Application performance research
of the NU classifier in industrial production

1 Introduction

The classifier is a core device in a recirculation grinding system, and has a great influence on product quality, output and power consumption. In order to reduce the power consumption of a grinding system and improve the product quality, the classifier has been thoroughly studied at home and abroad [1-5]. The classifier with U shape blades is the first generation of centripetal classifier originally proposed and developed by the Tianjin Cement Industry Design and Research Institute Co., Ltd., and is based on a large number of CFD theoretical studies [6-11]. Because the moving blades of the NU classifier have a unique aerodynamic shape, they can move the air flow between adjacent moving blades to the inside of the rotor during the rotation of the rotor. This action generates centripetal dynamic pressure and helps the air flow through the rotor of the separator, resulting in a reduction in resistance. The common classifier has only one kind of pulling force from the fan. In contrast, the outer wind wing of the U-shaped moving blade has an additional pulling force on the particles in the classifying area, which allows the product particles to pass through the classifying area faster and improves the classifier’s efficiency. In addition, at the same rotor speed, the airflow in the classifying area of the NU classifier has a higher tangential speed than that of a common classifier because the moving blade has a unique groove structure. The function of the unique structure not only improves the classifying efficiency, but also ensures clear classification. Moreover, the NU classifier can obtain the same product fineness as common classifiers at a lower speed, which is very helpful for further improving the efficiency of powder selection and system production capacity.

The classifier with U shape blades was first installed at the TRMS43.3 vertical roller mill of Hebei Qianjin Metallurgy Technology Co., Ltd in October 2015, and then successively obtained industrial applications on nearly 20 mills of various specifications in Hebei Qianbao, Southwest Wanzhou, Zhejiang Xinminghua, Anhui Dajiang, Qilianshan Yongdeng, etc. This classifier reduces system power consumption by 10-12% or higher and achieves good industrial application results in these cement plants [10, 11]. However, the industrial application performance of classifiers with U shape blades has not been studied in detail because it is very difficult to sample the feed and return of the classifier configured on the vertical mill. The LV classifier is a well-known advanced separator which was introduced by the Tianjin Cement Industry Design and Research Institute Co., Ltd. in 2014. The performance comparison between the NU classifier and LV classifier is attracting the attention of grinding technicians and the majority of cement manufacturers in the market. Therefore, given the fact that the #1 grinding system (LV5021) and the #2 grinding system (NU5026) of Yuzhou Lingwei Cement Plant have the similar operating conditions, the industrial application performance of the NU classifier and the LV classifier is studied in this paper.

2 Experiment

2.1 Sampling

2.1.1 Sampling materials to be fed into the classifier

It is more difficult to sample materials to be fed into the classifier in a vertical roller mill system than in a ball mill system, because the classifier is part of the vertical roller mill equipment and the materials are carried into the classifier by the updraft inside the mill. According to the structural characteristics of the classifier on the vertical roller mill, a sampling hole with a diameter of 100 mm is opened at the same height as the lower end face of the classifier rotor. Samples are drawn with constant speed using a dust meter. The positions of the two mills’ sampling holes are slightly different because of the difference between two working platforms. According to the rotation direction of the air flow in the mill, the sampling position of #1 mill is located on the leeward side of the mill feeding chute, and the sampling position of #2 mill is located on the windward side of the feeding chute. Figure 1 shows the position and size of the sampling holes.

2.1.2 Sampling materials returned from the classifier

It is also more difficult to sample materials returned from the classifier in a vertical roller mill system than in a ball mill system, because the materials returned from the classifier are mixed with the mill feed in the inner cone, and then fed onto the grinding table through the mill’s central feeding tube. According to the characteristics of the materials returned by the classifier, a 160×100 mm oval hole is opened on the top of the cone in the classifier, and a 100 mm diameter steel pipe is connected to the hole outside the mill housing, in which there are two air lock valves. Figure 2 shows the position and appearance of the sampling tube.

Before sampling, air lock valve 1 and air lock valve 2 are both opened to empty the materials into the sampling tube, after which the two air lock valves are closed. After about 5 min, the air lock valves 1 and 2 are opened to release part of the materials returned from the classifier and then air lock valves 2 and 1 are closed in that order. Next, the sample bag at the outlet of the sampling tube is covered, and then air lock valve 2 is opened, so that the material between the two air lock valves enters the sample bag under gravity. A sample is completed through the above steps.

