SU1642477A1 - Multichannel regression meter - Google Patents

Abstract

The invention relates to specialized digital computing devices for determining the statistical characteristics of random processes and can be used to study automation objects in various industries. The purpose of the invention is to improve the accuracy of determining regression parameters by adjusting the calculations to account for workplace noise. The regressiometer contains a control unit, averaging units, a subtraction unit and information processing channels, each of which contains a regression coefficient determination unit, a switch, a recovery node, a memory unit, an adder. Recovery nodes consist of memory blocks, a multiplier and an adder. 1 ZR-ON f-ly, 4 silt "with (L

Description

The invention relates to specialized digital computing devices for determining the statistical characteristics of random processes and can be used in the study of automation objects in various industries.

The purpose of the invention is to improve the accuracy of determining the regression parameters by adjusting the calculations taking into account the noise of the workplace.

Initially, the regression coefficients are calculated in each i-th (i 1, M) channel regression meter on an input sample of {Y9jl, j 1, M, -fixed signals in the memory block

When recalculating the regression, the input signal of each channel regression meter can be adjusted according to the formula

 Y.

-ps

(one)

 WITH

 -9J1 -j

where n is the j -e value of the noise estimate.

 workplace. As an estimate of workplace noise, one can choose the value of the noise component averaged over all signals

h- | h ;. ")

where n jj is the j-e sample value

noise component estimates in the i-th channel. The estimate of the noise component in the i-th channel can be determined by the formula

oh

4 N) -U J

n Y9J, - Y9Ji

(3)

where vq,; recovered value

input signal based on the evaluation of regression parameters „

In accordance with formulas (1) - (3), the input signal of the regression meter in the i-th channel is adjusted according to the formula

L. (- f- -I

v9Jl j; -44l + mLy3iJ (4) where is the expectation or averaging over all channels,

and implemented using the difference block and channel adder Averaging the sample values of the input signals of the in channels (J is performed in the first averaging block, averaging the sample recovered values or the output signals of the recovery nodes in the second averaging block

Channel recovery nodes calculate the restored input signal values based on the results of the first estimation of regression parameters in channel regression meters. Channel memory blocks are used to store sample values | Y9; j | 2j 1, NJ of the input signal

Figure 1 shows the scheme of a multichannel regressiometer; Fig. 2 is a control block diagram; FIG. 3 is a diagram of a channel recovery block; FIG. 4 is a block diagram of determining regression coefficients for a linear case.

The regressiometer contains information processing channels, each of which includes a regression coefficient determination unit 1, a switch 2, an adder 3, a memory unit 4 and a recovery node 5 (Fig. 1 3) consisting of memory blocks 6 and 7, multiplication unit 8 and adder 9, blocks 10 and 11 of averaging, block 12 of subtraction, block 13 of control (FIG. 2), states of triggers 14 and 15, comparison circuit 16, delay element 17, And 18, the single-oscillator 19 , the switch 20, the element And 21, the element OR 22, the generator 23 clock pulses, counter 24, inputs 25-27 and outputs 28-

33.

I

The regression coefficients determination unit 1 (Fig. 4) contains multiplier 34, adder 35, multipliers 36-39, subtractors 40 and 41, and adder 42. The outputs of the regression meter and the relationship between the blocks are labeled 43.1-43, t, 44.1- 440t and 45.1-45.t.

The channel block is designed to determine the parameters of the regression equation. For example, in the linear case the regression equation has the form

v

RO; +

i 1, M,

(five)

where the parameters of the regression equation

01

and P

it

calculated according to

Poi

a, tn; LYj +

a3m;

with formulas:

  (6) Y +, (7)

where a, 2 (2N + 1) / (M-1); a -6 / (N-1);

aE a2

a4 12 / (N-1) (N + 1);

t-

g- - i 1 m

; D YJ-i5gJV.

3JI

N

J

- selective input value

regression signal;

j 1, N is the number of samples;

i 1, M - the number of channels.

The normalizing blocks multiply the output signals of the drives by the values of the coefficients.

The regression meter operation is characterized by two modes: the first and the second, which is determined by the state of the trigger 14 in the control unit 13. Each mode is divided into two stages: Operation and Waiting, which is determined by the state of the trigger 15.

The start signal on input 27 triggers a 14 state to one (first mode). In addition, the Trigger signal through the element OR 22 is fed to the input of the trigger 15 and translates it into a single state (Step 1). From the output of the element OR 22, the signal arrives at the inlet of zeroing of the address counter 24 and brings it to the initial state. In addition, this signal enters output 29.

The signal 33 from the output of the control unit sets the channel switches and channel memory blocks to read the input signals. Switch 20 connects to the counting input of counter 24 an external signal. Synchronization on input 26 (element 21 at this stage of operation) is open. The device begins to accept sample values of the input signal arriving at the information inputs and accompanied by an external sync signal. When the next external sync signal arrives, the content of the counter increases. 24 per unit. At the same time, the output of the counter sets the address of the memory cell of the memory block and the numbers are processed in block 1 of the input signal values. In addition, the signal external synchronization enters through the synchronization inputs 17 of the channel memory block and block 1, ensuring that the current value of the input signal is recorded in the corresponding cell0. At the same time, in block 1, the next input signal is processed until the value of the 24 address counter exceeds the threshold level set in the comparison circuit 16, which corresponds to the processing of the last value of the input sample, a signal appears at the output of the And 18 element. This signal is present during the entire sync pulse lsa. At the end of the sync pulse, the one-shot generates a control pulse (on the falling edge), which arrives at trigger 15 and resets it, putting the regressometer into the Idle stage. In this case, the zero potential at the output of the trigger locks the element And 21, blocking the external clock signals. In addition, signals from the one-shot output are fed to output 32 and from there to the channel nodes of the recovery, Thus, the regression coefficients from the output of the channel block 1 are recorded in the internal memory elements.

