Method for identifying asphalt mixture wrapping state based on stirring motor torque signal
Patent Information
- Application Number
- CN202611166192.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-03
- Publication Date
- 2026-09-25
AI Technical Summary
该方法虽然实时性强，但高度依赖操作人员的个人经验，缺乏客观量化标准，难以保证判定的稳定性和一致性
[0024]本发明的有益效果在于：本发明通过采集搅拌电机的扭矩时序信号，在搅拌过程中即可完成裹覆状态的识别与判定，无需等待实验室抽提试验结果；针对干拌阶段骨料与矿粉机械混合的特点，提取均值、有效值、峰值因子、近似熵等反映负载水平和冲击强度的特征；针对喷油阶段沥青雾化与附着的动态过程，提取瞬态响应时间、扭矩跃升值、调制深度等反映粘滑振荡特性的特征；针对湿拌阶段沥青膜均匀化的平稳过程，提取波形因子、低频能量占比、扭矩方差、峭度、主频能量集中度等反映裹覆平滑度和均匀性的特征。各阶段特征的物理意义明确、与裹覆状态的关联性强，避免了将三阶段统一处理导致的信息丢失问题。通过转移概率矩阵中的因果约束条件，利用了搅拌工艺固有的物理因果链，抑制了违反物理规律的异常状态跳变，降低了单段信号异常导致的误判风险，提高了判定结果的鲁棒性。
Smart Images
Figure CN122817840A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of asphalt production and manufacturing, and specifically to a method for identifying the coating state of asphalt mixtures based on the torque signal of a mixing motor. Background Technology
[0002] In the production process of asphalt mixtures, the coating condition, that is, the degree to which asphalt is evenly coated on the surface of aggregates, is a core indicator for measuring the quality of the mixture. Poorly coated mixtures (commonly known as "white aggregates") are prone to early defects such as loosening, peeling, and water damage after paving, affecting the durability of the pavement.
[0003] In existing technologies, during the production process, operators periodically take samples from the mixing drum outlet and send them to a laboratory to determine the asphalt content and coating rate using centrifugal extraction or combustion methods. While this method provides accurate results, it suffers from a significant time lag—it typically takes more than 30 minutes from sampling to obtaining the test results. During this time, the corresponding batch of mix has often already been loaded and transported to the construction site. If substandard coating is found, the already-delivered mix must be reworked or scrapped, resulting in a serious waste of materials and energy.
[0004] The second method is manual visual inspection. Experienced operators judge the coating status by observing fluctuations in the agitator ammeter, listening to the sound of the agitator motor, and checking the color and shape of the mixture at the discharge port. While this method offers high real-time performance, it heavily relies on the operator's personal experience, lacks objective quantitative standards, and makes it difficult to guarantee the stability and consistency of the judgment. Significant differences may exist in the judgment results between different shifts and between different operators. Summary of the Invention
[0005] The purpose of this invention is to provide a method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor, replacing manual identification.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor includes the following steps:
[0008] S1. Obtain the material operating conditions and material quality for the current batch.
[0009] S2. Based on the material operating condition information and material quality, obtain the torque benchmark feature model corresponding to the current operating condition from the benchmark feature database. The torque benchmark feature model includes benchmark feature vectors corresponding to the dry mixing stage, the oil injection stage, and the wet mixing stage. The benchmark feature vectors are established by statistically analyzing the torque signals of historical qualified batches and are used to characterize the ideal torque characteristics of each mixing stage under the corresponding operating condition. The dimensions of the benchmark feature vector for the dry mixing stage include the mean torque, the effective torque value, the peak factor, and the approximate entropy. The dimensions of the benchmark feature vector for the oil injection stage include the transient response time, the torque jump value, and the modulation depth. The dimensions of the benchmark feature vector for the wet mixing stage include the waveform factor, the proportion of low-frequency energy, the torque variance, the kurtosis, and the concentration of main frequency energy. The proportion of low-frequency energy refers to the ratio of the energy of the torque signal in the 0.5~20Hz frequency band to the energy of the entire frequency band.
[0010] S3. Collect the torque timing signal of the stirring motor of the current batch, and extract the time domain and frequency domain features of the torque signals of the dry mixing stage, the oil injection stage and the wet mixing stage respectively to obtain the feature vector of the torque acquisition signal of each mixing stage of the current batch.
[0011] S4. Calculate the feature deviation between the feature vector collected at each stage and the corresponding reference feature vector to obtain the deviation vector of the dry mixing stage, the deviation vector of the oil injection stage, and the deviation vector of the wet mixing stage, and standardize each deviation vector.
