JPH09141116A - Vibration estimation device of crusher and control device based on vibration estimation of crusher - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the damage to a plant caused by the generation of violent vibration and to realize the stable operation of the plant by calculating the distribution of the vibration estimate values in the state space of a crusher and performing the estimation of vibration exceeding a threshold value and the calculation of operation necessary for avoiding the generation of vibration by using the distribution. SOLUTION: When the vibration of a crusher 25 is estimated, a first operation means 100 inputs a feeder command signal 3, a classifying characteristic command signal 13 and a roller pressing force command signal 14 and calculates a matter-to-be- crushed holding quantity estimate value 101 and a matter-to-be-crushed particle size distribution estimate value 102 on the basis of a matter-to-be crushed grinding property estimate value 202 to calculate a crusher difference pressure estimate value 103. A second operation means 200 inputs the deviation 201 to the measured value 19 of the crusher difference pressure estimate value 103 to correct the matter-to-be-crushed grinding property estimate value 202. By a third operation means 300, vibration estimate arrangement is renewed on the basis of the respective data 14, 101, 102, 202 and a vibration measured value 30 and the distribution 301 of the vibration estimate value in a state space is calculated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vibration prediction device for a crusher and a control device based on the vibration prediction for a crusher, and determines the distribution of the magnitude of vibration generated in the current operating condition and its surrounding operating conditions. By predicting, the load response of the crusher and the product particle size are secured, while enabling the operation operation that avoids the occurrence of vibration exceeding the threshold value, and the crusher suitable for realizing stable operation of the plant. The present invention relates to a vibration prediction device and a control device based on the vibration prediction.

[0002]

2. Description of the Related Art FIG. 10 shows a conventional crusher and its control device. The raw material 1 to be crushed is fed into the crusher by the feeder 2 according to the command signal 3 from the hopper 4 provided on the upper part of the crusher, and drops onto the holding means (rotating disc) 5,
It is mixed with the recirculation flow 10 of coarse particles generated by the classifying means described later to form the pulverized material 6.

The object to be crushed 6 is moved from the central portion of the holding means 5 toward the outer peripheral direction by the centrifugal force generated by the rotation of the holding means 5, is crushed between the holding means 5 and the crushing roller 7, and is held. The carrier gas 8 is blown up from the outer periphery of the means 5. The object to be pulverized blown up is located at the upper part of the pulverizer and is swirled by the rotary classifier 9 in which the rotary blades are rotated by the prime mover, and the particles having a large particle size are subjected to centrifugal force to generate the recirculation flow 10. Form. On the other hand, the particles having a small particle diameter pass through the rotary classifier 9 to become a crusher product 11, which is conveyed to a customer.

Since the rotary classifier can obtain sensitive classification characteristics by controlling the number of rotations of the rotary blades, it has been generally adopted in recent years in applications where the product particle size is strictly required. Has been done. In addition to this, as a classification means, there is a fixed type classifier that gives a swirl by a fixed vane, but it is possible to adjust the classification characteristics by controlling the angle of the vane, and the function is a rotary type classification. You can think in the same way as a container.

Further, the supply amount of the carrier gas 8 is adjusted by an adjusting means (damper) 12 provided on the way of the carrier gas supply path. The control of the crusher according to the prior art is generally configured to adjust the classification characteristic command signal 13, the pressing force command signal 14, and the carrier gas supply amount command signal 15 according to the supply amount of the crushed raw material 1.

Here, in many cases, the feeder command signal 3 is adjusted so that the flow rate of the crusher product 11 matches the demand amount of the customer, and when the flow rate of the crusher product 11 can be measured online. , PI (proportional integral formula) adjustment of the deviation between the measured value and the demanded amount of the customer. However, in many cases, it is difficult to measure the flow rate of the crusher product 11 online, and therefore, the same purpose is achieved by using the state quantity of the demand destination having a causal relationship with the flow rate. For example, considering coal as a raw material to be crushed and a pulverized coal burning drum boiler as a demand destination, the generated steam pressure of the drum boiler has a direct causal relationship with the pulverized coal flow rate supplied to the burner from the pulverizer. PI of deviation from target value of pressure
The feeder command signal 3 may be adjusted by the adjustment.

The classification characteristic command signal 13 has a tendency to increase coarse particles of the crusher product 11 due to the function element 16 in the high load operation of the crusher, so that the supply amount of the raw material 1 to be crushed increases. It is given in the direction of giving a stronger turning force to improve the collection efficiency of coarse particles. Pressure command signal 1
4 is given by the function element 17 in the direction of increasing according to the supply amount of the raw material 1 to be ground, based on the fact that the crushing capacity of the crusher is improved with the increase of the pressing force.

The carrier gas supply amount command signal 15 has a function element 18 which provides a sufficient transportation capacity for conveying the crusher product 11 to be produced, and therefore the carrier gas supply amount command signal 15 increases in accordance with the supply amount of the object to be crushed 1. Given in. These days, it is difficult to discuss the setting method of the function elements 16, 17, and 18 uniformly as described above, such as demanding further refinement of the product at the time of low load operation due to the demand of the customer. Command signal 13,
It is common that 14 and 15 are functions of the supply amount of the object to be ground 1.

Generally, in a crusher, a vibration with a very large amplitude, which is several times to several tens times that of normal operation, may occur during low load operation. If such violent vibration continues, the crusher may be damaged, and means for detecting the occurrence or possibility of occurrence of the crusher in advance is required. The above-mentioned violent vibration is considered to be the self-excited vibration of the crushing roller 7 based on the studies so far, and the mechanism of occurrence is considered as follows.

