# US 2013/0158738 A1 · Heat dissipation control system and control method thereof

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2013/0158738 A1
- Document ID: US20130158738A1
- Status: Published application
- Kind: A1
- Application number: 13/449,560
- Filed: 2012-04-18
- Published: 2013-06-20
- Earliest claimed: 2011-12-14
- Assignees: Inventec Corporation
- Inventors: Chien-An Chen; Chih-Chien Lin
- Web page: https://patentgrep.com/patent/US20130158738A1

## CPC

- G05D 23/1932: {to control the temperature of a plurality of spaces}
- G06F 1/20: Cooling means
- G06F 1/206: {comprising thermal management}
- G06N 3/045: Combinations of networks
- G06N 3/0499: Feedforward networks
- G06N 3/09: Supervised learning
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G05D: SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

A heat dissipation control system comprises a sensing unit, an artificial neural network computing unit, and two heat dissipation units. The artificial neural network computing unit performs computation for controlling based on a plurality of electronic-device temperatures sent out by the sensing unit. The computation for controlling performs a back propagation algorithm on an objective function which is defined a as the square of an error function. Accordingly, cooling effects suitable for the heat dissipation units are generated in order to achieve an optimum heat dissipation effect.

## Claims

### Claim 1

A heat dissipation control system for an electronic device, comprising:
a sensing unit for capturing a plurality of electronic-device temperatures of parts of the electronic device;
an artificial neural network computing unit, for receiving the electronic-device temperatures and comparing the electronic-device temperatures to a plurality of preset temperatures, respectively, wherein when the electronic-device temperatures are different from the corresponding preset temperatures the artificial neural network computing unit performs a computation for controlling based on the electronic-device temperatures, respectively, in order to generate a plurality of control signals, and wherein each of the computations for controlling include performing a back propagation algorithm on an objective function which is defined as a square of an error function; and
a heat dissipation unit, comprising a first heat dissipation element, the first heat dissipation element receiving the control signals and being driven by one of the control signals;
wherein the objective functions are defined as:
J
i
=
1
2

E
i
2
=
1
2

(
r
i
-
y
i
)
2
and wherein Ji is the objective function, Ei is the error function of the computations for controlling, ri is the preset temperature, and yi is the electronic-device temperature, for i: 1˜n.

### Claim 2 (depends on 1)

The heat dissipation control system as recited in claim 1, wherein the back propagation algorithm is adapted to adjust a weighted value of an artificial neural network for making the object function close to a minimal value, and the back propagation algorithm is defined as:
W
hy

(
k
+
1
)
=
W
hy

(
k
)
-
η

∂
J
i
∂
W
hy

(
k
)
and wherein Ji is the objective function, Whyi is the weighted value of the artificial neural network, k is sampling point, (k+1) is a next sampling point, and η is an adjustment speed rate, for i: 1˜n.

### Claim 3 (depends on 2)

The heat dissipation control system as recited in claim 2, wherein
∂
J
i
∂
W
hyi
=
∂
J
i
∂
E
i

∂
E
i
∂
y
i

∂
y
i
∂
v
i

∂
x
yi
∂
u
yi

∂
u
yi
∂
W
hyi
and wherein Ji is the objective function, Ei is the error functions of the computation for controlling, yi is the electronic-device temperature, νi is the control signals, xyi is an output of the computation for controlling, and uyi is an input of the computation for controlling, for i: 1˜n.

### Claim 4 (depends on 3)

The heat dissipation control system as recited in claim 3, wherein
∂
J
i
∂
E
i
=
(
r
i
-
y
i
)
∂
E
i
∂
y
i
=
-
1
∂
y
i
∂
v
i
=
sgn
[
(
y
i

(
k
)
-
y
i

(
k
-
1
)
)
*
(
v
i

(
k
-
1
)
-
v
i

(
k
-
2
)
]


∂
x
yi
∂
u
yi
=
1


∂
u
yi
∂
W
hyi
=
x
hi
and wherein Ji is the objective function, Ei is the error functions of the computation for controlling, ri is the preset temperatures, yi is the electronic-device temperature, k is a sampling point, (k−1) is a last sampling point, νi is the control signals, xyi is an output of the computation for controlling, uyi is an input of the computation for controlling, Whyi is the weighted value of the artificial neural network, and xhi is a gain variable.

### Claim 5 (depends on 4)

The heat dissipation control system as recited in claim 4, wherein each of the gain variables includes a ratio parameter, a differential parameter, and an integral parameter, and wherein the ratio parameter, the differential parameter, and the integral parameter are respectively defined as:
x
hi


1

(
k
)
=
{
1
,
u
hi


1

(
k
)
>
1
u
hi


1

(
k
)
,
-
1
≤
u
hi


1

(
k
)
≤
1
-
1
,
u
hi


1

(
k
)
<
-
1


x
hi


2

(
k
)
=
{
1
,
u
hi


2

(
k
)
>
1
u
hi


2

(
k
)
+
u
hi


2

(
k
-
1
)
,
-
1
≤
u
hi


2

(
k
)
≤
1
-
1
,
u
hi


2

(
k
)
<
-
1


x
hi


3

(
k
)
=
{
1
,
u
hi


3

(
k
)
>
1
u
hi


3

(
k
)
-
u
hi


3

(
k
-
1
)
,
-
1
≤
u
hi


3

(
k
)
≤
1
-
1
,
u
hi


3

(
k
)
<
-
1
and wherein xhi1 is the ratio parameter of one of the gain variables, xhi2 is the integral parameter of one of the gain variables, and xhi3 is the differential parameter of one of the gain variables, for i: 1˜n.

