An intelligent control system and method for filling slurry concentration based on multi-source information

Abstract

The present invention provides an intelligent control system and method for filling slurry concentration based on multi-source information, which belongs to the field of mine filling technology. The control system includes a stirring system, an online monitoring system for rheological parameters, a spectral analysis system, an image acquisition system and a concentration control system. When applied, firstly, data fusion is performed on three types of concentration data of rheological parameters, spectrum and image characteristics, and a multi-source heterogeneous data fusion method based on the background of uncertainty conflict is constructed. Then, the concentration of the slurry is analyzed, and the concentration is weighted and fused, and finally the real-time concentration value of the slurry is obtained. Finally, the filling slurry concentration control system realizes intelligent control of the slurry concentration through automatic adjustment of process links such as aggregate addition and cementitious material addition according to the comparison results. By integrating the results of multi-source information analysis, the system greatly improves the accuracy and timeliness of the concentration measurement results and greatly improves the production efficiency of the filling system.

Description

Technical Field

[0001] The present invention relates to the technical field of mine filling, and in particular to an intelligent control system and method for filling slurry concentration based on multi-source information. Background Art

[0002] Mining has made a huge contribution to the progress of the world economy. Mining activities generate a large amount of tailings, waste rock and other solid wastes when obtaining resources. Solid wastes are usually treated by the so-called "upstream method" to save costs, which is a relatively dangerous disposal method. If the tailings dam is damaged, it will pose a major risk to the safety of downstream personnel. At the same time, large-scale underground mining activities will form a large number of goafs. Goaf collapse and solid waste accumulation are the main problems facing mines today.

[0003] Backfill mining is an important way to achieve green development of mines. Mine backfilling can effectively prevent and control the collapse of goaf areas, eliminate tailings pond disasters, and realize a new disaster management model of "one waste to treat two harms". However, the concentration measuring instrument of the backfill slurry with ultra-high solid content has problems such as large error, low accuracy, and easy failure. In addition, it is difficult to accurately control the concentration of the backfill slurry, which seriously restricts the development of metal mine backfilling technology.

[0004] Generally speaking, filling materials include tailings (slurry), waste rock, fly ash, cement and water, etc., multiple items and multiple types of materials. There are many sources of materials, and the quality fluctuates (such as different water content), which directly affects the accurate supply of filling materials, causing the concentration of slurry to fluctuate up and down. The concentration of filling slurry is difficult to control and fluctuates greatly. In order to ensure the operation of the system and avoid precipitation blockage, sudden changes in transportation resistance, etc., the operating parameters (or indicators) are repeatedly relaxed, and sometimes even fail to meet the filling standards. As a result, the filling is stratified and segregated, the strength distribution of the filling body is uneven, and the stability of the filling body is poor, which endangers the safety of production operations.

[0005] Existing filling slurry concentration measurement instruments are mostly based on single modal information for measurement, which has many limitations and uncertainties. The filling slurry concentration measurement technology based on multivariate information fusion analysis proposed in the present invention realizes concentration information characterization for slurries of different concentrations from three different aspects (rheological data, spectral intensity information, filling slurry morphological characteristics). Fusion of three different discrimination models from a probabilistic perspective will inevitably achieve a strong combination effect, greatly increasing the recognition accuracy of the concentration detection system. Since the multivariate data fusion analysis technology in the present invention integrates advanced machine learning and artificial intelligence technologies, it will inevitably greatly improve the speed of concentration measurement analysis and control. It will provide good support for the study of filling slurry concentration fluctuation mechanism and precursor information identification, and has certain reference significance for the development of my country's intelligent filling mining technology. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide an intelligent control system and method for filling slurry concentration based on multi-source information. The system extracts multi-source information of slurry concentration and adopts data fusion analysis method to conduct a comprehensive analysis of the filling slurry data to obtain highly accurate analysis results, thereby realizing precise and intelligent control of the filling slurry concentration.