As described above, the load is adjusted by adjusting the feed amount of the raw material to be pulverized by the feeder 2,
When the load is low, the amount of raw material to be crushed decreases,
Along with this, the crushed material 6 on the holding means 5 decreases.
Therefore, the crushing capacity of the crushing roller 7 has a margin, the layer of the crushed object between the crushing roller 7 and the holding means 5 becomes relatively thin, and the particle size distribution of the crushed object is small. Move in the direction. For this reason, the friction coefficient in the layer of the crushed material is lowered, and slippage occurs in the layer of the crushed object. As a result, the crushing roller 7 vibrates greatly and causes self-excited vibration.

It is known that slippage in the coal seam tends to occur as the pressing force of the crushing roller 7 against the holding means 5 increases. For this reason,
When the load is low, it is conceivable to reduce the roller pressure in advance in order to prevent the occurrence of severe vibration. However, the reduction of the pressure applied to the roller causes a reduction in the pulverization ability of the pulverization roller, and the load response of the pulverizer (the ability to follow the production amount of the pulverizer product with respect to the change in the feed amount of the pulverizing material) is deteriorated. Therefore, as mentioned above, in cases where the load response of the crusher influences the load response of the entire plant, such as when the demand destination is a drum boiler, as described above, there is a demand to increase the roller pressure as much as possible. There is.

[0012] In the prior art, a trial run with an actual machine is carried out,
The function elements 16, 17 and 18 were finally adjusted so that the particle size distribution of the crusher product and the load responsiveness satisfy the requirements in the normal operation range, and the operation is such that vibration exceeding the threshold does not occur. Further, the detector 19 detects vibration of the crusher and the detector 2
The differential pressure of the crusher is 0, the temperature of the carrier gas supplied to the crusher by the detector 21 and the temperature of the carrier gas at the outlet of the crusher are measured by the detector 22, and they are used for abnormality monitoring inside the crusher. It was

[0013]

In the prior art, as described above, the function element 1 is obtained by obtaining the operation condition (combination of operation amounts) at which the vibration exceeding the threshold does not occur during the test operation of the actual machine.
6, 17, and 18 were set. Therefore, if the structure of the crusher and the properties of the raw material to be crushed (crushability, etc.) are the same as the conditions at the time of test operation, vibration exceeding the threshold value will not occur even in the conventional technique.

However, for example, when the properties of the material to be pulverized to be supplied are changed, the changes in the properties cause
The range of operation conditions in which vibration exceeding the threshold value occurs changes, and the operation condition that was safe at the time of test operation may be included in the region in which vibration exceeding the threshold value occurs. In such a case, in the related art, since the setting of the function element is fixed, it may not be possible to cope with this variation and vibration exceeding the threshold value may occur. In particular, in the case of a crusher that supplies pulverized coal to a coal-fired boiler, it is known that the coal, which is a raw material for pulverization, has greatly different properties depending on its brand (mainly determined by the place of origin).

On the other hand, in the prior art, when there are a plurality of coal brands used in the plant, a trial run is carried out for each brand and the function generators 16, 17, 18 are carried out.
Each brand has its own dedicated setting, and it is used by switching to the setting that corresponds to the coal currently supplied to this crusher. However, even in this case, the following problems remain.

First, in the prior art, it is difficult to deal with variations in properties within the same brand of pulverized raw material. In general, naturally occurring solid raw materials such as coal have relatively large variations in properties even with the same brand. For example, even if coals of the same brand are used, coal crushability (HG
The variation of I) may be 20%. When the crushability changes, the crushing ability required of the crushing roller 7 also changes relatively, and along with this, the amount of crushed material held and the particle size distribution of the crushed object also change. It is also known that the easiness of slippage in the coal seam has a strong correlation with the pulverizability. Therefore, it cannot be said that the conventional fixed measures for each coal brand are sufficient.

Secondly, it is difficult for the prior art to deal with the case where a plurality of brands of crushed raw materials are mixed and supplied. When a mixture of crushed materials of multiple brands is supplied to the crusher, the combination of the brands of crushed raw materials to be mixed and the respective mixing ratios will be huge, and it is possible to avoid the occurrence of vibrations that exceed the threshold under all conditions. It is difficult to set the function generating elements 16, 17, and 18 to do so.

Thirdly, it is difficult for the prior art to deal with a transient state. The conventional technique seeks a combination of manipulated variables that achieves good driving by a test. In the steady state, the internal state quantity of the crusher is uniquely determined by the combination of manipulated variables (assuming constant coal properties). However, in the transient state, the internal state may be different even if the manipulated variable is the same. For example, the internal state is completely different in the process of increasing the pulverized raw material input amount and the process of decreasing it. For this reason, when the load changes rapidly, the internal state of the crusher is likely to deviate from the test range during the test run, and vibration exceeding the threshold value may occur.

The problems of the prior art are eventually summarized as follows. (1) There is no means for knowing online the amount of internal state of the crusher (amount of crushed object, particle size distribution of crushed object, etc.) that is closely related to vibration. (2) There is no means for knowing the properties of the raw material to be pulverized supplied to the pulverizer online. (3) A means is provided for knowing the direction and distance of a violent vibration generation region as seen from the current state of the crusher in the crusher state space determined by the crusher internal state quantity and the property of the crushed material. Therefore, the direction and size of the correction operation necessary and sufficient for avoiding vibration cannot be obtained.

The present invention provides a vibration prediction device for a crusher and a control device based on the vibration prediction of the crusher, which solves the above problems and realizes stable operation especially in a plant where the properties of the material to be crushed frequently change. It was made to do.