Stage Wait until an external signal is received. Recalculate input 25. When this signal arrives at trigger input 14, the latter is brought to the zero state, the channel memory blocks are switched to read information, the channel switch is processed by the corrected values input signals, and the switch 20 - for synchronization from the internal clock generator “In this case, the signal is recalculated over the element OR 22, but the counter 24

five

0

five

and goes to output 29, from there to channel blocks 1, returns them to the initial state. The 15 g trigger translates to the state corresponding to the Operation stage, and its output signal is started by the generator 23. The further operation of the control unit is similar to that of the first 0. Modes with the difference that, instead of external synchronization, the generator pulses 23 clock pulses are used.

| In the Recalculate mode, the block 1. signals are received from the adder, which forms the corrected input signal values as a sum of signals from channel memory blocks and a signal from the difference between the values of signals from averaging blocks in accordance with the formula (4) "Recovery nodes based on stored regression coefficients and current implementations of the address counter form the theoretical values of the input signals in accordance with the formula (5) о These values are averaged in averaging block 11, | From the averaging result in block 12, the result is tat averaging Bloch 0 girdle 10 input signals read out of the channel blocks of memory values, and thus obtained a correction signal, identical for all channels of

5 The correction signal is added to the channel input signal in the channel adders and fed to the input of block 1. Blocks 1 process the corrected values of the input signals of the regression meter, form the corrected values of the regression coefficients. At the end of the second mode, the regressometer output contains the corrected values of 5 regression coefficients, which can be used by external devices. External devices can also use output signals from recovery nodes, with 0 being the first signal, Recalculating, the recovered Y values will be killed without taking into account workplace noise; if the signal is recalculated. Recalculating output signals will be 5 input signal estimates taking into account workplace noise.

Upon subsequent connections of the recalculation signal, further approximations of the coefficients will be generated.

regression and restored values of the input signal

Claims (1)

Invention Formula 1. A multichannel regression meter comprising a control unit, a subtraction unit and a first information processing channel consisting of a first memory block, a first adder, a switch, the first and second outputs of a control unit are connected to a synchronization input and a recording control input of a first memory block, different By the fact that, in order to increase accuracy, two averaging blocks and additional information processing channels, similar to the first one, are introduced into it, each of which has two memory blocks, a second adder, a multiplication unit and determining the regression coefficients, the first, second and third outputs of the control unit are connected respectively to the synchronization inputs and control inputs of the recording of the first memory blocks of additional information processing channels and the control inputs of reading the first memory blocks of all information processing channels whose address inputs are connected to the fourth output of the control unit; the outputs of the first memory blocks of all information processing channels are connected to the first inputs of the first adders of the same name channels information processing and with the corresponding inputs; the first averaging unit, the output of which is connected to the input of the subtracted subtraction unit, the output of which is connected to the second inputs of the first adders of all information processing channels, the outputs of each of which are connected to the first information input of the switch of its channel, the second information input of which is connected information processing channel and is the corresponding channel information input of the regressiometer, the first and second control inputs of the switches with dineny respectively with the second and third outputs of the control unit, the output of each channel switch information processing unit is connected to data input determining factor ten 15 20 25 30, 24778 regression of their own channel, the first and second outputs of which are related to the corresponding channel outputs of the first and second coefficients of the corresponding linear regression equation and connected to the information inputs of the second and third memory blocks of the same name, respectively, synchronization inputs, setting and address inputs of the coefficient determining blocks regressions of all information processing channels are connected respectively to the first, fifth and fourth outputs of the control unit, the fourth output of which connected to the first inputs of the multiplication blocks of all information processing channels; the sixth output of the control unit is connected to the read control inputs of the second and third memory blocks of all channels; the first and second outputs of the second memory block of each information processing channel are connected respectively to the first input of the second adder and address the input of the third memory block of this channel, the output of which is connected to the second input of the multiplication unit of its information processing channel, the output of which is connected to the second input of the second sum the information processing channel of the same name, the outputs of the second adders of all information processing channels are connected to the inputs of the second averaging unit, the output of which is connected to the input of the decremented subtraction unit, the start, recalculation and synchronization inputs of the control unit are the same names of the regimenter 35 40 0 five 2 "The regression meter that contains the control contains a switch, an element under item 1, about t - the fact that the block has two triggers, OR, the elements of the delay element, 0 five And, the comparison circuit, one-shot, counter and clock generator, the output of which is connected to the first information input of the switch, the output of which is the first output of the block and connected to the counting input of the counter and through the delay element to the first input of the first element AND whose output is connected to the one-shot input, the output of which is the sixth output of the block and connected to the zero input of the first trigger, the direct output of which is connected to the trigger input of the clock generator and the first input of the second element And, the second input of which is the synchronization input of the block, and the output is connected to the second information input of the switch, the zero input of the second trigger is connected to the first input of the OR element and is the start input of the block, the recalculation input of which is the single input of the second trigger connected to the second the input of the OR element, the output of which is the fifth output five block, and connected to the zero input of the first trigger, the forward output of the second trigger is the second output of the block and connected to the first control input of the switch, the inverse output of the second trigger is the third output of the block and connected to the second control input of the switch, the output of the OR element connected to the zeroing input of the counter, the output of which is the fourth output of the block, and through a comparison circuit connected to the second input of the first element AND, 28 30 33 1 / F h / GGT 28 30 33 13 , 28 2Ј -thirty --3 / -z 33 45.1 ... 4-5. t F 31 28 333 $ 43 t No. t OSHU1 33 30 ZЈ 9Z 2 GF about si oil LLKW ( FIG L