[0007] The system includes four online subsystems: stirring system, rheological parameter online monitoring system, spectral analysis system, image acquisition system and concentration control system, as well as a concentration analysis system for slurry concentration calculation and a front-end feeding system; the stirring system includes a primary mixer and a secondary mixer;

[0008] The secondary mixer is placed below the discharge port of the primary mixer, and the rheological parameter online monitoring system is set below the discharge port of the secondary mixer; the spectral analysis system is set just above the observation port of the secondary mixer to perform non-contact spectral analysis on the evenly stirred slurry; the image acquisition system is set just above the slurry observation port of the secondary mixer to acquire images and videos of the slurry in real time; the outlet of the front-end feeding system is set above the primary mixer; the concentration analysis system is set in the central control room to transmit data with the rheological parameter monitoring system, spectral analysis system and image acquisition system through signals; the concentration control system controls the primary mixer, the secondary mixer and the front-end feeding system through signals.

[0009] The rheological parameter online monitoring system is a micro annular tube formed by connecting three pipelines of different diameters in series. The rheological parameters of the slurry are calculated by measuring the slurry flow rate flowing into the annular tube and the pressure data of the front and rear ends of the three diameter pipelines.

[0010] The three pipelines with different diameters are all 1 m long, and have diameters of 30 mm, 40 mm and 50 mm respectively. The 30 mm, 40 mm and 50 mm pipelines are connected in series in sequence to form a ring pipe.

[0011] The spectral analysis system uses a high-spectral scanning device to extract and analyze the spectral component characteristics of the slurry.

[0012] The image acquisition system uses a high-speed camera to obtain the slurry image and then identifies and extracts the image features based on the optical flow method.

[0013] The concentration analysis system has a built-in Dempster combined decision-making method based on the combination of probability distance function and classical conflict coefficient to conduct a comprehensive analysis of the concentration characteristics of multi-source filling slurry.

[0014] The rheological parameter monitoring system inverts the rheological parameters of the slurry based on the flow and pressure measured by the bypass small loop pipe. The bypass small loop pipe system is installed at the outlet of the secondary mixer and connected by a gate valve. The slurry can return to the secondary mixer after flowing through the small loop pipe device.

[0015] The concentration control system mines multi-source real-time concentration data to fuse information, establishes a precise concentration calculation model, combines the PLC cabinet with online automatic control of the gate valve and information remote communication, adjusts the system operating parameters in real time, and realizes precise and intelligent control of the front-end feeding system.

[0016] The concentration control system can also display the tailings slurry intelligent control strategy, filling slurry stirring intelligent control strategy and concentration control value in real time through control strategy visualization technology.

[0017] The application method of the system comprises the following steps:

[0018] S1: Multi-source information acquisition:

[0019] Through online rheological property testing, hyperspectral scanning and high-speed camera shooting, a collection of evidence of filling slurry concentration characteristics based on rheological parameters, spectrum and image concentration information is extracted and established, and then the matching probability is normalized so that the judgment value of each characteristic value falls between [0,1], forming a basic probability distribution value of each concentration, i.e., concentration characteristic evidence; the evaluation of the three characteristic information data of rheological parameters, spectrum and image concentration information is recorded as m1, m2, and m3 respectively, and the basic trust probability distribution (BPA) of these three characteristic information data is: m1(A1), m2(A2) and m3(A3), where A1, A2, A3 are the evidence collections of the above three characteristic information data respectively and m1(A1)+m2(A2)+m3(A3)=1;

[0020] S2: Reliability judgment of multi-source information:

[0021] The uncertainty of the evidence set of the three feature information data in S1 is measured by the belief function Bel(A) and the likelihood function Pl(A):

[0022] The credibility function is expressed as:

[0023]

[0024] The likelihood function is expressed as:

[0025] Pl(A)=∑m(B),B∩A≠φ

[0026] Among them, A represents a certain type of evidence set of feature data; B represents a certain piece of evidence in the evidence set A; m(B) represents the basic trust probability of a certain piece of evidence; the belief function value Bel(A) represents the degree of trust that the evidence set A is true, and the likelihood function value Pl(A) represents the degree of trust that the evidence set A is not false, thereby judging the reliability of each piece of evidence set;

[0027] S3: Use the Dempster combination rule that combines the support probability distance function and the classical conflict coefficient to determine the conflict degree of evidence based on the conflict value K among various pieces of evidence. The calculation formula for the K value is as follows:

[0028] K = ∑m1(A1)m2(A2)m3(A3), A1 ∩ A2 ∩ A3 = φ

[0029] Among them, when 0 < K < 0.65, it is a non-complete conflict and data fusion can be performed; when 0.65 < K ≤ 1, it is a high conflict and not applicable to data fusion.