2.1.3 Sampling classifier product

The instant samples are taken directly from the sampling port of the chute that transports the product.

2.2 Performance test of powder concentrator

In this paper, the system test method is used to test the performance of the classifier in order to ensure the reliability and systemicity of the performance parameters for the classifier. Therefore, in addition to testing the resistance, efficiency, circulation load, k (resolution of classification), and product particle fineness of the #1 and #2 classifiers under basically the same working conditions, this paper also analyzes the operating data of the process system such as air volume, output, and power consumption during the test and historical periods.

2.2.1 Sample fineness testing

Under the stable operation and similar working conditions, #1 grinding system and #2 grinding system are sampled eight times respectively and three samples are collected each time. The R_45μm, R_80μm and R_200μm of each sample are tested with a negative pressure sieve analyzer on the production site. Particularly, the materials to be fed into the classifier and the materials returned from the classifier are firstly sieved by a 200 μm sieve, and the materials passing through the sieve are used to detect R_45μm and R_80μm.

2.2.2 Performance test of the classifier

The classifier is a typical process unit, whose performance is not only affected by its own process structure parameters, but also by the operating parameters of the other equipment in the grinding system. In order to improve the comparability of the performance parameters of the two separators, the air volumes of the two mill systems are first calibrated and the fan valves are adjusted to ensure that the two powder separators have basically the same working conditions during the test. Then, the two mills are set with the same operating parameters: the feed amount is 310 t/h, the roller pressure is 8.6-8.8 MPa, the speed of the classifier is 770 rpm. After the above operations are completed, the materials to be fed into the classifier, the materials returned from the classifier, and classifier products are respectively taken according to the sampling method of 2.1 under stable mill operation conditions, and a total of 8 effective samples are taken from each mill. In order to avoid errors, the test results of 8 groups of samples are statistically averaged, and the average values are used to calculate the performance parameters such as the powder selection efficiency and circulation load of each classifier. Table 2 shows the results.

The classifier performance test can be based on parameters such as classifying efficiency, circulation load, β (bypass value) and k (resolution of classification). At the same time, it is important to check the synergistic effect of the classifier on the mill and the grinding system, which is whether the classifier can reduce the power consumption of the grinding system. For this reason, the power consumption of the two mills during the test period of 2019.5.12 to 2019.5.15 and the monthly statistical power consumption of April 2015 are recorded and studied in detail. Table 3 shows the results.

3 Results and discussion

3.1 Product fineness comparison

Whether for raw materials or cement, the fineness of the finished product can affect not only the quality of the finished product, but also the power consumption of the grinding system. For cement raw materials, as long as the R_200μm can be controlled below 2%, R_80μm can be relaxed to 20% or even higher, so that the raw meal mill system can achieve the best energy saving effect. Under the same fineness control conditions, a fine classifier with excellent fineness control capability needs low speed and circulation load, which can reduce the power consumption of the mill and fan. In order to compare the fineness performance of the LV and NU classifiers, Table 1 shows the fineness of finished products of the samples obtained from #1 mill and #2 mill, which are tested by different test equipment and methods and then averaged statistically. The samples are taken under the conditions of feeding rate of 310 t/h, classifier speed of 770 rpm, roller pressure of 8.6-8.8 MPa and similar system air volume (the air volume of #1 mill is 461851m³/h, and the air volume of #2 mill is 495545 m³/h.). It can be found in Table 1 that no matter what kind of screening equipment and method is used, the fineness of the finished product of the NU classifier is finer than that of the LV classifier. The results show that R_80μm is reduced by 2%-3% while R_200μm is reduced by 0.15%-0.83%, and the decrease range reaches 12%-33%, which proves that the NU classifier has a significant technical advantage in fineness control.

3.2 Classifying efficiency comparison

The classifying efficiency is the core technical parameter of the classifier, which not only affects the production capacity and power consumption of the grinding system directly, but also affects the particle distribution of the product. The particle distribution has an important influence on the specific surface area and performance of cement, slag and other products. The classifying efficiency is closely related to the particle size of the particles. Strictly speaking, the evaluation of the classifying efficiency must limit the corresponding particle size. The Trump curve of the classifier reflects the classifying efficiency at different particle sizes, and it is also the performance curve that evaluates the classifying efficiency comprehensively. By processing the values of this curve, some important performance parameters such as the β (bypass value), the k (resolution of classification), and the X₅₀ (cut size) can be obtained.

