Asphalt mixing plant monitoring system and method

Abstract

The invention relates to the technical field of asphalt concrete production, in particular to an asphalt mixing plant monitoring system and method. The method comprises the following steps: acquiring real-time production operation data of the asphalt mixing plant, and constructing a real-time production state vector; acquiring production process data of all superior product batches in history as a training data set; performing clustering processing on the training data set to obtain a qualified mode set; when the real-time production state vector does not belong to any qualified production mode cluster, calculating a mode deviation degree, and further calculating a comprehensive risk level; performing grading early warning based on a size relationship between the comprehensive risk grade and a preset risk threshold value, and monitoring the asphalt mixing plant; and the monitoring accuracy of the asphalt stirring process is improved.

Description

Technical Field

[0001] This application relates to the field of asphalt concrete production technology, and in particular to a monitoring system and method for asphalt mixing plants. Background Technology

[0002] Asphalt mixing plants are the core equipment for producing asphalt mixtures in infrastructure projects such as highway construction. The quality of the finished mixture, especially the stability and uniformity of the asphalt-aggregate ratio and aggregate gradation, directly determines the final performance and durability of the pavement. The production process requires precise metering, heating, and uniform mixing of various specifications of aggregates, mineral powder, recycled materials, and high-temperature liquid asphalt.

[0003] Currently, the quality control of asphalt mixing plants mainly relies on programmable logic controllers (PLCs). The existing system uses a formula setting and static threshold alarm system. This system controls the amount of each material added by pre-setting the production formula, triggering an alarm when sensor readings exceed upper and lower limits. However, this control method has significant limitations. Because it is a passive response, problems are usually only detected when deviations are already large, at which point several tons of substandard mixture may have already been produced, resulting in economic losses and resource waste. Furthermore, this response method cannot identify complex abnormal operating conditions caused by the combined effect of small fluctuations in multiple parameters within permissible ranges. Existing monitoring methods lack the ability to recognize and assess the overall state pattern of the production process, and cannot provide early warnings of quality risks caused by the coupling of multiple factors.

[0004] Therefore, there is an urgent need for a monitoring method that can autonomously learn normal production patterns from massive amounts of production data and identify atypical high-risk production conditions in real time, so as to achieve forward-looking and refined assurance of the uniformity of asphalt mixture quality. Summary of the Invention

[0005] To address the technical problem of being unable to provide early warnings of quality risks caused by the coupling of multiple factors due to a lack of understanding and assessment of the overall state pattern of the production process, this application provides an asphalt mixing plant monitoring system and method.

[0006] Firstly, this application provides a monitoring method for asphalt mixing plants, employing the following technical solution: Real-time production and operation data of asphalt mixing plants are obtained to construct a real-time production status vector; production process data of all batches of superior products in history are obtained as a training dataset; clustering is performed on the training dataset to obtain a set of qualified patterns, wherein the set of qualified patterns contains the centroid of each qualified production pattern cluster. When the real-time production state vector does not belong to any qualified production mode cluster, the following operations are performed: Calculate the average of the standard deviations of all production operation data within the collection period to obtain the process volatility; calculate the mode deviation, which is positively correlated with the distance from the current real-time production state vector to the centroid of the nearest qualified production mode and with the process volatility; calculate the comprehensive risk level, which is positively correlated with the mode deviation and with the dot product of the absolute values ​​of the difference vectors between the preset weight vector and the current real-time production state vector and the centroid of the nearest qualified production mode; and perform graded early warning based on the relationship between the comprehensive risk level and the preset risk threshold to monitor the asphalt mixing plant.

[0007] The beneficial effects are as follows: By comparing real-time production operation data with historical high-quality batch data, and using the qualified pattern set formed by clustering, it is possible to determine whether the current production status is in a healthy mode. In abnormal situations, the pattern deviation degree and comprehensive risk level are calculated. Combined with a graded early warning mechanism, potential production risks can be identified in advance. This approach breaks through the traditional static alarm mode based on single-parameter thresholds, realizing a holistic understanding and forward-looking monitoring of complex working conditions, thereby effectively improving the monitoring accuracy and reliability of the asphalt mixing process.

[0008] Furthermore, the step of clustering the training dataset to obtain a qualified pattern set includes: processing the training dataset using the DBSCAN clustering algorithm, taking all non-noise point clusters as qualified production pattern clusters, and taking the set of all qualified production pattern clusters as the qualified production pattern set.