[0030] S4: Multi-source data fusion:

[0031] When the evidence bodies are not in complete conflict, use the Dempster combination rule for data fusion. When there are conflicts among the evidence bodies, use an improved combination method based on the support probability distance for data fusion, and construct a multi-source information data set with the fused concentration characteristics;

[0032] S5: Multi-source data analysis:

[0033] Build a hybrid deep neural network concentration analysis method DNN-CNN based on matlab. This analysis method consists of a rheological parameter analysis network, a hyperspectral feature analysis network, and an image feature analysis network. Among them, the concentration characteristics based on rheological parameters are analyzed by a deep neural network, and the hyperspectral and image features are analyzed by convolutional neural networks respectively, so as to obtain three kinds of concentration data with labels based on rheological parameters, spectra, and images; finally, adjust the sub-networks of the above three features according to the weight values calculated by the AHP (Analytic Hierarchy Process) to balance the contribution sizes of each concentration sub-network;

[0034] S6: Input the three kinds of concentration data with labels based on rheological parameters, spectra, and images obtained in S5 into the DNN-CNN model. Then train the DNN-CNN model to make it reach a relatively high training concentration analysis value, and then input real-time data into the model through the multi-source information data set to realize the real-time detection of the filling slurry concentration;

[0035] S7: Compare and analyze the standard value of the filling slurry concentration with the calculated value of the DNN-CNN model, and feedback it to the concentration control system;

[0036] S8: Realize the precise addition of materials through the control device, and realize the real-time display of the intelligent control strategy of the slurry and the concentration control value based on the control strategy visualization technology.

[0037] The relatively high training concentration analysis value in S6 refers to the concentration analysis value when the determination coefficient R 2 > 0.97 of the DNN-CNN model.

[0038] The beneficial effects of the above technical solution of the present invention are as follows:

[0039] 1. In terms of distributed information acquisition. Existing filling slurry concentration measuring devices mostly use a single method to detect the filling slurry concentration, which makes it difficult to achieve online accurate measurement. The three data acquisition systems of the present invention (rheological parameter measurement, image features, and spectrum recognition) adopt a distributed arrangement, with independent data acquisition processes and integrated data analysis processes, which can effectively improve the system reliability and the accuracy of the measured concentration data.

[0040] 2. Real-time analysis of multi-source information. The present invention involves multi-source information such as pipeline flow, pressure difference, image, video and spectrum, integrates deep learning algorithm into machine vision technology, and can realize real-time analysis of multi-source information through the computing power of high-performance servers and GPU resources. At the same time, the analysis results of multi-source information are integrated, which greatly improves the accuracy and timeliness of concentration measurement results.

[0041] 3. In terms of intelligent concentration control system. The low degree of automation of existing traditional concentration measurement, filling system industrial control technology and supporting equipment has become the main bottleneck restricting the filling process. The concentration measurement instrument and technology of the present invention adopts advanced machine vision, artificial intelligence, deep learning and other modern computer science and technology. Through the research and development of relevant algorithms and control programs, it can eventually realize the "precision, unmanned, and intelligent" control of the filling system concentration and even the entire filling system, greatly improving the production efficiency of the filling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the structure of the intelligent control system of filling slurry concentration based on multi-source information of the present invention;

[0043] Figure 2 This is a flow chart of multi-source information analysis of the present invention.

[0044] Among them: 1-first-stage mixer; 2-second-stage mixer; 3-rheological parameter online monitoring system; 4-spectral analysis system; 5-image acquisition system; 6-concentration control system; 7-concentration analysis system; 8-front-end feeding system. DETAILED DESCRIPTION

[0045] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0046] The present invention provides a filling slurry concentration intelligent control system and method based on multi-source information.

