding to the same evolutionary segment are arranged in ascending order. The corresponding orthogonal deviation distance quantiles are determined according to the preset coverage ratio. The orthogonal deviation distance quantiles are determined as the upper limit of normal deviation for the corresponding evolutionary segment. The region where the orthogonal deviation distance is not greater than the upper limit of normal deviation is determined as the normal deviation range for the corresponding evolutionary segment.

5. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 1, characterized in that, Determine the target evolution stage, specifically including: The dynamic odor feature vector of the bamboo powder to be tested is projected along the normal response change direction of each evolution segment. The relative position of the projection position in the corresponding evolution segment is determined based on the directed projection length of the projection position relative to the starting reference feature center of the corresponding evolution segment, and the distance between the starting reference feature center and the ending reference feature center of the corresponding evolution segment. Evolutionary segments with projected positions within the positional interval defined by the starting reference feature center and the ending reference feature center are identified as candidate evolutionary segments, and the degree of deviation between the dynamic odor feature vector of the bamboo powder to be tested and the corresponding projected positions of each candidate evolutionary segment is determined. If there is one candidate evolutionary segment, that candidate evolutionary segment is determined as the target evolutionary segment; if there are multiple candidate evolutionary segments, the candidate evolutionary segment with the smallest deviation is determined as the target evolutionary segment. When no candidate evolutionary segment exists, the degree of boundary deviation between each projection position and the nearest boundary of the corresponding position interval is determined, and the evolutionary segment with the smallest boundary deviation is determined as the initial evolutionary segment. When there are multiple preliminary evolutionary segments with the same degree of boundary deviation, the preliminary evolutionary segment with the smallest degree of deviation between the dynamic odor feature vector of the bamboo powder to be tested and the corresponding projection position is determined as the target evolutionary segment.

6. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 1, characterized in that, The determination of the processing state transition components and residual components specifically includes: Determine the feature difference vector between the dynamic odor feature vector of the bamboo powder to be tested and the initial reference feature center of the target evolution segment; The projection vector of the feature difference vector along the response decomposition direction of the target evolution segment is determined as the processing state migration component. The processing state migration component is used to characterize the odor feature changes that can be explained by the normal processing state migration of the target evolution segment. The vector difference between the feature difference vector and the processing state migration component is determined as the residual component, and the degree of deviation represented by the residual component is determined according to the vector length of the residual component. The residual component is used to represent the odor feature changes that cannot be explained by the normal processing state migration of the target evolution segment.

7. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 6, characterized in that, Identifying abnormal gas-sensitive channels specifically includes: According to the feature dimension corresponding to each gas-sensitive channel in the dynamic odor feature vector, the remaining components constituting the abnormal odor shift are divided into channel remaining sub-vectors corresponding to each gas-sensitive channel, and the channel shift degree of the corresponding gas-sensitive channel is determined according to the vector length of each channel remaining sub-vector. Using the starting reference feature center of the target evolution segment as the base point, the feature changes of each normal reference feed bamboo powder corresponding to the starting and ending processing states of the target evolution segment are decomposed into normal processing state migration components along the response decomposition direction and normal residual components deviating from the response decomposition direction. The normal channel offset degree corresponding to each normal residual component is determined according to the feature dimension corresponding to each gas-sensitive channel. Arrange the offset degrees of multiple normal channels corresponding to the same gas-sensitive channel in ascending order, determine the corresponding channel offset degree quantile according to the preset channel coverage ratio, determine the channel offset degree quantile as the upper limit of the normal channel offset of the corresponding gas-sensitive channel, and determine the range not greater than the upper limit of the normal channel offset as the normal channel offset range. Gas-sensitive channels whose channel deviation exceeds the upper limit of the corresponding normal channel deviation are identified as abnormal gas-sensitive channels.

8. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 7, characterized in that, The determination of the direction and degree of deviation of abnormal odors in each abnormal gas-sensitive channel, as well as the combination of deviation directions and the relationship of deviation degrees between abnormal gas-sensitive channels, specifically includes: The channel residual sub-vectors corresponding to each abnormal gas-sensitive channel are normalized in direction. The channel residual sub-vectors after direction normalization are determined as the offset direction of the corresponding abnormal gas-sensitive channel, and the vector length of the channel residual sub-vectors before direction normalization is determined as the offset degree of the corresponding abnormal gas-sensitive channel. The offset directions of the abnormal gas-sensitive channels are combined according to their arrangement order in the dynamic odor feature vector to form a combination of offset directions between the abnormal gas-sensitive channels. The degree of deviation of different abnormal gas-sensitive channels is compared, and the relationship between the degree of deviation of different abnormal gas-sensitive channels is determined according to the ratio of the degree of deviation and the order of the degree of deviation.