[0021]

The invention claimed in this application to achieve the above object is as follows. (1) Holding means for crushing an object to be crushed that rotates in a horizontal plane in a crusher; a plurality of crushing rollers that rotate while being pressed by the holding means to constitute crushing means for the object to be crushed; The coarse particles of the material to be crushed that has passed through the means are recirculated to the holding means and the fine particles are taken out as a crusher product, a means for adjusting the classification characteristics of the classifying means, and a crusher differential pressure In the vibration prediction device of the crusher that has a means for measuring, the measured value of the amount of operation to the crusher at the present time, the estimated value of the crusher state amount at the time of the previous calculation, and the previous calculation of the property of the crushed raw material held by the crusher Based on the estimated value at the time point, a first calculation means for calculating the current estimated value of the pulverizer state quantity, one or more estimated values of the crusher state quantity calculated by the first arithmetic means, and measurement of the state quantity Grind by using the matching status with the value A second calculating means for estimating the property of the crushed object held at the time of the calculation, a part or all of the state amount of the crusher calculated by the first calculating means, and the operation amount of the crusher. And a third arithmetic means for calculating the distribution of the predicted value of the vibration magnitude of the crusher in a certain range centered on the current coordinates in the multidimensional space formed by part or all Vibration prediction device for crushers.

(2) Holding means for rotating the crushed object in the crusher, and a plurality of crushing rollers that rotate while being pressed by the holding means by the pressurizing means and constitute crushing means for the crushed object together with the holding means. And a classification means for recirculating the coarse particles of the material to be crushed that has passed through the crushing means to the holding means and discharging the fine particles as a crusher product,
Based on the crushed raw material supply command signal to the crusher, means for adjusting the amount of raw material supplied to the crusher, means for adjusting the classification characteristics of the classifying means, means for adjusting the pressing force of the crushing means and the crusher In a vibration predicting device of a crusher equipped with a carrier gas supply amount adjusting means, a current supply amount of a material to be crushed to a crusher, a roller pressing force for pressing a crushing roller against a holding means, and a classification characteristic of a classifying means. Of the pulverizer at the present time by using the classification characteristic command signal that determines the value, the carrier gas supply amount signal 15, the estimated value of the pulverized object possessed by the pulverizer at the time of the previous calculation, and the estimated pulverizability of the pulverized object. First computing means for calculating an estimated value of the crushed material holding amount, an estimated value of the crushed object particle size distribution on the holding means, and an estimated value of the crusher differential pressure of the crusher based on the estimated crushed object holding amount value And the estimated value of the crusher differential pressure calculated by the first calculation means and Based on the deviation from the measured value of the crusher differential pressure, the second calculation means for calculating the crushability of the crushed object at the present time, and the current estimated value of the crushed object held by the crusher calculated by the first calculating means. , The crushing constituted by these four variables, based on the particle size distribution estimated value of the object to be crushed on the holding means, the pulverizability estimated value of the object to be crushed calculated by the second operation, and the roller pressing force at the present time. In the state space of the machine, a third arithmetic means for calculating the distribution of the predicted value of the magnitude of the crusher vibration in the crusher state space within a certain range centering on the current crusher state is provided. Vibration prediction device for crushers.

(3) Holding means for the object to be crushed which horizontally rotates in the crusher, and a crushing roller which rotates while being pressed by the holding means by the pressing means and constitutes a crushing means for the object to be crushed together with the holding means. Classifying means for recirculating the coarse particles of the material to be crushed that has passed through the crushing means to the holding means and discharging fine particles to the outside of the crusher, means for adjusting the classification characteristics of the classifying means, and a crusher In a vibration predicting device for a crusher having a means for measuring a differential pressure, a dynamic characteristic model of the crusher is used, and a raw material supply amount to be crushed to the crusher at the present time,
A roller pressing force for pressing the crushing roller against the holding means,
Using the classification characteristic command signal, the estimated value of the crushed object holding amount at the time of the previous calculation, and the crushability estimated value of the crushed object, the current estimated value of the crushed object held amount in the crusher and the crushed object on the holding means A first calculation means for calculating the particle size distribution estimated value and further for calculating the pulverizer differential pressure estimated value based on the pulverized material holding amount estimated value; the pulverizer differential pressure estimated value and the differential pressure actual measurement A second calculation means for estimating the pulverizability of the pulverized material at the present time from the deviation from the value, the roller pressing force and the pulverized material holding amount estimated value, the pulverized material particle size distribution estimated value on the holding means, Based on the estimated crushability of the crushed object and the history of past vibrations that are successively learned, the predicted value of the magnitude of vibration in the crusher state quantity space within a certain range centered on the current crusher state quantity A vibration predicting apparatus for a crusher, comprising: a third calculating means for calculating a distribution. (4) Means for calculating the roller pressing force correction signal and the classification characteristic correction signal based on the vibration predicted value distribution calculated by the crusher vibration prediction device according to any one of (1) to (3), and the above calculation. A controller based on vibration prediction of a crusher, characterized in that it is provided with means for adjusting the roller pressure of the crusher and the classification characteristics of the classifying means based on the correction signal.

According to the present invention, the amount of internal state of the crusher (for example, the amount of crushed material retained, which is considered to have a strong correlation with the vibration generation,
Assuming a multidimensional state space based on the crushed object particle size distribution), the crusher operation amount (for example, crushing pressure roller pressure), and the crushed object property (for example, crushed object crushability), the multidimensional state is assumed. It provides a means for learning and referring to the distribution of the magnitude (amplitude) of the vibration generated in the actual crusher machine in the space online.

However, in the actual crusher, it is difficult to measure most of the internal state quantities online.
Therefore, in the present invention, the required crusher internal state quantity is calculated as a dynamic characteristic model of the crusher. In calculating the internal state quantity of the crusher using the dynamic characteristic model, the undetermined quantity (hereinafter referred to as unknown parameter) such as the property of the crushed object (eg, crushability)
Exists, which causes a calculation error. In order to deal with this, in the present invention, an amount that can be measured in an actual crusher machine and has a strong correlation with a typical crusher internal state (for example, crusher differential pressure, crusher power) is estimated in a dynamic characteristic model. , The unknown parameters are sequentially identified so that there is no deviation between the estimated value and the actual measured value.