[0012] S5. Establish a stage state model corresponding to the mixing process. Using the standardized dry mixing stage deviation vector, oil injection stage deviation vector, and wet mixing stage deviation vector as observations, jointly infer the state of the dry mixing stage, oil injection stage, and wet mixing stage to obtain the state of each stage of the current batch and the final coating state. The stage state model satisfies the sequential constraints between the mixing process stages, so that the state of the later stage is constrained by the state of the previous stage.
[0013] S6. Based on the final coating state output by the stage state model and the corresponding judgment confidence level, determine the coating qualification of the current batch.
[0014] When it is determined that the coating of the current batch is unqualified and the confidence level of the determination reaches the first preset threshold, the stirring process parameters of the current batch are adjusted. The stirring process parameters include at least one or more of the following: wet mixing time, stirring mode, and stirring speed.
[0015] After compensation is completed, the torque timing signal is reacquired, and steps S3 to S5 are repeated.
[0016] If the coating fails to meet the pre-set criteria after a series of consecutive tests, and the confidence level of each test reaches the first pre-set threshold, the current batch of pots is prohibited from entering the loading process.
[0017] Furthermore, the deviation vector in step S4 is obtained by subtracting the current batch feature vector from the corresponding stage baseline feature vector, and then a standardized deviation vector is obtained by standardization processing. The standardization processing method includes any one of Z-score standardization, covariance whitening processing, or Mahalanobis distance standardization.
[0018] Furthermore, the method for establishing the benchmark feature database includes:
[0019] Acquire the timing signals of the stirring motor torque of multiple historical batches under different material working conditions, and screen qualified batches based on manual inspection results, laboratory coating rate test results, or production quality test results;
[0020] Based on the stirring process signal, the torque timing signal of each historical batch is divided into the dry mixing stage, the oil injection stage, and the wet mixing stage.
[0021] The feature vectors of the torque signals at each stage are extracted and classified and statistically analyzed according to the material working condition information and material quality.
[0022] The statistical mean of the characteristic vectors of the corresponding stages under various working conditions is calculated as the benchmark characteristic vector for that working condition and stored in the benchmark characteristic database.
[0023] Among them, the same material working conditions include at least working conditions where the aggregate gradation, asphalt-aggregate ratio, mineral powder content, asphalt grade and material quality are consistent or meet the preset similarity requirements.
[0024] The beneficial effects of this invention are as follows: By collecting the torque timing signal of the mixing motor, the invention can identify and determine the coating state during the mixing process, without waiting for laboratory extraction test results. For the characteristics of mechanical mixing of aggregates and mineral powder in the dry mixing stage, features reflecting load level and impact intensity, such as mean, effective value, peak factor, and approximate entropy, are extracted. For the dynamic process of asphalt atomization and adhesion in the spraying stage, features reflecting stick-slip oscillation characteristics, such as transient response time, torque jump value, and modulation depth, are extracted. For the stable process of asphalt film homogenization in the wet mixing stage, features reflecting coating smoothness and uniformity, such as waveform factor, low-frequency energy ratio, torque variance, kurtosis, and dominant frequency energy concentration, are extracted. The physical meaning of the features in each stage is clear, and their correlation with the coating state is strong, avoiding the information loss problem caused by processing the three stages uniformly. By utilizing the causal constraints in the transition probability matrix, the inherent physical causal chain of the mixing process is utilized to suppress abnormal state jumps that violate physical laws, reduce the risk of misjudgment caused by single-segment signal anomalies, and improve the robustness of the judgment results. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is a process flow diagram of the present invention.
[0027] Figure 2 This is a flowchart illustrating the specific steps of step S5 in this invention. Detailed Implementation
[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the invention, not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.
[0029] like Figure 1 and Figure 2 As shown, the asphalt mixture coating state identification method based on the torque signal of the mixing motor of this application includes the following steps:
[0030] S1. Obtain the material condition information and material quality of the current batch. The material condition information includes one or more of the following: aggregate particle size distribution, asphalt-aggregate ratio, mineral powder content, asphalt grade, and asphalt temperature.
[0031] S2. Based on the material operating condition information and material quality, obtain the torque benchmark feature model corresponding to the current operating condition from the benchmark feature database. The torque benchmark feature model includes benchmark feature vectors corresponding to the dry mixing stage, the oil injection stage, and the wet mixing stage. The benchmark feature vectors are established by statistically analyzing the torque signals of historical qualified batches of coating, and are used to characterize the ideal torque characteristics of each mixing stage under the corresponding operating condition. The dimensions of the benchmark feature vector for the dry mixing stage include the mean torque, the effective torque value, the peak factor, and the approximate entropy.