[0047] like Figure 1As shown, the system includes four online subsystems, namely, a stirring system, an online monitoring system for rheological parameters 3, a spectral analysis system 4, an image acquisition system 5 and a concentration control system 6, as well as a concentration analysis system 7 for calculating slurry concentration and a front-end feeding system 8; wherein the stirring system includes a primary mixer 1 and a secondary mixer 2;

[0048] The secondary mixer 2 is placed below the discharge port of the primary mixer 1, and the rheological parameter online monitoring system 3 is arranged below the discharge port of the secondary mixer 2 to measure the rheological parameters of the filling slurry that is about to enter the pipeline for transportation; the spectral analysis system 4 is arranged just above the observation port of the secondary mixer 2 to perform non-contact spectral analysis on the evenly stirred slurry; the image acquisition system 5 is arranged just above the slurry observation port of the secondary mixer 2 to acquire images and videos of the slurry in real time; the outlet of the front-end feeding system 8 is arranged above the primary mixer 1; the concentration analysis system 7 is arranged in the central control room, and data is transmitted through signals with the rheological parameter monitoring system 3, the spectral analysis system 4, and the image acquisition system 5, and multi-source information data is processed and analyzed to obtain the slurry concentration value; the concentration control system 6 controls the primary mixer 1, the secondary mixer 2 and the front-end feeding system 8 through signals.

[0049] The rheological parameter online monitoring system 3 is a micro-annular tube formed by three different diameter pipelines connected in series. The rheological parameters of the slurry are calculated by measuring the slurry flow rate flowing into the annular tube and the pressure data at the front and rear ends of the three diameter pipelines. The length of the three different diameter pipelines is 1m, and the diameters are 30mm, 40mm and 50mm respectively. The 30mm, 40mm and 50mm pipelines are connected in series in sequence to form an annular tube.

[0050] The spectral analysis system 4 uses a high-spectral scanning device to extract and analyze the spectral component characteristics of the slurry.

[0051] After the image acquisition system 5 obtains the slurry image using a high-speed camera, it identifies and extracts the image features based on the optical flow method.

[0052] The concentration analysis system 7 has a built-in Dempster combined decision-making method based on a combination of a probability distance function and a classical conflict coefficient to conduct a comprehensive analysis of the concentration characteristics of the multi-source filling slurry.

[0053] The rheological parameter monitoring system 3 inverts the rheological parameters of the slurry based on the flow and pressure measured by the bypass small annular pipe. The bypass small annular pipe system is installed at the outlet of the secondary mixer 2 and is connected through a gate valve. The slurry can return to the secondary mixer 2 after flowing through the small annular pipe device.

[0054] The concentration control system 6 mines multi-source real-time concentration data fusion information, establishes a concentration precision calculation model, combines the PLC cabinet online automatic control gate valve and information remote communication, adjusts the system operation parameters in real time, and realizes precise and intelligent control of the front-end feeding system 8. The concentration control system 6 can display the tailings slurry intelligent control strategy, the filling slurry stirring intelligent control strategy and the concentration control value in real time through the control strategy visualization technology.

[0055] like Figure 2 As shown, the system application method includes the following steps:

[0056] S1: Multi-source information acquisition:

[0057] Through online rheological property testing, hyperspectral scanning and high-speed camera shooting, a collection of evidence of filling slurry concentration characteristics based on rheological parameters, spectrum and image concentration information is extracted and established, and then the matching probability is normalized so that the judgment value of each characteristic value falls between [0,1], forming a basic probability distribution value of each concentration, i.e., concentration characteristic evidence; the evaluation of the three characteristic information data of rheological parameters, spectrum and image concentration information is recorded as m1, m2, and m3 respectively, and the basic trust probability distribution (BPA) of these three characteristic information data is: m1(A1), m2(A2) and m3(A3), where A1, A2, A3 are the evidence collections of the above three characteristic information data respectively and m1(A1)+m2(A2)+m3(A3)=1;

[0058] S2: Reliability judgment of multi-source information:

[0059] The uncertainty of the three evidence sets in S1 is measured by the belief function Bel(A) and the likelihood function Pl(A):

[0060] The credibility function is expressed as:

[0061]

[0062] The likelihood function is expressed as:

[0063] Pl(A)=∑m(B),B∩A≠φ

[0064] Among them, A represents a certain type of evidence set of feature data; B represents a certain piece of evidence in the evidence set A; m(B) represents the basic trust probability of a certain piece of evidence; the belief function value Bel(A) represents the degree of trust that the evidence set A is true, and the likelihood function value Pl(A) represents the degree of trust that the evidence set A is not false, thereby judging the reliability of each piece of evidence set;