Hereinafter, the crushed object 6 is defined as the internal state quantity of the crusher.
When the amount of possession and the particle size distribution of the object to be pulverized 6 can be measured online, the differential pressure of the pulverizer which has a strong correlation with the amount of the object to be pulverized 6 is adopted, and the object 6 pulverizability of the object to be pulverized is used as an unknown parameter. Taking the example as an example, a specific calculation procedure will be described. In the first computing means, the operation signal of the target crusher actual machine and the crushability estimated value of the crushed object fixed at the time of the previous calculation are given to the kinetic model of the crusher as a crushability assumed value, and the crushed object is crushed. Estimated values of the amount of internal state of the crusher, such as the amount of the object 6 held and the particle size distribution of the object 6 to be crushed are obtained. Further, the crusher differential pressure estimated value is calculated based on the estimated amount of the crushed object 6 held.

The second computing means updates the pulverizability estimation value of the object to be pulverized 6 based on the deviation of the estimated pulverizer differential pressure value calculated by the first computing means from the actual measurement value. Generally, when the estimated value of the crusher differential pressure is larger than the actual measured value (equivalent to estimating that the crushed object 6 possession amount is larger than the actual machine), the pulverizability is too small (cracking is smaller than actual). It is judged that it is difficult to evaluate, and the estimated pulverizability of the pulverized material 6 is corrected in the increasing direction, and the same is true in the opposite case.

The operating forms of the first calculating means and the second calculating means are such that, at each calculation point, convergence calculation is performed until the deviation between the estimated value and the measured value of the crusher differential pressure is within a specified value. The convergence calculation may be repeated once or may be terminated once or a prescribed number of times. These are determined in consideration of the calculation speed of the computer used and the time interval at each calculation time point. The third calculation means has two functions. The first function is a vibration distribution updating function, and the current crusher vibration measurement value is the operation amount of the roller pressing force, the crushed object 6 estimated amount of the crushed object 6 calculated by the first calculating means, and the crushed object. It is configured with the amount having a strong correlation with the vibration of the crusher, such as the estimated value of 6 particle size distribution and the estimated value of 6 crushability of the object to be crushed. Record as the actual value at the current coordinates in the multidimensional empty state space. By sequentially performing this operation during the operation of the crusher, a vibration distribution in the multidimensional state space as shown in FIG. 2 can be obtained after a sufficient time has elapsed. The second function is a reference function of the vibration distribution, and refers to the vibration distribution in the crusher in a multidimensional state space within a certain range (hatched range in FIG. 3) centering on the condition, It is output as a vibration prediction distribution within the range.

In the third means, the range for obtaining the distribution of the predicted vibration value increases the computer load if it is obtained over a wide range, but if the range is narrowed, there is no time margin to reach the vibration generation area, and sufficient avoidance is possible. It may not be possible to operate, so set it in consideration of both. With the above operation, for the current state of the crusher, the distribution of the magnitude of vibration predicted to occur at the current position in the multidimensional state space and the surrounding states can be obtained. By referring to the position and the distribution of vibration, when the state of the crusher approaches a vibration area exceeding a threshold value, an operator is notified by an alarm or the like to prompt an avoidance operation, or to avoid an area where it occurs. It is possible to calculate the correction operation amount necessary for

[0030]

FIG. 1 shows an embodiment of a vibration prediction method according to the present invention. The same components as those in FIG. 10 are designated by the same reference numerals and the description thereof will be omitted. First computing means 10
0 is the same operation amount as that given to the actual machine at present (feeder command signal 3, classification characteristic command signal 13, roller pressing force command signal 14, carrier gas supply amount command signal 1).
5) as an input, and based on the assumed value of the pulverizability of the pulverized material 6 (estimated pulverizability of the pulverized material at the time of the previous calculation 202), use the built-in dynamic characteristic model to estimate the amount of pulverized material currently held Value 101, crushed object particle size distribution estimated value 102,
The crusher differential pressure estimated value 103 is obtained from the stored coal amount estimated value.

As the dynamic characteristic model, it is possible to use a statistical model in which relations of various quantities of interest are arranged by actual measurement data, but various processes of the crusher (crushing, classification, mixing, etc.) are faithfully simulated. Since this physical model has high accuracy under the widest range of conditions, the physical model is adopted in this embodiment. Details of the physical model of the crusher applicable to this embodiment will be described later.

The second computing means 200 receives the deviation 201 of the estimated pulverizer differential pressure value 103 with respect to the measured value 19 as an input, and assumes the pulverizability assumed value of the crushed object 6 (previously estimated crushability value of the crushed object 202). Is corrected by the PI function to calculate the crushed material crushability estimated value 202 at the present time. As described above, it is known that the crusher differential pressure has a strong correlation with the crushed object 6 holding amount of the crusher, and monotonically increases with respect to the crushed object 6 holding amount under the same operation amount. ing. Therefore, when the deviation 201 is a positive value, that is, when the estimated value of the crusher differential pressure is larger than the actually measured value, it indicates that the amount of the object to be crushed 6 is estimated more than it actually is. On the other hand, if the manipulated variables are the same, it is considered that the change in the amount of the pulverized material 6 held is due to the change in the pulverizability of the pulverized material 6. For example, as grindability increases (becomes more fragile), the proportion of particles that disintegrate to the product particle size in a single pass through the grinding mechanism increases, thus reducing the coarse particles recycled by the classification mechanism. For this reason, the amount of the pulverized material 6 formed by the recirculation portion and the new input portion is reduced. Based on the above knowledge, the correction of the crushability estimation value 202 of the crushed object 6 in the second calculation means is performed in the increasing direction if the deviation 201 is a positive value, and in the decreasing direction if the deviation 201 is a negative value. (Proportional-integral type) The controller may be set.