[0032] Where approximate entropy :
[0033]
[0034] The calculation steps are as follows:
[0035] The torque sequence of the dry mixing section with a length of N. Reconstructed into an m-dimensional vector:
[0036]
[0037] Define vector and The distance is the maximum value of the difference between its corresponding elements:
[0038] Statistical satisfaction The number of j (including k=j) is denoted as And calculate:
[0039]
[0040] The approximate entropy is: .
[0041] m is the embedding dimension, which is set to 2 in this implementation; r is the similarity tolerance; and N is the number of sampling points for the torque signal in the dry mixing section. The approximate entropy reflects the complexity of the torque signal in the dry mixing section; a larger value indicates a more irregular aggregate collision pattern and less uniform mixing. The mean torque reflects the average load level of the aggregate and mineral powder mixture in the dry mixing section, while the effective torque value reflects the total energy of load fluctuations in the dry mixing section.
[0042] The reference characteristic vector of the injection section includes transient response time, torque surge value, and modulation depth. Transient response time reflects the speed at which asphalt adheres to the aggregate after atomization; a longer response time indicates higher asphalt viscosity or insufficient atomization pressure. Torque surge value reflects the magnitude of the additional load caused by asphalt adhesion and is positively correlated with the asphalt-aggregate ratio and asphalt viscosity. It can be calculated by the difference between the steady-state torque value at the end of the injection section and the torque value at the beginning of the injection.
[0043] The modulation depth reflects the oscillation intensity of the asphalt adhesion-slip (stick-slip) effect, with a value range of [0,1]. The larger the value, the more intense the stick-slip oscillation, which is usually related to the lower temperature and higher viscosity of the asphalt.
[0044]
[0045] max(env) is the maximum value of the envelope of the torque signal in the fuel injection section; min(env) is the minimum value of the envelope of the torque signal in the fuel injection section; env(•) is the Hilbert envelope operator, obtained by performing a Hilbert transform on the torque signal and taking the modulus; env is the Hilbert envelope of the torque signal in the fuel injection section.
[0046] The dimensions of the baseline characteristic vector for wet-mix sections include waveform factor, low-frequency energy proportion, torque variance, kurtosis, and dominant frequency energy concentration. The waveform factor reflects the shape of the torque waveform. When the coating is good, the asphalt film is smooth, and the waveform factor is close to 1; when the coating is poor, there are spikes caused by uncoated particles, and the waveform factor increases.
[0047] The low-frequency energy ratio refers to the ratio of the torque signal energy in the 0.5~20Hz frequency band to the total energy across the entire frequency band. It reflects the proportion of low-frequency load fluctuation energy to the total energy. When the coating is good, stick-slip oscillations are reduced, and the low-frequency energy ratio decreases; when the coating is poor, the macroscopic movement of the aggregate is hindered, and the low-frequency energy ratio increases.
[0048] Torque variance reflects the intensity of torque signal fluctuation in the wet mixing section; the larger the variance, the more severe the load fluctuation.
[0049] Kurtosis reflects the degree of spikes or impact components in the torque signal. A normal distribution has a kurtosis of ≈3; however, if impact spikes caused by uncoated aggregates exist, the kurtosis is significantly greater than 3.
[0050] Methods for establishing a benchmark feature database include:
[0051] Acquire the timing signals of the stirring motor torque of multiple historical batches under different material working conditions, and screen qualified batches based on manual inspection results, laboratory coating rate test results, or production quality test results;
[0052] Based on the stirring process signal, the torque timing signal of each historical batch is divided into the dry mixing stage, the oil injection stage, and the wet mixing stage.
[0053] The feature vectors of the torque signals at each stage are extracted and classified and statistically analyzed according to the material working condition information and material quality.
[0054] The statistical mean of the characteristic vectors of the corresponding stages under various working conditions is calculated as the benchmark characteristic vector for that working condition and stored in the benchmark characteristic database.
[0055] Among them, the same material working conditions include at least working conditions where the aggregate gradation, asphalt-aggregate ratio, mineral powder content, asphalt grade and material quality are consistent or meet the preset similarity requirements.
[0056] S3. Collect the torque timing signal of the stirring motor of the current batch, and extract the time domain and frequency domain features of the torque signals of the dry mixing stage, the oil injection stage and the wet mixing stage respectively to obtain the feature vector of the torque acquisition signal of each mixing stage of the current batch.