[0065] S3: Determine the conflict degree of evidence by the conflict value K of each evidence using the Dempster combination rule that combines the support probability distance function and the classical conflict coefficient. The calculation formula of the K value is as follows:

[0066] K = ∑m1(A1)m2(A2)m3(A3), A1 ∩ A2A3 = φ

[0067] Among them, when 0 < K < 0.65, it is a non-complete conflict and data fusion can be performed; when 0.65 < K ≤ 1, it is a high conflict and not applicable to data fusion.

[0068] S4: Multi-source data fusion:

[0069] When the evidence bodies are not completely in conflict, use the Dempster combination rule for data fusion. When there are conflicts between the evidence bodies, use an improved combination method based on the support probability distance for data fusion, and construct a multi-source information data set with the fused concentration characteristics;

[0070] S5: Multi-source data analysis:

[0071] Construct a mixed deep neural network concentration analysis method DNN-CNN based on matlab. This analysis method consists of a rheological parameter analysis network, a hyperspectral feature analysis network, and an image feature analysis network. Among them, the concentration characteristics based on rheological parameters are analyzed by a deep neural network, and the hyperspectral and image features are analyzed by convolutional neural networks respectively, so as to obtain three kinds of concentration data with labels based on rheological parameters, spectra, and images; finally, adjust the sub-networks of the above three features according to the weight values calculated by the AHP analytic hierarchy process to balance the contribution sizes of each concentration sub-network;

[0072] S6: Input the three kinds of concentration data with labels based on rheological parameters, spectra, and images obtained in S5 into the DNN-CNN model. Then train the DNN-CNN model to make it reach a relatively high training concentration analysis value, and then input real-time data into the model through the multi-source information data set to realize the real-time detection of the filling slurry concentration;

[0073] S7: Compare and analyze the standard value of the filling slurry concentration (usually a standard concentration value is preset in advance) with the calculated value of the DNN-CNN model, and feedback it to the concentration control system;

[0074] S8: Realize precise material addition through the control device, and realize the real-time display of the intelligent control strategy of the slurry and the concentration control value based on the control strategy visualization technology.

[0075] The relatively high training concentration analysis value in S6 above refers to the concentration analysis value when the determination coefficient R 2 > 0.97 of the DNN-CNN model.

[0076] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent control system for filling slurry concentration based on multi-source information, characterized in that: It includes four online subsystems: stirring system, rheological parameter online monitoring system, spectral analysis system, image acquisition system and concentration control system, as well as concentration analysis system and front-end feeding system for slurry concentration calculation; the stirring system includes a primary mixer and a secondary mixer; The secondary mixer is placed below the discharge port of the primary mixer, and the rheological parameter online monitoring system is set below the discharge port of the secondary mixer to measure the rheological parameters of the filling slurry that is about to enter the pipeline for transportation; the spectral analysis system is set just above the observation port of the secondary mixer to perform non-contact spectral analysis on the evenly stirred slurry; the image acquisition system is set just above the slurry observation port of the secondary mixer to obtain images and videos of the slurry in real time; the outlet of the front-end feeding system is set above the primary mixer; the concentration analysis system is set in the central control room, and the signal is used to transmit data with the rheological parameter monitoring system, spectral analysis system, and image acquisition system, and the multi-source information data is processed and analyzed to obtain the slurry concentration value; the concentration control system controls the primary mixer, the secondary mixer, and the front-end feeding system through signals; The rheological parameter online monitoring system is a micro-annular tube formed by connecting three different diameter pipelines in series. The rheological parameters of the slurry are calculated by measuring the slurry flow rate flowing into the annular tube and the pressure data at the front and rear ends of the three diameter pipelines. The spectrum analysis system uses a high-spectrum scanning device to extract and analyze the spectrum component characteristics of the slurry; The image acquisition system obtains the slurry image with a high-speed camera and then identifies and extracts the image features based on the optical flow method.