The third calculation means 300 measures the roller pressing force command signal 14, the crushed object holding amount estimation value 101, the crushed object particle size distribution estimated value 102, the crushed object crushability estimated value 202, and the crusher vibration measurement. While inputting the value 30, the vibration prediction array described later is updated, and the past record of vibration occurrence recorded in the array is referred to, and within a certain range of the crusher state space centering on the current crusher state The distribution 301 of the predicted vibration value (hereinafter, referred to as predicted vibration distribution value) is calculated. FIG. 3 is an example of vibration distribution prediction in a two-dimensional state space, and the third calculation means is a range (hatching of hatching) necessary and sufficient for vibration avoidance control centered on the current state indicated by the black dots in FIG. The vibration distribution of (range) is output.

The input / output relationship similar to that of the vibration prediction array can be realized by sequentially learning the hierarchical neural network, and the selection of both is made in consideration of the target property and the subtractor load. The output of the vibration prediction array has a staircase distribution as shown in FIG. 4 due to its structure, but a continuous distribution as shown in FIG. 2 is obtained by performing appropriate interpolation, function approximation, etc., if necessary. It is possible. In addition, FIG.
FIG. 4 shows the case where only the crushed object holding amount and the roller pressing force are selected as the state quantities used for vibration prediction in the present embodiment, but other state quantities are selected from the multidimensional state space (n> 2). It can also be regarded as a diagram of a fixed cross section.

Next, the dynamic characteristic model of the crusher and the vibration prediction array will be described. 1. In this example, the dynamic characteristic model of the crusher is involved in the research of the inventor himself, and the method of simulating the particle size distribution with only four variables and having high accuracy and low calculation amount (Measurement Automatic Control Society China Branch Academic Lecture Meeting; Flat 3/12
/ 13 will be adopted by the inventor and will be described in detail below. 1.1 Symbols / abbreviations / terms 1.1.1 Subscripts The following subscripts indicate specific parts, mechanisms, functions, and time points.

B: mixing mechanism ib: crusher inlet ij: j-th classifying mechanism inlet ip: crushing mechanism inlet j: j-th classifying mechanism ob: mixing mechanism outlet op: crushing mechanism outlet rj: j-th number Circulating flow by classifying mechanism s: Grinding distribution constant oj: Throughflow by j-th classifying mechanism 1.1.2 Symbol β _k : k-th cumulant μ: average ξ: logarithmic particle size (nominal value) σ: standard deviation b ( ξ): Grinding rate constant c (ξ): Partial classification efficiency f (ξ): Generally probability density function g (ξ): Particle size distribution probability density function g (ξ | η): Conditional probability density function Ξ: Logarithm of particles Particle size (random variable) s (ξ): Milling distribution constant v _k (λ, ρ): kth moment normalized to (λ, ρ) E (・): Expected value E (・ | ・): Conditional Expected value Var (•): Variance Pr (•): Probability Q: Weight Flow rate Θ: Indicator (random variable) that a particle passes through (Θ = 1) or collects (Θ = 0) in the classification process. 1.1.3 Words Random variable: Because of “fluctuation” or lack of information, Requires probabilistic evaluation of variable values (values cannot be known deterministically; noise,
Variables are powder flow rate and coal properties). Therefore, the probability Pr {x <X ≦ x + Δx} (probability that the value of the random variable X is in the vicinity of x) is given by f _X (X) Δx (f _X (X): probability density function of x) Calculate statistics.

Cumulant: When the logarithm of the characteristic function (Fourier transform of the probability density function) is Laurent expanded, the kth expansion coefficient is called a kth cumulant. In terms of application, it has convenient properties such as "interchangeable with moment" and "superimposed integral (change in particle size distribution in the crushing mechanism) results in addition". Skewness: Translated as skewness and indicates the degree of left-right symmetry of the distribution.

Excess: Translated as kurtosis, the center of the distribution,
Indicates the proportion of the skirt (= 3 for normal distribution). Conditional probability density: Considering the causal relationship between two events
The probability density of the other when the information of one is obtained is shown. Ingredient
Physically, the particle size is η <Θ at the entrance of the crushing mechanism._ip≦ η + dη grain
After the child is crushed ξ <Θ_ipThe probability that ≦ ξ + Δξ is Θ_opΘ
_ipProbability density function for (conditional probability d
ensity function of Θ_op given by Θ _ip) With g
_{op | ip}It is expressed as (ξ | η) dξ. 1.2 Phenomena inside the crusher 1.2.1 Indication of particle size distribution Mass of particle size ξ or less among particles passing through the cross section in a short time
The particle size distribution can be defined by the ratio, and its density function is g
It is written as (ξ), and a subscript indicating the place is appropriately added. Sa
The sampled static mass particle size distribution density f (ξ)
The relationship is expressed by the following equation using the mass flow rate Q.

[0039]

(Equation 1)

1.2.2 Grinding Mechanism When the subscripts ip and op are given to the quantities before and after the crushing, the particle size distribution has the following relationship.

[0041]

(Equation 2)

[0042] Here, the particle size probability density conditionally take on a logarithmic axis ξ g _op |.. _Ip is, L Austin, et al. Elucidated the grinding distribution constant (Power Technology, Vol 29, pp.263-275 (1981 )
, Ibid, Vol. 33, pp. 113-125 (1982), ibid, Vol. 33, p.
p.127-134 (1982)), and this is designated as s.

[0043]

## _EQU3 ## g _{OP | ip} (ξ | η) = s (ξ−η) (3) It is known that this pulverization distribution constant s depends on the pulverizability of particles passing through the pulverization mechanism. As for the mass flow rate, the following equation is obtained assuming that there is no accumulation in the grinding mechanism.

[0044]

E {Q _op } = E {Q _ip } (4) 1.2.3 Classification Mechanism For the j-th classification mechanism, each “particle passage” is an independent event, and Θj can be used as an indicator. For example, the classification efficiency c _j (ξ) (Mari et al.
Powder Engineering Journal, Vol.25, pp. 480-436 (1988)) and the following relationships.