In order to compare the performance of the NU and LV classifiers more intuitively, the sieve residual curve of the materials to be fed into the classifier, the materials returned from the classifier, and the product are sampled, as shown in Figures 3, 4, and 5 in that order. Figure 3 shows that the sieve residue of the LV classifier feed is about 10% lower than that of the NU in the range of particle size larger than 25 μm, and the sieve residue of the LV classifier feed is about 5% lower than that of the NU in the range of particle size smaller than 25 μm, which shows that the fineness of the materials fed into the NU classifier is much coarser than that fed into the LV classifier. Under the premise of the same mill feed rate, this result indicates that the air volume of the #1 mill (LV) system is smaller than that of the #2 mill (NU) system, which is also confirmed by the calibration results of the system air volume shown in Table 2 (The air volume of the #2 mill is 33694 m³ more than that of the #1 mill). In theory, the finished product will be finer, due to the finer materials fed into the classifier and the lower amount of air used in the system at the same speed of the classifier. However, the test results of the NU and LV classifier show the opposite conclusion, which further proves that the finished product fineness control of the NU classifier is excellent.

Figure 4 shows that the sieve residue of the materials returned from the NU classifier is about 2%-3% lower than that of the LV in the particle size range 50 μm to 200 μm, the sieve residue of the materials returned from the NU classifier is about 3%-5% higher than that of the LV in the particle size range 200 μm to 250 μm, and the sieve residue of the materials returned from the NU classifier is about 5%-10% higher than that of the LV in the particle size range larger than 250 μm. This result shows that the NU classifier has more advantages in preventing the entry of coarse particles above 200 μm into the product. For the fine powder below 50 μm, there is not much difference in the sieve balance of the two kinds of classifiers, which indicates that the two classifiers have the same ability to select fine powder. The NU classifier has not greatly reduced the efficiency of selecting fine powder due to the selection of coarse particles, so it has not affected the production capacity of the mill and the power consumption of the grinding system. Figure 5 intuitively shows that the fineness of the finished product of the NU classifier is smaller than that of the LV classifier. More specifically, the R_80μm of the NU classifier is about 2% to 3% lower than that of the LV, and the sieve residue of the NU classifier product is about 3%-5% lower than that of the LV in the particle size range 100 μm to 200 μm. This result shows that the product of the NU classifier is thinner than the product of the LV, especially in the particle size range of 100 μm to 200 μm, which is conducive to improving the burnability of raw meal.

Figure 6 shows the Trump curve of the NU and LV classifiers under the same conditions. As shown in Figure 6, the β (bypass value) of the LV classifier is 1.8%, which is better than the β (bypass value) of the NU classifier of 3.6%. However, it can be roughly estimated that the β (bypass value) of the LV classifier should be at least 5% lower than that of the NU, according to the previous conclusion that the material fed to the LV classifier is 5% to 10% finer than the material fed to the NU classifier. In fact, the β (bypass value) of the LV classifier is only 1.8% lower than that of the NU, so the LV classifier has no advantage over the NU classifier in the β (bypass value). In addition, if the finished product fineness is set to be the same, the NU classifier can obtain a clear advantage in the β (bypass value) compared with the LV by reducing the speed of operation. The k (resolution of classification) of the NU classifier is 0.6 and the k (resolution of classification) of the LV classifier is 0.58, which are not much different. Under the action of static electricity or molecular force, fine powder can adhere to the surface of large particles and adversely affect the k (resolution of classification) of the classifier. Therefore, it can basically be concluded that the NU classifier has great advantages in k (resolution of classification) if the factor is considered that the fineness of the material fed to the NU classifier is much coarser than that of the LV. The X₅₀ (cut size) of the NU classifier is 170 μm, which is significantly better than the X₅₀ (cut size) of the LV classifier of 183 μm. The result guarantees that the finished product fineness of the NU classifier must be better than that of the LV classifier from the classifying characteristics.

3.3 Comparison of classifier resistance and power consumption of the grinding system

As shown in Table 3, the resistance of the NU classifier is 2165 Pa and the resistance of the LV classifier is 2850 Pa under similar conditions. The resistance of the NU classifier is 685 Pa lower than that of the LV classifier, which is in line with the result of CFD theoretical calculations in the development stage that the resistance of the NU classifier is 700 Pa lower than that of the LV classifier [8, 10]. This result verifies the correctness of the CFD theoretical calculation method. At the same time, it also confirms the technical advantages of the first generation centripetal classifier NU classifier in reducing the resistance of powder selection through an advanced principle that moving blades help instead of blocking airflow through the rotors.

