SLUG FLOW VELOCITY ESTIMATION DURING PNEUMATIC CONVEYING OF BULK SOLID MATERIALS BASED ON IMAGE PROCESSING TECHNIQUES

. The article presents the use of fast cameras and image processing and analysis methods to determine the velocity of pneumatic transport of bulk material. The presented solution is dedicated to the analysis of dense flow in the form of slugs moving in a horizontal section of the pipeline. The developed image-processing algorithm is based on the estimation of the bulk solid material level in the pipeline in particular moments of the process. The acquisition of an image containing the area of two separate sections of the pipeline allows to determine the shift-time of material between two pre-defined pipeline regions. The obtained results indicate the proper action of the developed system.


Introduction
The dynamic development of many industries branches, like chemical, food, construction, mining, plastics, directly connected with new measurements techniques as answer to the need to monitor and control multiphase flows in technological processes [11]. For chemical process engineering a pneumatic transport, with particular emphasis on the transport of bulk solid materials over long distances, is very useful and safety type of gas-solid flow [19,13]. Pneumatic conveying, process of transporting bulk materials in a gas stream through a pipeline, is applied in industry, where granular material is main component of final product or occurs in intermediate stage of technological process. The biggest advantage of used of this type of flow is its efficiency and robustness in transfer of granular materials over considerable distances. Additionally, proper designed pneumatic conveying system allows to achieve very low energy consumption. In order to achieve optimal flow control necessary is to determine the conveying characteristics of the bulk solid material. Results of better understanding of flow behavior are directly visible in better flow control system, providing new knowledge about process. The proper design of conveying system, in based on knowledge about process, causes increasing of flow efficiency. However, even correct designed conveying system needs flow control unit. The possibility of directly application of new measurements techniques, which are capable to enrich data in control unit, are very useful to prevent unwanted situation. This is extremely important due to the possibility e.g. of reduced throughput of conveying system, or even blockage of installation. Such situations can causes stopping production process. Other troubleshooting problems are abrasive wear of pipeline, granular particle attrition, which can lead to low quality of product. The occurrence of such a situations may mean a halt in the production process and, in the worst case, damage to the installation. That is why it is so important to constantly monitor the entire process so that situations can be eliminated or properly controlled so that the whole process of flow to react to such phenomena [1]. The control unit in pneumatic conveying system mainly based on analysis of relation between the volume of air in the transport system and amount of bulk solid material flowing through the installation.
In order to better understand process necessary is to have full information about spatial-temporal changes of transported bulk solid material during changes of process conditions. Such information can be provide by apply industrial imaging techniques like process tomography or vision systems [4,3,8,9,20].

Industrial flow process visualization
The flow imaging technologies include two main development paths, based on camera vision and process tomography systems. Both solutions provide information about process in case of images, which can be processed and analyzed to determined flow parameters [5,23,2,7]. The obtained image of process is formed in different way. The camera visions system needs transparent installation; process tomography system can be applied for opaque installation. Depend on type of installation and aim of flow imaging, process investigation or control, choice of proper imaging technologies can be crucial for achieving the assumed objectives.
Often, techniques using capacitance of electrical tomography systems (ECT) are used to analyze such processes, [21,10,22,12,14,18]. It is a helpful to visualize the distribution of transported material. In this way, you can study many different parameters, such as the level of occupancy the pipeline or calculate the speed of moving slugs. However, this is not the only method used to analyze flow parameters. Another equally interesting method is the acquisition of flow using CCD cameras. These are cameras that can record at 100 fps. The camera records the entire flow in the form of a film. Then, the recorded data is processed further using image processing algorithms. It is with their help that we are able to extract from the picture relevant information about the course. A more detailed description of the processing is in [6]. New image processing methods are still being developed to solve various problems. One example is the use of image processing to calculate the translational speed of the screw and the length of the liquid screw in the horizontal flow of the air water pipe (developed to quantitatively compute the translational velocity and slug liquid length in horizontal air-water pipe flow) [24].
Another example of the use of fast cameras is their use in flows where the granularity of loose materials is not of a regular shape [14]. The high-speed camera also works well in solids deposition in low-velocity slug flow pneumatic conveying [16]. They are also used to study Low-velocity pneumatic transport of bulk solids [15]. As present, there is a wide range of CCD cameras in analyzing multi-phase flows, both air-solid and air-water.
The present article describes the use of CCD camera and image processing for determine slug velocity during pneumatic conveying of bulk solid materials. The measurement section contains transparent sections of the horizontal pipeline. There is visible Electrical Capacitance Tomography sensor (ECT sensor), in the middle of the section, which divide image into two areas, A-in front of sensor, Bin behind of sensor.