[0045]

Equation 5] _{Pr {Θ j = 0 | (} ξ, ξ + dξ)} = c j (ξ) (5) Pr {Θ j = 1 | (ξ, ξ + dξ)} = 1-c j (ξ) (6) When the subscripts ij, rj, and oj are given to the quantities related to the classifier inlet, the circulation flow, and the passing flow, respectively, the following equation representing the particle size distribution density is obtained by the Bayes theorem.

[0046]

(Equation 6)

It is The flow rate around the classification mechanism is calculated as follows.

[0048]

E {Q _rj } = E {Θ _j Q _ij } = _rj E {Q _ij } (10) E {Q _oj } = E {(1-Θ _j ) Q _ij } = (1- _rj ) E {Q _ij } (11) 1.2.4 Mixing mechanism The circulating flow from the classifying mechanism (j = 0, ..., N) and the raw material to be crushed (subscript ib) are mixed to form an outflow (subscript ob). Think about the mechanism. Here, the following relationship is assumed between the objects to be ground G _b and Q _ob held by the mixing mechanism.

[0049]

E {Q _ob } = E {P} E {G _b } (12) P is independent of particle size, and in order to justify this assumption, a hypothesis exists between the mixing mechanism and the subsequent grinding mechanism. Considering the ξ-dependent crushing rate constant elucidated by L. Austin et al., Mentioned above, by providing a general classification mechanism (j = 0).

Here, paying attention to the above equations (1) and (12), considering that the particle size does not change due to mixing,
A mass balance formula of particles belonging to (ξ, ξ + dξ) is obtained.

[0051]

(Equation 9)

1.3 Mathematical Description of Model 1.3.1 Parameterization of Distribution Density Consider a moment normalized (affine transformation) by λ and ρ when Ξ follows distribution density g (ξ).

[0053]

(Equation 10)

At this time, the cumulant β _k (λ, ρ) is correspondingly obtained. In this model, the distribution density is organized by the following parameters.

[0055]

(11) μ = v ₁ (0,1), σ = [v ₂ (0,1)] ^1/2 (15) Skewness: β ₃ (μ, σ) = v ₃ (μ, σ) (16 ) Excess: β ₄ (μ, σ) = v ₄ (μ, σ) ⁻³ (17) From these, the edge-worth expansion coefficient α _k can be uniquely obtained, and the distribution density can be specifically displayed.

[0056]

(Equation 12)

Where p (ξ; μ, σ) is a Gaussian distribution,
h _k is a Hermitian polynomial of degree k. 1.3.2 Grinding mechanism Substituting equation (3) above into equation (2) gives a convolution integral, which results in the sum of cumulants, and obtains the following.

[0058]

(13) v _1op (μ _ip, σ _ip ) = v _1s (μ _ip, σ _ip ) (19) v _2op (μ _ip, σ _ip ) = 1 + v _2s (μ _ip, σ _ip ) (20) v _3op (μ _ip, σ _ip ) = v _3ip (μ _ip, σ _ip ) + v _3s (μ _ip , σ _ip ) (21) v _4op (μ _ip, σ _ip ) = v _4ip (μ _ip, σ _ip ) + v _4s (μ _ip , σ _ip ) + 6v _2s (μ _ip , σ _ip ) (22) Here, the subscript s indicates the pulverization distribution constant s, and the others indicate each particle size distribution density g. Furthermore, μ _op , σ _op , β ₃ (μ _op , σ
_op ), β ₄ (μ _op , σ _op ) are the above-mentioned equations (15) to (1)
It can be calculated using the equation 7) and the following equation.

[0059]

[Equation 14]

1.3.3 Classification Mechanism c _j (ξ) can be approximated by using appropriate τ _mj , λ _mj and ρ _mj .

[0061]

(Equation 15) g _ij (ξ) is in the form of the expression (18), and various quantities of the circulating flow are specifically obtained from the expressions (7) and (10).

[0062]

(Equation 16)

Here are the following functions:

[0064]

[Equation 17]

Further, α _Orjmk can be obtained by applying the Hermitian polynomial addition theorem to the following equation and arranging the coefficients.

[0066]

(Equation 18)

The above equation (25) is a weighted mixture of distribution densities, where α _{Orjm k} to v _k (λ _mrj,
ρ _mrj ) is uniquely obtained, and v _k of the same λ and ρ can be weighted and added, so that the expression (23), the expression (1
Using equations (5) to (17), μ _rj , σ _rj, β ₃ (μ
_rj, σ _rj ), β ₄ (μ _rj , σ _rj ) can be calculated. The same argument applies to the passing flow with the subscript oj. 1.3.4 Mixing mechanism When normalized by properly selected λ _{b and} ρ _b , the above (1)
From equation 3), a differential equation for v _kob is obtained.

[0068]

[Equation 19]

By applying the equations (15) to (17), in general, μ, σ, β ₃ (μ, σ), β ₄ (μ,
Since σ) and v _k (λ, σ) can be mutually transformed, 1.
(30) can be solved by substituting the conclusions of 3.2 and 1.3.3. At this time, stable numerical calculation became possible by adopting the Pade approximation. Therefore, as an initial condition, E
By giving {Q _ob }, v _kob (λ _b, ρ _b ), giving the same amount of operation as the actual crusher, and solving the above equations sequentially,
The amount E {G _b } of the object to be crushed 6 and its particle size distribution v _kb (λ _ib, ρ _ib ) can be obtained. 2. Vibration prediction array Obtaining the vibration distribution in the target n-dimensional state space, which is the object of the present invention, is nothing but obtaining the following n-input function.

[0070]

Equation 20] _{h (x 1, x 2,} ........., x n) (31) Here, h (x): n input function x ₁ representing the vibration distribution, ......, x _{n: n-dimensional} state space It is the coordinate within. As a means for obtaining the shape of the function h in a computer, the shape of the function h is approximated by a known mathematical function (logarithmic function, parabola, etc.) based on some assumption, and the parameters of the approximation function are minimized so that the approximation error is minimized. The method of adjusting is often used. However, such a method
It is effective only when the shape of the function is expected to some extent, and is effective only when the number of inputs is small, and it is difficult to apply it to an object, such as the object of the present invention, which has many inputs and whose shape is difficult to predict.