2. The intelligent control system for filling slurry concentration based on multi-source information according to claim 1 is characterized in that: The three pipelines with different diameters are all 1 m long, and have diameters of 30 mm, 40 mm and 50 mm respectively. The 30 mm, 40 mm and 50 mm pipelines are connected in series in sequence to form a ring pipe.

3. The intelligent control system for filling slurry concentration based on multi-source information according to claim 1, characterized in that: The rheological parameter monitoring system inverts the rheological parameters of the slurry based on the flow and pressure measured by the bypass small annular pipe, wherein the bypass small annular pipe system is installed at the outlet of the secondary mixer and connected by a gate valve. The slurry can return to the secondary mixer after flowing through the small annular pipe device.

4. The application method of the intelligent control system for filling slurry concentration based on multi-source information according to claim 1 is characterized in that: The steps include: S1: Multi-source information acquisition: Through online rheological property testing, hyperspectral scanning and high-speed camera shooting, a collection of filling slurry concentration feature evidence based on rheological parameters, spectrum and image concentration information is extracted and established, and then the matching probability is normalized so that the judgment value of each characteristic value falls between [0,1], forming a basic probability distribution value of each concentration, i.e., concentration feature evidence; the evaluation of the three characteristic information data of rheological parameters, spectrum and image concentration information is recorded as m1, m2, and m3 respectively, and the basic trust probability distribution of these three characteristic information data is: m1(A1), m2(A2) and m3(A3), where A1, A2, A3 are respectively the evidence collections of the above three characteristic information data and m1(A1)+m2(A2)+m3(A3)=1; S2: Reliability judgment of multi-source information: The uncertainty of the evidence set of the three feature information data in S1 is measured by the belief function Bel(A) and the likelihood function Pl(A): The credibility function is expressed as: The likelihood function is expressed as: Pl(A) = ∑m(B), B∩A≠φ Where, A represents the evidence aggregate of a certain characteristic data; B represents a certain evidence in the evidence aggregate A; m(B) represents the basic belief probability of a certain evidence; the belief function value Bel(A) represents the degree of belief that the evidence is true for the evidence aggregate A, and the plausibility function value Pl(A) represents the degree of belief that the evidence is not false for the evidence aggregate A, so as to judge the reliability of each evidence aggregate; S3: Use the Dempster combination rule that combines the support probability distance function and the classical conflict coefficient to determine the conflict degree of evidence through the conflict value K between each evidence. The calculation formula of the K value is as follows: K = ∑m1(A1)m2(A2)m3(A3), A1∩A2∩A3 = φ Where, when 0 < K < 0.65, it is a non-complete conflict and data fusion can be performed; when 0.65 < K ≤ 1, it is a high conflict and is not applicable to data fusion; S4: Multi-source data fusion: When the evidence bodies are non-completely conflicting, use the Dempster combination rule for data fusion. When there are conflicts between the evidence bodies, use an improved combination method based on the support probability distance for data fusion, and construct a multi-source information data set with the fused concentration characteristics; S5: Multi-source data analysis: Build a hybrid deep neural network concentration analysis method DNN-CNN based on matlab. This analysis method consists of a rheological parameter analysis network, a hyperspectral feature analysis network, and an image feature analysis network. Among them, the concentration characteristics based on rheological parameters are analyzed by a deep neural network, and the hyperspectral and image features are analyzed by a convolutional neural network respectively, so as to obtain three kinds of concentration data with labels based on rheological parameters, spectra, and images; finally, adjust the subnetworks of the above three features according to the weight values calculated by the AHP analytic hierarchy process to balance the contribution sizes of each concentration subnetwork; S6: Input the three kinds of concentration data with labels based on rheological parameters, spectra, and images obtained in S5 into the DNN-CNN model, then train the DNN-CNN model to make it reach a high training concentration analysis value, and then input the real-time data into the model through the multi-source information data set to realize the real-time detection of the filling slurry concentration; The higher training concentration analysis value refers to the DNN-CNN model determination coefficient R 2 >0.97 concentration analysis value S7: Compare and analyze the standard value of the filling slurry concentration with the calculated value of the DNN-CNN model, and feedback it to the concentration control system; S8: Realize precise material addition through the control device, and realize the real-time display of the intelligent control strategy of the slurry and the concentration control value based on the control strategy visualization technology.

Images