The positive effect of the classifier on the grinding system is mainly reflected in the aspects of increasing output and reducing power consumption. Based on the data in Table 3, the output of #2 mill is calculated after setting the product fineness of #2 mill to be the same as that of #1 mill. The theoretical calculation results show that #2 mill ratio #1 mill improves the output by 305.1×[1+ (22.48-19.59) × 2/100]-308.4=14.3 t/h. The output of the #2 grinding system may increase by 15 t/h-20 t/h if the positive effect of the fineness increase and the reduction of the rotation speed on the efficiency of classifying and grinding is considered.

The classifier has a positive effect on mill power consumption, fan power consumption and system power consumption. According to Table 3, the power consumption during the test period of 2019.5.12 - 2019.5.15 and the power consumption in April both show that the average power consumption of #2 mill is 0.4-0.5 kWh/t lower than that of #1 mill. However, the NU classifier has not shown an advantage in the power consumption of the fan because of the large air volume of #2 mill and the difference in efficiency between the two fans. In order to keep the air flow of the two mills similar, the fan valve of the #1 mill is fully open and the valve of the #2 mill is 80% open, which results in the deviation in the fan efficiency of #2 system and the increase in fan power. In April 2018, the cement plant owner opened the fan valve in #2 system to about 95%, which caused the power consumption of the fan in #2 system to be 1.5 kWh/t higher than that of #1 system.

The power consumption of #2 grinding system is roughly converted according to a coefficient of 1+ (22.48-19.59) × 2/100 = 1.058 after setting the product fineness of #2 mill to be the same as that of #1 mill. The converted power consumption of #2 grinding system is 14.98/1.058=14.16 kWh/t, so the power consumption of #2 grinding system is 0.5 6kWh/t (0.56=14.72-14.16) less than that of #1 grinding system. After the product fineness is converted to be the same, the power consumption of the mill equipped with the NU classifier is 0.92 kWh/t (0.92=7.42-6.88/1.058) less than that of the mill equipped with the LV classifier. It can be inferred that the power consumption of the grinding system may decrease by at least 1 kWh/t if there is no problem with fan efficiency.

In general, the NU classifier is superior to the LV classifier in terms of resistance, classifying efficiency, β (bypass value), fineness of finished products, and reduction of power consumption of the mill and system. This result verifies the advanced nature of the unique centripetal classifying principle of the NU classifier, and also provides a good technical solution for the technical upgrade of the grinding system.

4 Conclusion

In this paper, the industrial application performance of the NU classifier is studied by means of experimental tests, while the method and device for sampling feedstock, return materials and product of the vertical roller mill classifier are designed. At the same time, the experimental research method for the performance of the vertical roller mill classifier is also established. In addition, the performance comparison study of the NU classifier and the LV classifier is conducted under basically the same operating conditions. The conclusions reached are as follows:

1) The sampling method of the vertical roller mill classifier using the dust meter and the inner cone guide tube design is effective and practical, enabling the experimental research of important performance parameters such as powder selection efficiency and Trump curve of vertical roller mill classifier to be realized basically.

2) Under the same conditions, the resistance of the NU classifier is 24% lower than that of the LV classifier, and the decrease is about 700 Pa. The β (bypass value) of the NU classifier is 3.6%, which is far lower than the β (bypass value) of 10% for a traditional classifier. The k (resolution of classification) of the NU classifier is as high as 0.6. The X₅₀ (cut size) of the NU classifier is small, which results in finer finished products under the same conditions. Compared with the LV classifier, the R_80μm of the NU classifier has been reduced by 2%-3% and the R_200μm has been reduced by 12%-33%, indicating that the NU classifier has a strong control ability in the particle size range above 200 μm.

3) Compared with the raw material mill system equipped with the LV classifier, the mill of the raw material grinding system equipped with the NU classifier saves 0.4 kWh/t-0.5 kWh/t of power consumption and  0.92 kWh/t under the condition that the product fineness is the same. The system equipped with the NU classifier saves at least 1 kWh/t if the fan has no efficiency problem meaning that this classifier provides a good technical solution for the technical upgrading of a grinding system.