Therefore, in the present invention, the n-dimensional state space is
Is divided into meshes of appropriate size, and
Simulates vibration distribution by recording and referencing generated vibration
The vibration prediction array is adopted. This vibration prediction array is
Physically, it is actually implemented as an array secured in the memory on the computer.
Will be revealed. In the simplest configuration, the target n-dimensional state
As shown in Fig. 5 (two-dimensional example), the space should have a uniform size.
In the area included in each mesh,
N dimensions (n is the number of inputs) that stores the predicted value of the generated vibration
Array Y (y₁, Y _Two, ……, y_n) Is used.

The vibration predicted value is updated every calculation cycle of the third calculating means. Further, there are various setting methods such as the maximum value in the past and the average value in the past for the calculation means of the vibration predicted value, but here, the maximum value of the vibration that has occurred in a certain recent period is used as the vibration predicted value. adopt. For example, the maximum value of vibrations generated in the past one month in the area included in the target mesh is used as the vibration predicted value of the mesh.

[0073]

Y (y ₁ , y ₂ , ..., Y _n ) = MAX {H (y ₁ , y ₂ , ..., Y _n )} (32) where MAX {K} is the set K. Is a function that outputs the maximum value included in H (y ₁ , y ₂ , ..., Y _n ).
Is a set of crusher vibration values measured in the corresponding region over a recent period of time.

This emphasizes the evaluation on the safety side and considers that the vibration distribution moves due to abrasion of the crushing mechanism (holding means 5 and crushing roller 7) due to continuous operation. . As the operation amount of the target crusher changes, the internal state changes every moment and moves in the multidimensional state space. Along with this, the element of the vibration prediction array corresponding to the coordinate is updated, and the vibration distribution in the multidimensional state space as shown in FIG. 4 is obtained.

When the operation of the present estimation device is started, it is sufficient to roughly give an empirically known vibration region as an initial value of the vibration prediction array. After that, the vibration prediction array is updated sequentially, so after a sufficient period of operation,
An array is formed that reflects the characteristics of the plant. FIG.
Is an embodiment when the vibration prediction method according to the present invention is applied to a control device.

The fourth computing means 400 calculates, based on the vibration distribution predicted value 301,
The distance and direction viewed from the current state of the crusher are calculated, and the roller pressing force correction signal 401 and the classification characteristic correction command signal 402 are obtained by fuzzy reasoning based on the distance and direction. Generally, the roller pressing force is reduced as it approaches the area where vibration exceeding the threshold value occurs.
Or loosen the classification characteristics (increasing the 50% passing particle size)
It may be corrected in the direction of

FIG. 7 shows the correspondence when the vibration occurs according to the prior art in FIG. In the prior art, when trying to shift from a state represented by a black dot to a state represented by a white dot, there is no means to know that even if the transition path includes a vibration generation area exceeding a threshold value. Cannot be avoided. For this reason, after detecting the occurrence of severe vibration by the detector 19, a vibration suppressing operation such as sharply and drastically reducing the roller pressing force is performed by manual intervention or automatic operation. FIG. 8 shows the behavior of the pulverizer product flow rate and the same particle size distribution obtained by simulation when such an avoiding operation is performed, but the effect of the avoiding operation is clearly shown.

In the present embodiment, as shown in FIG. 9, referring to the vibration distribution around the current state, when a vibration area exceeding the threshold value is approached, a correction operation is performed so as to move away from the vibration area. As shown in FIG. 9, it is possible to perform a driving operation that detours a region where vibration exceeding a threshold value occurs to the minimum limit. In this case, it is not necessary to drastically change the operation amount, and the above-mentioned adverse effect does not occur. Further, for example, it is possible to suppress the vibration avoiding operation that affects the characteristics of the crusher, such as reduction of the pressure applied to the roller, to the minimum necessary, so that the characteristics of the crusher can be kept good.

[0079]

EFFECTS OF THE INVENTION According to the present invention, even in a plant where the properties of the material to be ground frequently change, the distribution of the predicted vibration value in the state space of the grinding machine can be provided. By using this distribution, it is possible to predict vibrations that exceed the threshold and to calculate the operation required to avoid the occurrence of the vibrations. It has the effect of avoiding it in advance and realizing stable operation of the plant.

[Brief description of the drawings]

FIG. 1 is a diagram showing a first embodiment of the present invention.

FIG. 2 is a diagram showing a schematic diagram of vibration distribution in a multidimensional state space.

FIG. 3 is an explanatory diagram of prediction of vibration distribution in a certain range.

FIG. 4 is a diagram showing an example of a vibration prediction distribution given by a vibration prediction array.

FIG. 5 is an explanatory diagram of a vibration prediction array.

FIG. 6 is an explanatory view of an embodiment in which the vibration prediction result according to the present invention is applied to a control device for a crusher.

FIG. 7 is an explanatory diagram of a vibration avoidance operation according to a conventional technique.

FIG. 8 is an explanatory diagram showing an example of the behavior of the product flow rate and particle size distribution of the crusher during the vibration avoidance operation according to the conventional technique.

FIG. 9 is an explanatory diagram of a vibration avoidance operation according to the present invention.

FIG. 10 is an explanatory view of a control device for a crusher according to a conventional technique.