[0009] The beneficial effects are as follows: By using the DBSCAN clustering algorithm to process historical superior batch data, a set of qualified production patterns can be automatically formed without relying on a preset number of clusters. Furthermore, noise points can be removed, ensuring that the qualified patterns are more representative. This improves the model's ability to cover real production conditions, enabling the system to flexibly adapt to process fluctuations under different batches and operating conditions.

[0010] Furthermore, the formula for calculating the mode deviation is: In the formula, The deviation of the model at each time point, This represents the real-time production status vector at each point in time. For the set of qualified patterns, For the set of qualified patterns The Middle The centroid of a qualified production mode cluster Real-time production state vector and qualified mode set The Middle The centroid of a qualified production model cluster The distance between them Minimum value function The process fluctuation of the data collection cycle at each time point is the last time point.

[0011] The beneficial effects are: the calculation of the mode deviation introduces the minimum distance between the real-time state vector and the centroid of the qualified mode cluster, and is corrected by combining the process volatility, so that the risk assessment reflects the degree of geometric deviation while taking into account the stability of the production process.

[0012] Furthermore, the method for obtaining the comprehensive risk level is as follows: for each time point, calculate the vector composed of the absolute values ​​of the differences between the elements of the real-time production state vector and the centroid of the qualified production mode cluster closest to the real-time production state vector, and calculate the product of the dot product of the vector and the preset sensitivity weight vector and the mode deviation degree as the comprehensive risk level for each time point.

[0013] The beneficial effects are as follows: The calculation method for the comprehensive risk level combines the difference vector after weighting the model deviation and parameter sensitivity, highlighting the key parameters that have a greater impact on the quality of finished asphalt products. When core parameters are abnormal, the risk level increases significantly, thus making the risk assessment more consistent with the actual importance distribution in production and avoiding false alarms caused by fluctuations in non-critical parameters.

[0014] Furthermore, the warning includes: issuing a warning for parameters whose weight values ​​in the preset sensitivity weight vector are within the preset weight percentage, and whose absolute values ​​in the vector formed by the absolute values ​​of the differences between the elements of the real-time production state vector and the centroid of the nearest qualified production mode cluster are within the preset weight percentage.

[0015] The beneficial effects are: by focusing on parameters with high sensitivity weights in the early warning mechanism and combining real-time deviation to filter out the most critical abnormal parameters, the source of the problem can be quickly located, thus improving the efficiency of fault diagnosis.

[0016] Furthermore, the real-time production operation data includes at least: the frequency of the feeding belt conveyor of each cold material silo; the flame temperature of the dryer burner; the aggregate temperature at the outlet of the dryer; the instantaneous weight of the asphalt weighing hopper; the asphalt supply temperature; and the motor current of the mixing cylinder.

[0017] Furthermore, the preset sensitivity weight vector is determined based on the degree of influence of production parameters on the quality of finished asphalt products.

[0018] Furthermore, the hierarchical early warning based on the relationship between the comprehensive risk level and the preset risk threshold includes: normalizing the comprehensive risk level at all time points, and issuing low-risk, medium-risk, and high-risk alarms respectively when the normalized comprehensive risk level exceeds the preset low-risk, medium-risk, and high-risk thresholds.

[0019] Furthermore, the values ​​of each element in the preset sensitivity weight vector correspond to the order of the types of real-time production and operation data.

[0020] Secondly, this application provides a monitoring system for an asphalt mixing plant, which adopts the following technical solution: An asphalt mixing plant monitoring system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an asphalt mixing plant monitoring method as described above is implemented.

[0021] The above-mentioned method for monitoring asphalt mixing plants is used to generate a computer program, which is then stored in a memory for loading and execution by a processor. This allows for the creation of a system based on the memory and processor, making it convenient to use.

[0022] This application has the following technical advantages: This application achieves more refined and forward-looking monitoring of the asphalt mixing plant production process by introducing a DBSCAN-based clustering method and an improved risk assessment mechanism. This application departs from passive alarms based on single sensor thresholds. It utilizes historical high-quality batch data to form a set of qualified production patterns, enabling proactive identification of whether production is in a healthy state, thus avoiding the waste of large quantities of substandard mixtures under traditional methods. Furthermore, it introduces a process volatility index to capture the stability of operating conditions before anomalies occur, allowing risk assessment to identify potential systemic instability and sudden failures while focusing on the degree of deviation, improving the ability to judge complex operating conditions. Based on this, a pattern deviation degree is constructed to quantify geometric deviation and process instability risks, ensuring keen detection of quality risks. Furthermore, a preset sensitivity weight vector is constructed to focus comprehensive risk level assessment on core quality indicators, thereby avoiding false warnings caused by fluctuations in non-critical parameters and improving the accuracy and guidance of warnings. Finally, through a tiered warning mechanism, anomalies are promptly alerted while clearly indicating the key parameter dimensions where risks are mainly concentrated, helping operators quickly locate problems and take targeted adjustment measures. This application improves the ability to monitor complex asphalt mixing conditions, provides early warnings for abnormal situations in the asphalt mixing process, improves monitoring accuracy, and reduces economic losses and resource waste. Attached Figure Description