Image processing algorithm
The first important task of the designed image processing algorithm is the percentage estimate of the settled layer of granular material in the lower part of the horizontal transparent section of the pipeline. The area monitored by the camera, divided into two sections, allows to analysis the flow in separate two localizations. Using this fact, the algorithm responsible for image processing estimates flow state based on image area in the behind of sensor (area A) and in the front of sensor (area B).  Figure 2 shows the successive steps of the algorithm, which lead to the analysis of the flow inside the pipeline registered by the camera. The algorithm tries to detect and isolate transported material in both parts from recorded images. Then the average occupancy percentage is calculated in the section in front of the sensor and in the section after the sensor in real time. On the basis of these two information, the average percentage of material settled on the bottom of the pipeline is also calculated. In the first step of the presented algorithm, Fig. 2, the raw frame of the registered image is read in. It is shown in Fig. 4. The goal is to isolate and process only those parts of the image on which the transported material is located. Therefore, subsequent operations are aimed at achieving this result. The next step in the process includes initial correction, image preparation and focuses on solving the previously described assumption. As part of this process, pre-processing operations are carried out to facilitate and speed up the further image processing. This process involves separating those parts of the image that bring unnecessary noise and increase the number of operations that need to be done in the next steps [17].
The video frame with an empty pipeline ( Figure 3) is used to remove the background from images on which the material being transported is seen (Figure 4). The result of this operation is shown ( Figure 5) where the areas in which the material is transported are shown, realizing the assumed goal. In the next step, the algorithm monitors and processes both pipeline sections before and after the ECT sensor.
The next stage focuses on operations related to image thresholding and mask preparation. The result of this step is adequately depicted in the Fig 6. More details can be found in article [16]. This is an important element of the whole algorithm, because thanks to well-chosen filters and the thresholding method it helps to achieve the intended goal. The goal is to separate the image containing only material in the stream from the rest of the image. As a result of this stage of the algorithm, two areas have been designated which form the basis for further image processing.
The next stage of the algorithm involves processing the area in front of (area A) and behind the sensor (area B), the purpose of which is to determine the percentage occupancy level in both sections for each image frame being processed. Based on the level of material filling and determining the height of the pipeline, the algorithm is able to determine the percentage proportions of the material to the height of the pipeline. The result from this stage is shown in Fig. 7  Each section is additionally marked with a red contour in order to better visualize the achieved effect. Under each section there is a calculated percentage of filling in each of the monitored sections. The whole process is repeated for each new image frame loaded by the algorithm so that for each section, the average occupancy level of the pipeline is calculated in real time as shown in Figure 7. The algorithm calculates the percentage occupancy in the section in front of the sensor behind the sensor and outlines the areas that are processed At each moment, a processed image representing the level of fulfillment the pipeline is presented. The processed image is presented for each loaded raw frame, it facilitated verification of the extent to which the areas A and B were filled. Placing the obtained results on the graph showing changes in percentage fill in time where on the x-axis the time from the moment of starting the installation to transfer bulk material is presented. On the other hand, the y axis will display the percentages of pipeline occupancy for each section.  The material is transported in direction with the arrows in the drawing (from right to left). Both sections have a different occupancy level section on the right (before the sensor) 87.4% and the section on the left (behind the sensor) 23.1%. Above the pipeline is a chart showing how the received data is placed on the graph.
The final step of the algorithm deals with placing the all processed data on the graph in order to present the entire process flow. The selected flow fragment is presented below to present the results achieved. The result is presented in Fig. 9.

Result and discussion
In figure 9 is shown the graph of the percentage occupancy in time for the area in front of the sensor and after the ECT sensor representing the final results. The figure shows a fragment of the processed bulk material transfer process. Three slugs of moving material are visible on it. The Δt1, Δt2 and Δt3 offsets have been marked on the graph. Analyzing the shift between the readings from both monitored sections Δt1, Δt2 and Δt3 based on the difference in the number of frames for each Δt1, Δt2 and Δt3 offset, the approximate velocities of material movement were determined. Knowing the length of the ECT sensor, the sliding speed was calculated for each slug flow. The approximate speed was calculated for each displacement separately. They are within the range of 0.8-1.2 m/s.

Conclusion
The purpose of this article was to show how the speed of moving slug in bulk materials can be determined using a recorded image using fast cameras. A fast 100kps recording camera was used for the acquisition. The installed lighting in the tested part of the pipeline allows for a better image quality. Its contents present individual steps of the process that were carried out with the sample results obtained at each step of the process.
Also shown is the method in which the final result was obtained in the form of a graph presenting the percentage level of occupancy for each moment of the whole process, on which readings from both parts of the pipeline are presented. Based on the displacement between the readings from sections A and B which marked on the chart, the velocity of the slug flow was pre-calculated.