[Explanation of symbols]

1 ... Material to be ground, 2 ... Feeder, 3 ... Feeder command signal, 4 ... Hopper, 5 ... Holding means, 6 ... Object to be ground, 7 ... Grinding roller, 8 ... Carrier gas, 9 ... Rotary classifier, 10 ... Re Circulating flow, 11 ... Grinder product, 12 ... Damper, 13 ... Classification characteristic command signal, 14 ... Pressurization command signal, 15 ... Carrier gas supply command signal, 16, 17, 18 ... Function element, 19 ...
(Vibration) detector, 20 ... (Differential pressure) detector, 21 ... (Carrier gas inlet temperature) detector, 22 ... (Carrier gas outlet temperature) detector, 25 ... Grinder, 100 ... First computing means, 101
... Estimated amount of crushed object held, 102 ... Estimated value of particle size distribution of crushed object, 103 ... Estimated value of crusher differential pressure, 200 ... Second computing means, 201 ... Estimated deviation of crusher differential pressure, 202 ... crushed object Grindability estimated value, 300 ... Third calculation means, 301 ... Vibration predicted value distribution.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Mikio Yamanaka 6-9 Takaracho, Kure City, Hiroshima Prefecture Babcock Hitachi Kure Factory

Claims (4)

[Claims] 1. A holding means for holding an object to be ground which rotates in a horizontal plane in a grinder, and a plurality of grinding rollers which rotate while being pressed by the holding means to constitute a means for grinding the object to be ground together with the holding means. Classifying means for recirculating the coarse particles of the material to be crushed that has passed through the crushing means to the holding means and for taking out fine particles as a crusher product, a means for adjusting the classification characteristics of the classifying means, and a crusher difference In the vibration predicting device of the crusher having a means for measuring the pressure, the operation amount measurement value to the crusher at the present time, the crusher state amount estimated value at the time of the previous calculation, and the crushed raw material property held by the crusher. Based on the estimated value at the time of the previous calculation, first calculating means for calculating the current estimated value of the crusher state amount, one or more estimated values of the crusher state amount calculated by the first calculating means, and the state amount Using the matching situation with the measured value of A second calculation means for estimating the property of the crushed object held by the crusher at the time of the calculation, a part or all of the state quantity of the crusher calculated by the first calculation means, and And a third calculation means for calculating the distribution of the predicted value of the magnitude of the crusher vibration within a certain range centered on the coordinates at the present time in the multidimensional space formed by a part or all of the manipulated variables. A vibration prediction device for a crusher, which is characterized in that 2. A holding means for rotating an object to be ground which rotates in a crusher, and a rotating means while being pressed by the holding means by a pressurizing means,
A plurality of crushing rollers which constitute a crushing means for crushing the object to be crushed, and coarse particles of the crushed object which have passed through the crushing means are recirculated to the holding means, and fine particles are discharged as a crusher product. And a means for adjusting the amount of raw material supplied to the crusher based on the crushed raw material supply command signal to the crusher, a means for adjusting the classification characteristics of the classifying means, and a crushing means. In a vibration predicting device of a crusher equipped with means for adjusting pressure and means for adjusting the amount of carrier gas supplied to the crusher, the current supply amount of the material to be crushed to the crusher and the crushing roller are pressed against the holding means. Roller pressing force, classification characteristic command signal that determines the classification characteristics of the classification means, carrier gas supply amount signal, estimated value of crushed material possessed by the crusher at the time of previous calculation, and estimated crushability of the crushed object Using the current grinding First calculation for calculating the crushed object holding amount estimated value, the crushed object particle size distribution estimated value on the holding means, and the crusher differential pressure estimated value of the crusher based on the crushed object holding amount estimated value Means for calculating the pulverizability of the object to be pulverized at the present time based on the deviation between the estimated value of the pulverizer differential pressure calculated by the first arithmetic means and the actual measured value of the pulverizer differential pressure; The current estimated value of the crushed object possessed by the crusher calculated by one calculating means, the estimated value of the particle size distribution of the crushed object on the holding means, and the estimated pulverizability of the crushed object calculated by the second operation And the pressure of the roller at the present time, in the state space of the crusher constituted by these four variables, the magnitude of the crusher vibration in the crusher state space within a certain range centered on the current crusher state And a third calculation means for calculating the distribution of predicted values of Prediction device. 3. A holding means for the object to be crushed which horizontally rotates in the crusher, a crushing roller which rotates while being pressed by the holding means by a pressure means and constitutes a crushing means for the object to be crushed together with the holding means, Classifying means for recirculating the coarse particles of the material to be crushed that has passed through the crushing means to the holding means and discharging fine particles to the outside of the crusher, means for adjusting the classification characteristics of the classifying means, and crusher difference In a vibration predicting device for a crusher having a means for measuring pressure, a dynamic characteristic model of the crusher is used, and the amount of raw material to be crushed supplied to the crusher at present, and a roller addition for pressing the crushing roller against the holding means. Using the pressure, the classification characteristic command signal, the estimated value of the crushed object holding amount at the time of the previous calculation, and the crushability estimated value of the crushed object, the current estimated value of the crushed object held in the crusher and the holding means Calculate the estimated particle size distribution of the crushed object In addition, the first calculation means for calculating the crusher differential pressure estimated value based on the crushed object holding amount estimated value, and the current crushed object based on the deviation between the crusher differential pressure estimated value and the measured differential pressure value. A second calculation means for estimating the pulverizability of the object, a roller pressing force, an estimated value of the amount of the object to be pulverized, an estimated value of the particle size distribution of the object of pulverization on the retaining means, an estimated value of the pulverizability of the object A third calculating means for calculating a distribution of predicted values of vibration magnitude in a crusher state quantity space within a certain range centered on the crusher state quantity at the present time, based on the past vibration occurrence history that is sequentially learned And a vibration predicting device for a crusher. 4. A means for calculating a roller pressing force correction signal and a classification characteristic correction signal based on the vibration predicted value distribution calculated by the crusher vibration prediction device according to claim 1, and the above calculation. A controller based on vibration prediction of a crusher, characterized in that it is provided with means for adjusting the roller pressure of the crusher and the classification characteristics of the classifying means based on the correction signal.