[0057] S4. Calculate the feature deviation between the feature vector collected at each stage and the corresponding reference feature vector to obtain the deviation vectors for the dry mixing stage, the oil injection stage, and the wet mixing stage. Then, standardize each deviation vector. The deviation vector is obtained by subtracting the feature vector collected in the current batch from the reference feature vector of the corresponding stage. Standardization is then applied to obtain the standardized deviation vector. The standardization method includes any one of Z-score standardization, covariance whitening, or Mahalanobis distance standardization.
[0058] S5. Establish a stage state model corresponding to the mixing process. Use the standardized dry mixing stage deviation vector, oil spraying stage deviation vector and wet mixing stage deviation vector as observations to jointly infer the state of the dry mixing stage, oil spraying stage and wet mixing stage to obtain the state of each stage of the current batch and the final coating state.
[0059] The stage state model is established using a Hidden Markov Model, with the dry mixing stage, oil injection stage, and wet mixing stage serving as three sequential state nodes. The model transition probabilities satisfy the sequential constraints of the mixing process, making the state of the previous stage a priori condition for inferring the state of the next stage, thus obtaining the final state of the wet mixing stage as the coating state of the current batch.
[0060] like Figure 2 As shown, the specific steps include the following:
[0061] A three-stage hidden Markov model corresponding to the mixing process is established. The dry mixing stage, oil injection stage, and wet mixing stage are defined as the first stage, second stage, and third stage hidden states, respectively. The state sequence is represented as follows:
[0062]
[0063] in, This represents the current pot's stage state sequence. These represent the process states of the dry mixing stage, the oil spraying stage, and the wet mixing stage, respectively. The value of each stage state is taken from the set. , where 0 indicates that the process requirements are met in this stage, and 1 indicates that the process requirements are not met in this stage;
[0064] The standardized bias vector obtained in step S4 is used as the observation sequence of the hidden Markov model:
[0065]
[0066] Where O represents the observation sequence, , , These are the standardized deviation vectors for the dry mixing stage, the oil injection stage, and the wet mixing stage, respectively.
[0067] Establish the parameters of the Hidden Markov Model, including the initial state probability, state transition probability, and observation emission probability:
[0068] The initial state probability of the dry mixing stage, as the first stage, is expressed as follows: ,in, This represents the initial probability of being in state i during the dry mixing stage; i represents the state number. ∈ ; This represents the initial probability that the process requirements are met during the dry mixing stage; This represents the initial probability that the dry mixing stage does not meet the process requirements, and it satisfies... The initial state probability is obtained based on historical production data statistics;
[0069] Establish the transition probabilities between phase states:
[0070] The state transition probability from the dry mixing stage to the fuel injection stage is expressed as: The state transition probability from the fuel injection stage to the wet mixing stage is expressed as: ,in, This represents the probability of transitioning to state j in the fuel injection stage when the dry mixing stage is state i. This represents the probability of transitioning to state j in the wet mixing stage, given that state i is the fuel injection stage; where i and j are both state numbers, and The state transition probabilities for each stage satisfy the following:
[0071] , The state transition probability is obtained based on historical batch statistics and satisfies the stirring process sequence constraint.
[0072] Establish the observed launch probability:
[0073] The observation vectors at each stage follow a multivariate Gaussian distribution, and their emission probabilities are expressed as:
[0074]
[0075] Where k represents the mixing stage number. This indicates the dry mixing stage. Indicates the fuel injection stage. Indicates the wet mixing stage; This represents the standardized deviation vector actually calculated in stage k of the current batch. This represents the probability density of the current deviation vector occurring when the stage state is i; This represents the mean of the observation vector corresponding to stage k in state i; This represents the covariance matrix of the observation vector corresponding to stage k in state i; The mean vector represents the multivariate Gaussian probability density function. Covariance Matrix The standardization deviation vector of the corresponding stage of the historically calibrated coating state pots was obtained by statistical analysis.
[0076] Establish the current three-stage joint probability of the satellite launch based on the initial state probability, state transition probability, and observed launch probability:
[0077]
[0078] in, This represents the joint probability of the current boiler stage state sequence and the observation sequence occurring simultaneously;
[0079] The Viterbi algorithm is used to calculate the state path with the highest joint probability:
[0080]
[0081] in, This represents the sequence of stage states with the highest joint probability.