9. The method for detecting the odor of bamboo powder used in animal feed based on an electronic nose according to claim 8, characterized in that, Determine the direction of any abnormal odor in the bamboo powder being tested, specifically including: Based on the sensitivity and response change direction of each gas-sensitive channel to different categories of volatile odor substances, a differential sensitivity relationship between gas-sensitive channels and categories of volatile odor substances is established. Based on the odor properties of different categories of volatile odor substances, a correlation relationship between categories of volatile odor substances and odor direction is established. Based on the differential sensitivity relationship between each abnormal gas-sensitive channel and the category of volatile odor substances, candidate volatile odor substance categories associated with abnormal odor shifts are screened, and the candidate abnormal odor directions are determined based on the correlation between each candidate volatile odor substance category and the odor direction. For each candidate abnormal odor direction, the category of volatile odor substances associated with the candidate abnormal odor direction is determined, and gas-sensitive channels that have a sensitive response to the determined volatile odor substance category are screened. Based on the sensitivity and response change direction of the screened gas-sensitive channels to the determined volatile odor substance category, the sensitive gas-sensitive channels, channel response directions, and relative response degrees between channels for the corresponding candidate abnormal odor direction are determined, forming a sensitive gas-sensitive channel response pattern. The abnormal gas-sensitive channels, the combination of the offset directions of the abnormal gas-sensitive channels, and the relationship of the offset degree are matched with the sensitive gas-sensitive channels, the channel response directions, and the relative response degree between channels corresponding to each candidate abnormal odor direction. The matching degree of each candidate abnormal odor direction is determined based on the degree of overlap between the abnormal gas-sensitive channels and the sensitive gas-sensitive channels, the degree of consistency between the offset direction and the channel response direction, and the degree of consistency between the offset degree relationship and the relative response degree between channels. The candidate abnormal odor direction with the highest matching degree is determined as the abnormal odor direction of the feed bamboo powder to be tested.

10. An electronic nose-based system for detecting the odor of bamboo powder used in animal feed, for performing the detection method as described in any one of claims 1-9, characterized in that, Specifically, it includes: The odor feature acquisition unit is used to acquire the volatile gases formed by the test feed bamboo powder and the untreated, enzymatically hydrolyzed and post-enzymatically fermented normal reference feed bamboo powder under the same headspace equilibrium conditions. The electronic nose continuously collects the dynamic response sequence of each gas-sensitive channel during the detection and cleaning stages, and extracts the dynamic response features of each gas-sensitive channel to form a dynamic odor feature vector. The evolutionary benchmark construction unit is used to determine the starting reference feature center, ending reference feature center, normal response change direction and normal deviation range of each evolutionary segment according to the processing sequence of untreated, enzymatic hydrolysis and post-enzymatic hydrolysis fermentation, based on the dynamic odor feature vector of normal reference feed bamboo powder in adjacent processing states. The normal odor evolution benchmark is formed by each evolutionary segment. The target evolution segment determination unit is used to project the dynamic odor feature vector of the bamboo powder to be tested along the normal response change direction of each evolution segment, determine the target evolution segment based on the positional relationship and deviation between the projection position and the corresponding starting and ending reference feature centers, and determine the response decomposition direction of the bamboo powder to be tested based on the normal response change direction of the target evolution segment. The response decomposition unit is used to decompose the characteristic changes corresponding to the tested feed bamboo powder into processing state migration components along the response decomposition direction and residual components deviating from the response decomposition direction, with the starting reference feature center of the target evolution segment as the base point. The degree of deviation represented by the residual components is compared with the normal deviation range of the target evolution segment, and the residual components with a degree of deviation exceeding the normal deviation range are identified as abnormal odor shifts. An abnormal channel determination unit is used to determine abnormal gas-sensitive channels based on abnormal odor deviation, and to determine the deviation direction, deviation degree, combination of deviation directions and deviation degree relationships between abnormal gas-sensitive channels in each abnormal gas-sensitive channel. The odor direction determination unit is used to determine the abnormal odor direction of the feed bamboo powder to be tested based on the offset direction and degree of the abnormal gas-sensitive channel, the combination of offset directions and degree of offset between abnormal gas-sensitive channels, and the differential sensitivity of abnormal gas-sensitive channels to different types of volatile odor substances, and to form the odor detection result of feed bamboo powder.

Images