[0023] Figure 1 This is a flowchart of a monitoring method for an asphalt mixing plant according to this application. Detailed Implementation

[0024] This application discloses a method for monitoring asphalt mixing plants. The method acquires real-time production and operation data of the asphalt mixing plant and constructs a real-time production state vector. It obtains historical production process data of all superior batches as a training dataset. The training dataset is clustered to obtain a set of qualified patterns. When the real-time production state vector does not belong to any qualified production pattern cluster, the pattern deviation is calculated, and then the comprehensive risk level is calculated. Based on the relationship between the comprehensive risk level and a preset risk threshold, a graded early warning is issued to monitor the asphalt mixing plant. This improves the accuracy of asphalt mixing process monitoring.

[0025] Reference Figure 1 A monitoring method for asphalt mixing plants includes steps S1-S3.

[0026] Step S1: Obtain real-time production and operation data of the asphalt mixing plant and construct a real-time production status vector; obtain production process data of all historical high-quality batches as a training dataset.

[0027] The system acquires real-time production and operation data of the asphalt mixing plant collected by the monitoring system within a preset collection period. In one embodiment of this application, the preset period is 5 seconds, but the implementer can choose other values ​​based on actual conditions. The real-time production and operation data of the asphalt mixing plant selected in this application includes: the frequency of the feeding belt conveyor of each cold aggregate bin, the flame temperature of the dryer burner, the fuel consumption rate, the aggregate temperature at the outlet of the dryer, the instantaneous weight of the asphalt weighing hopper, the asphalt supply temperature, and the motor current of the mixing cylinder.

[0028] For each data acquisition cycle, after normalizing all types of real-time production operation data at each time point within the acquisition cycle, the real-time production status vector at each time point within each acquisition cycle is obtained. The normalization process is a well-known technique and will not be elaborated upon in this application.

[0029] At the same time, production process data of all batches of asphalt mixing plants that have been tested as superior products in the laboratory are collected as superior historical data, and the set of real-time production status vectors at each time point in the superior historical data is used as training dataset.

[0030] Step S2: Cluster the training dataset to obtain a set of qualified patterns, which contains the centroids of each qualified production pattern cluster. When the real-time production state vector does not belong to any qualified production pattern cluster, perform the following operations: calculate the average of the standard deviations of all production operation data within the collection period to obtain the process volatility, and calculate the pattern deviation. The pattern deviation is positively correlated with the distance from the current real-time production state vector to the nearest qualified production pattern centroid and is positively correlated with the process volatility.

[0031] Furthermore, this application uses Clustering algorithms are used to cluster the training dataset, dividing it into clusters and removing noise. Each cluster, excluding noise, is designated as a qualified production pattern cluster. Each qualified production pattern cluster represents a validated production pattern capable of stably producing high-quality asphalt mixtures. The set of all qualified production pattern clusters is denoted as the qualified pattern set. And calculate the centroid of each qualified production mode cluster. The method for obtaining the centroid is a well-known technique and will not be elaborated upon in this application.

[0032] For the real-time production status vector at each time point within each collection cycle If the real-time production state vector If it does not belong to the qualified production mode cluster, then it is... If a clustering algorithm identifies a point as a noise point, it indicates that the production status at that point in time is a state that has not been verified by historical data and poses a quality risk. Therefore, this application analyzes the real-time production status vector that is identified as a noise point.

[0033] To further assess the potential impact of this anomalous state, it is necessary to analyze the process stability prior to its occurrence. Based on this, this application constructs a process volatility index. It should be noted that the process volatility index This measure assesses the dispersion of key metrics within a short time window preceding an anomaly. An anomaly occurring suddenly during a stable process carries a different risk than one occurring during sustained fluctuations. The former may represent a sudden failure, while the latter may represent systemic instability. Specifically, process volatility indicators... The calculation formula is: ; In the formula, This represents the process fluctuation within the data collection period, with each time point serving as the final time point. This represents the number of elements in the real-time production state vector. The length of the data collection period. The first in the real-time production state vector at each time point The value of a real-time production operation data point. This represents the first-order vector in the real-time production status vector of all time points within the acquisition period, with each time point being the last time point. The average of the values ​​of real-time production operation data. This represents the current time at each point in time. It should be noted that, to prevent the denominator from being zero, a time length greater than 1 should be selected when choosing the duration of the data collection period.