[0082] The state of the wet mixing stage in the optimal state path is taken as the final coating state of the current batch;
[0083] The forward-backward algorithm is used to calculate the posterior probability of the final state in the wet mixing stage:
[0084]
[0085] in, This indicates the forward probability of being in state i during the wet mixing stage, calculated by the forward algorithm. This indicates that the backward algorithm calculates the state of the wet mixing stage. The backward probability; the posterior probability corresponding to the final decision state is used as the decision confidence: Where c represents the confidence level of the current pot coating state determination, and The larger the value of c, the higher the reliability of the current judgment result.
[0086] S7. Based on the final coating state output by the stage state model and the corresponding judgment confidence level, determine the coating qualification of the current batch.
[0087] When it is determined that the coating of the current batch is unqualified and the confidence level of the determination reaches the first preset threshold, the stirring process parameters of the current batch are adjusted. The stirring process parameters include at least one or more of the following: wet mixing time, stirring mode, and stirring speed.
[0088] After compensation is completed, the torque timing signal is reacquired, and steps S3 to S6 are repeated.
[0089] If the coating fails to meet the pre-set criteria after a series of consecutive tests, and the confidence level of each test reaches the first pre-set threshold, the current batch of pots is prohibited from entering the loading process.
[0090] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor, characterized in that, Includes the following steps: S1. Obtain the material operating conditions and material quality for the current batch. S2. Based on the material working condition information and material quality, obtain the torque benchmark feature model corresponding to the current working condition from the benchmark feature database. The torque benchmark feature model includes benchmark feature vectors corresponding to the dry mixing stage, oil injection stage and wet mixing stage. The benchmark feature vectors are established by statistically analyzing the torque signals of historical qualified batches of coating, and are used to characterize the ideal torque characteristics of each mixing stage under the corresponding working condition. The dimensions of the dry mixing section's reference feature vector include the mean torque, effective torque value, peak factor, and approximate entropy; the dimensions of the fuel injection section's reference feature vector include transient response time, torque jump value, and modulation depth; and the dimensions of the wet mixing section's reference feature vector include waveform factor, low-frequency energy ratio, torque variance, kurtosis, and main frequency energy concentration. The low-frequency energy ratio refers to the ratio of the torque signal's energy in the 0.5~20Hz frequency band to the energy across the entire frequency band. S3. Collect the torque timing signal of the stirring motor of the current batch, and extract the time domain and frequency domain features of the torque signals of the dry mixing stage, the oil injection stage and the wet mixing stage respectively to obtain the feature vector of the torque acquisition signal of each mixing stage of the current batch. S4. Calculate the feature deviation between the feature vector collected at each stage and the corresponding reference feature vector to obtain the deviation vector of the dry mixing stage, the deviation vector of the oil injection stage, and the deviation vector of the wet mixing stage, and standardize each deviation vector. S5. Establish a stage state model corresponding to the mixing process. Using the standardized dry mixing stage deviation vector, oil injection stage deviation vector, and wet mixing stage deviation vector as observations, jointly infer the state of the dry mixing stage, oil injection stage, and wet mixing stage to obtain the state of each stage of the current batch and the final coating state. The stage state model satisfies the sequential constraints between the mixing process stages, so that the state of the later stage is constrained by the state of the previous stage. S6. Based on the final coating state output by the stage state model and the corresponding judgment confidence level, determine the coating qualification of the current batch. When it is determined that the coating of the current batch is unqualified and the confidence level of the determination reaches the first preset threshold, the stirring process parameters of the current batch are adjusted. The stirring process parameters include at least one or more of the following: wet mixing time, stirring mode, and stirring speed. After compensation is completed, the torque timing signal is reacquired, and steps S3 to S5 are repeated. If the coating fails to meet the pre-set criteria after a series of consecutive tests, and the confidence level of each test reaches the first pre-set threshold, the current batch of pots is prohibited from entering the loading process.
2. The method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor according to claim 1, characterized in that: The stage state model is established using a hidden Markov model. The dry mixing stage, the oil injection stage, and the wet mixing stage are respectively regarded as three sequential state nodes, and the standardized deviation vector of each stage is used as the corresponding observation. Among them, the model transition probability satisfies the sequential constraint of the stirring process, so that the state of the previous stage is used as the prior condition for the inference of the state of the next stage, so as to obtain the final state of the wet mixing stage as the current pot coating state.
3. The method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor according to claim 2, characterized in that: Step S5 specifically includes the following steps: A three-stage hidden Markov model corresponding to the mixing process is established. The dry mixing stage, oil injection stage, and wet mixing stage are defined as the first stage, second stage, and third stage hidden states, respectively. The state sequence is represented as follows: in, This represents the current pot's stage state sequence. These represent the process states of the dry mixing stage, the oil spraying stage, and the wet mixing stage, respectively. The value of each stage state is taken from the set. , where 0 indicates that the process requirements are met in this stage, and 1 indicates that the process requirements are not met in this stage; The standardized bias vector obtained in step S4 is used as the observation sequence of the hidden Markov model: Where O represents the observation sequence, , , These are the standardized deviation vectors for the dry mixing stage, the oil injection stage, and the wet mixing stage, respectively. Establish the parameters of the Hidden Markov Model, including the initial state probability, state transition probability, and observation emission probability: The initial state probability of the dry mixing stage, as the first stage, is expressed as follows: ,in, This represents the initial probability of being in state i during the dry mixing stage; i represents the state number. ∈ ; This represents the initial probability that the process requirements are met during the dry mixing stage; This represents the initial probability that the dry mixing stage does not meet the process requirements, and it satisfies... The initial state probability is obtained based on historical production data statistics; Establish the transition probabilities between phase states: The state transition probability from the dry mixing stage to the fuel injection stage is expressed as: The state transition probability from the fuel injection stage to the wet mixing stage is expressed as: ,in, This represents the probability of transitioning to state j in the fuel injection stage when the dry mixing stage is state i. This represents the probability of transitioning to state j in the wet mixing stage, given that state i is the fuel injection stage; where i and j are both state numbers, and The state transition probabilities for each stage satisfy the following: , The state transition probability is obtained based on historical batch statistics and satisfies the stirring process sequence constraint. Establish the observed launch probability: The observation vectors at each stage follow a multivariate Gaussian distribution, and their emission probabilities are expressed as: Where k represents the mixing stage number. This indicates the dry mixing stage. Indicates the fuel injection stage. Indicates the wet mixing stage; This represents the standardized deviation vector actually calculated in stage k of the current batch. This represents the probability density of the current deviation vector occurring when the stage state is i; This represents the mean of the observation vector corresponding to stage k in state i; This represents the covariance matrix of the observation vector corresponding to stage k in state i; The mean vector represents the multivariate Gaussian probability density function. Covariance Matrix The standardization deviation vector of the corresponding stage of the historically calibrated coating state pots was obtained by statistical analysis. Establish the current three-stage joint probability of the satellite launch based on the initial state probability, state transition probability, and observed launch probability: in, This represents the joint probability of the current boiler stage state sequence and the observation sequence occurring simultaneously; The Viterbi algorithm is used to calculate the state path with the highest joint probability: in, This represents the sequence of stage states with the highest joint probability. The state of the wet mixing stage in the optimal state path is taken as the final coating state of the current batch; The forward-backward algorithm is used to calculate the posterior probability of the final state in the wet mixing stage: in, This indicates the forward probability of being in state i during the wet mixing stage, calculated by the forward algorithm. This indicates that the backward algorithm calculates the state of the wet mixing stage. The backward probability; the posterior probability corresponding to the final decision state is used as the decision confidence: Where c represents the confidence level of the current pot coating state determination, and The larger the value of c, the higher the reliability of the current judgment result.
4. The method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor according to claim 1, characterized in that: The deviation vector in step S4 is obtained by subtracting the feature vector collected in the current batch from the benchmark feature vector of the corresponding stage, and then a standardized deviation vector is obtained by standardization processing. The standardization processing method includes any one of Z-score standardization, covariance whitening processing or Mahalanobis distance standardization.
5. The method for identifying the coating state of asphalt mixture based on the torque signal of a mixing motor according to claim 1, characterized in that: The method for establishing the benchmark feature database includes: Acquire the timing signals of the stirring motor torque of multiple historical batches under different material working conditions, and screen qualified batches based on manual inspection results, laboratory coating rate test results, or production quality test results; Based on the stirring process signal, the torque timing signal of each historical batch is divided into the dry mixing stage, the oil injection stage, and the wet mixing stage. The feature vectors of the torque signals at each stage are extracted and classified and statistically analyzed according to the material working condition information and material quality. The statistical mean of the characteristic vectors of the corresponding stages under various working conditions is calculated as the benchmark characteristic vector for that working condition and stored in the benchmark characteristic database. Among them, the same material working conditions include at least working conditions where the aggregate gradation, asphalt-aggregate ratio, mineral powder content, asphalt grade and material quality are consistent or meet the preset similarity requirements.