[0034] It should be noted that process variability Recent process instability is quantified by calculating the average of the standard deviations of multiple key parameters within a specified time window. (Process volatility) The larger the value, the more drastic the fluctuations in the metering process during production before the abnormal state occurs, and the worse the system stability.

[0035] Furthermore, obtain the process volatility. Next, further analysis is needed based on the scenario content. It should be noted that, in addition to the instability of the asphalt mixing process, it is also necessary to analyze the degree to which the current abnormal state deviates from the qualified production mode. Based on this, the deviation degree of the construction mode in this application is determined. Pattern deviation This analysis combines the geometric distance between the current outlier and the nearest successful production model, as well as process volatility. It's important to note that an outlier that is far from a successful model and exhibits significant process volatility carries a significantly higher risk than an outlier that is close by and has a stable process. This is used to construct the model deviation. The aim is to quantify this complex risk; specifically, pattern deviation. The calculation formula is: ; In the formula, The deviation of the model at each time point, This represents the real-time production status vector at each point in time. For the set of qualified patterns, For the set of qualified patterns The Middle The centroid of a qualified production mode cluster Real-time production state vector and qualified mode set The Middle The centroid of a qualified production model cluster The distance between them Minimum value function The process fluctuation of the data collection cycle at each time point is the last time point; This means that for each point in time, the minimum distance between the real-time production state vector and the centroid of all qualified production mode clusters in the qualified mode set is obtained. In addition, in one embodiment of this application, Euclidean distance is selected as the distance measurement method. The implementer may select other distance measurement methods based on the actual situation.

[0036] It should be noted that the pattern deviation... It is proportional to the distance from the abnormal real-time production state vector to the nearest qualified mode, and also to the process variability. Proportional. When the process volatility When the process increases, i.e. when the process itself is unstable, the same geometric deviation distance will be amplified, thus leading to a greater mode deviation. The higher value reflects a greater increase in the risk of deviant behavior in an unstable context.

[0037] Step S3: Calculate the comprehensive risk level. The comprehensive risk level is positively correlated with the mode deviation and positively correlated with the dot product of the absolute values ​​of the difference vectors between the preset weight vector, the current real-time production state vector, and the centroid of the most recent qualified production mode. Based on the relationship between the comprehensive risk level and the preset risk threshold, a graded early warning is performed to monitor the asphalt mixing plant.

[0038] Get pattern deviation Furthermore, the varying weights of different production parameters on the final quality must be considered. For instance, deviations in the asphalt-aggregate ratio and aggregate weight are more critical than deviations in dust removal system parameters. Based on this, this application constructs a pre-defined sensitivity weight vector and establishes a comprehensive risk level. Furthermore, by combining the comprehensive risk level with the DBSCAN clustering algorithm, assessment defects caused by noise points are identified, thus improving the original algorithm. It should be noted that risk assessment needs to focus on the dimensions that have the greatest impact on the quality of the finished product, achieving a precise profile of the risk.

[0039] Based on the above analysis, the overall risk level is determined. The calculation formula is as follows: In the formula, The overall risk level at each point in time. The deviation of the model at each time point, For one A predefined sensitivity weight vector of dimension, where each element of the vector... Representing the The degree of influence of each production parameter on the quality of the finished product The larger the value, the more critical the impact. This vector can be set by process engineers based on experience and historical data analysis. The distance to the real-time production state vector The centroid of the most recent qualified production pattern cluster, This represents the real-time production status vector at each point in time. express and The element-wise absolute value of the difference vector. This represents the vector dot product operation. In one embodiment of this application, the preset sensitivity weight vector is (0.05, 0.05, 0.1, 0.1, 0.2, 0.2, 0.3), and the implementer can select other values ​​based on the actual situation.

[0040] It should be noted that the final risk level In mode deviation Based on this, adjustments are made using a weighted term. It calculates the deviation of the current state vector in each dimension and sums them using their sensitivity to quality. If a deviation mainly occurs in parameters with high sensitivity weights, the value of that dot product term will increase, thus significantly increasing the final risk level. Conversely, if the deviation mainly occurs in non-critical parameters, then The increase will decrease.

[0041] Based on the comprehensive risk level... The value is compared with a preset risk threshold to achieve tiered early warning. The comprehensive risk level at all time points is normalized; this application sets three thresholds: low risk, medium risk, and high risk. When the normalized comprehensive risk level... When the risk exceeds preset low-risk, medium-risk, and high-risk thresholds respectively, the system issues low-risk, medium-risk, and high-risk alarms to the operator respectively, based on the preset sensitivity weight vector. The weighting is in the top 10%, and in The parameters with deviations within the top 10% are given an early warning to guide operators to conduct targeted checks and adjustments. In one embodiment of this application, the values ​​of the low-risk threshold, medium-risk threshold, and high-risk threshold are 0.3, 0.6, and 0.8, respectively. Implementers can select other values ​​based on actual circumstances.

[0042] This application also discloses an asphalt mixing plant monitoring system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an asphalt mixing plant monitoring method according to this application is implemented.

[0043] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0044] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A monitoring method for an asphalt mixing plant, characterized in that, The steps include: acquiring real-time production and operation data of the asphalt mixing plant and constructing a real-time production status vector; acquiring production process data of all historical high-quality batches as a training dataset; Clustering is performed on the training dataset to obtain a set of qualified patterns, which contains the centroids of each qualified production pattern cluster. When the real-time production state vector does not belong to any qualified production mode cluster, the following operations are performed: calculate the average of the standard deviations of all production operation data within the collection period to obtain the process volatility, and calculate the mode deviation. The mode deviation is positively correlated with the distance from the current real-time production state vector to the centroid of the nearest qualified production mode and is positively correlated with the process volatility. Calculate the comprehensive risk level, which is positively correlated with the mode deviation and positively correlated with the dot product of the absolute values ​​of the difference vectors between the preset weight vector, the current real-time production state vector, and the centroid of the most recent qualified production mode. A tiered early warning system is implemented based on the relationship between the comprehensive risk level and the preset risk threshold to monitor asphalt mixing plants.

2. The monitoring method for an asphalt mixing plant according to claim 1, characterized in that, The process of clustering the training dataset to obtain a qualified pattern set includes: processing the training dataset using the DBSCAN clustering algorithm, taking all non-noise point clusters as qualified production pattern clusters, and taking the set of all qualified production pattern clusters as the qualified production pattern set.

3. The monitoring method for an asphalt mixing plant according to claim 1, characterized in that, The formula for calculating the pattern deviation is: In the formula, The deviation of the model at each time point, This represents the real-time production status vector at each point in time. For the set of qualified patterns, For the set of qualified patterns The Middle The centroid of a qualified production mode cluster Real-time production state vector and qualified mode set The Middle The centroid of a qualified production model cluster The distance between them Minimum value function The process fluctuation of the data collection cycle at each time point is the last time point.

4. The monitoring method for an asphalt mixing plant according to claim 1, characterized in that, The method for obtaining the comprehensive risk level is as follows: For each time point, calculate the vector composed of the absolute values ​​of the differences between the elements of the real-time production state vector and the centroid of the qualified production mode cluster closest to the real-time production state vector, and calculate the product of the dot product of the vector and the preset sensitivity weight vector and the mode deviation degree as the comprehensive risk level for each time point.

5. The monitoring method for an asphalt mixing plant according to claim 4, characterized in that, The warning includes: issuing warnings for parameters whose weight values ​​in the preset sensitivity weight vector are within the preset weight percentage, and whose absolute values ​​in the vector formed by the absolute values ​​of the differences between the elements of the real-time production state vector and the centroid of the nearest qualified production mode cluster are within the preset weight percentage.

6. The monitoring method for an asphalt mixing plant according to claim 1, characterized in that, The real-time production operation data includes at least: the frequency of the feeding belt conveyor of each cold material silo; the flame temperature of the dryer burner; the aggregate temperature at the outlet of the dryer; the instantaneous weight of the asphalt weighing hopper; the asphalt supply temperature; and the motor current of the mixing cylinder.

7. The monitoring method for an asphalt mixing plant according to claim 4, characterized in that, The preset sensitivity weight vector is determined based on the degree of influence of production parameters on the quality of finished asphalt products.

8. The monitoring method for an asphalt mixing plant according to claim 1, characterized in that, The method of hierarchical early warning based on the relationship between the comprehensive risk level and the preset risk threshold includes: normalizing the comprehensive risk level at all time points, and issuing low-risk alarm, medium-risk alarm and high-risk alarm respectively when the normalized comprehensive risk level exceeds the preset low-risk threshold, medium-risk threshold and high-risk threshold respectively.

9. A monitoring method for an asphalt mixing plant according to claim 4, characterized in that, The values ​​of each element in the preset sensitivity weight vector correspond in order to the types of real-time production and operation data.

10. A monitoring system for an asphalt mixing plant, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement an asphalt mixing plant monitoring method according to any one of claims 1-9.

Images