### Claim 6 (depends on 1)

The heat dissipation control system as recited in claim 1, wherein the first heat dissipation element has a plurality of weighted values corresponding to each part of the electronic device, and wherein each of the control signals is multiplied by each of the weighted values generated by the artificial neural network computing unit for generating a plurality of driving signals.

### Claim 7 (depends on 6)

The heat dissipation control system as recited in claim 6, wherein the first heat dissipation element receives the driving signals at the same time and is driven by the driving signal corresponding to a highest rotational speed value.

### Claim 8 (depends on 1)

The heat dissipation control system as recited in claim 1, wherein the heat dissipation unit further comprises a second heat dissipation element, the second heat dissipation element having a plurality of weighted values corresponding to each part of the electronic device, wherein each of the control signals is multiplied by each of the weighted values generated by each of the artificial neural network computing units for generating a plurality of driving signals.

### Claim 9 (depends on 1)

The heat dissipation control system as recited in claim 1, wherein the first heat dissipation element is a fan.

### Claim 10

A heat dissipation control method, comprising:
capturing a plurality of electronic-device temperatures corresponding to different parts of an electronic device by a sensing unit;
receiving the electronic-device temperatures by an artificial neural network computing unit respectively, and comparing each of the electronic-device temperatures with a corresponding preset temperature;
when the electronic-device temperatures are different from the preset temperatures, performing a computation for controlling by the artificial neural network computing unit based on the electronic-device temperatures for generating a plurality of corresponding control signals, and performing a back propagation algorithm by the computation for controlling on an objective function which is defined as a square of an error function, wherein the error function is generated by having the corresponding electronic-device temperature subtracted by a corresponding one of the preset temperatures;
outputting the control signals to a heat dissipation unit simultaneously from the artificial neural network computing unit; and
driving the heat dissipation unit by a corresponding control signal.

### Claim 11 (depends on 10)

The heat dissipation control method as recited in claim 10, wherein the step of outputting the control signals to the heat dissipation unit simultaneously from the artificial neural network computing unit further comprises:
multiplying each of the control signals by weighted values which reflect the heat dissipation effect of the dissipation unit to elements to be cooled in order to generate a plurality of corresponding driving signals, and transmitting the driving signals to the heat dissipation unit, respectively; and
driving the heat dissipation unit by the corresponding driving signals.

### Claim 12 (depends on 11)

The heat dissipation control method as recited in claim 11, wherein the step of driving the heat dissipation unit by the corresponding driving signals further comprises a step of:
driving the heat dissipation unit by the driving signals corresponding to the highest rotational speed.

### Claim 13

A heat dissipation control system for a server, comprising:
a sensing unit for capturing a plurality of electronic-device temperatures of parts of the server;
an artificial neural network computing unit, for receiving the electronic-device temperatures and comparing the electronic-device temperatures to a plurality of preset temperatures, respectively, wherein when the electronic-device temperatures are different from the corresponding preset temperatures the artificial neural network computing unit performs a computation for controlling based on the electronic-device temperatures, respectively, in order to generate a plurality of control signals, and wherein each of the computations for controlling include performing a back propagation algorithm on an objective function which is defined as a square of an error function; and
a fan unit, comprising a first fan, the first fan receiving the control signals and being driven by one of the control signals;
wherein the objective functions are defined as:
J
i
=
1
2

E
i
2
=
1
2

(
r
i
-
y
i
)
2
and wherein Ji is the objective function, Ei is the error function of the computations for controlling, ri is the preset temperature, and yi is the electronic-device temperature, for i: 1˜n.

### Claim 14 (depends on 13)

The heat dissipation control system as recited in claim 13, wherein the back propagation algorithm is adapted to adjust a weighted value of an artificial neural network for making the object function close to a minimal value, and the back propagation algorithm is defined as:
W
hy

(
k
+
1
)
=
W
hy

(
k
)
-
η

∂
J
i
∂
W
hy

(
k
)
and wherein Ji is the objective function, Whyi is the weighted value of the artificial neural network, k is sampling point, (k+1) is a next sampling point, and η is an adjustment speed rate, for i: 1˜n.

### Claim 15 (depends on 14)

The heat dissipation control system as recited in claim 14, wherein
∂
J
i
∂
W
hyi
=
∂
J
i
∂
E
i

∂
E
i
∂
y
i

∂
y
i
∂
v
i

∂
x
yi
∂
u
yi

∂
u
yi
∂
W
hyi
and wherein Ji is the objective function, Ei is the error functions of the computation for controlling, yi is the electronic-device temperature, νi is the control signals, xyi is an output of the computation for controlling, and uyi is an input of the computation for controlling, for i: 1˜n.

### Claim 16 (depends on 15)

The heat dissipation control system as recited in claim 15, wherein
∂
J
i
∂
E
i
=
(
r
i
-
y
i
)
∂
E
i
∂
y
i
=
-
1
∂
y
i
∂
v
i
=
sgn
[
(
y
i

(
k
)
-
y
i

(
k
-
1
)
)
*
(
v
i

(
k
-
1
)
-
v
i

(
k
-
2
)
]


∂
x
yi
∂
u
yi
=
1


∂
u
yi
∂
W
hyi
=
x
hi
and wherein Ji is the objective function, Ei is the error functions of the computation for controlling, ri is the preset temperatures, yi is the electronic-device temperature, k is a sampling point, (k−1) is a last sampling point, νi is the control signals, xyi is an output of the computation for controlling, uyi is an input of the computation for controlling, Whyi is the weighted value of the artificial neural network, and xhi is a gain variable.

### Claim 17 (depends on 16)

The heat dissipation control system as recited in claim 16, wherein each of the gain variables includes a ratio parameter, a differential parameter, and an integral parameter, and wherein the ratio parameter, the differential parameter, and the integral parameter are respectively defined as:
x
hi


1

(
k
)
=
{
1
,
u
hi


1

(
k
)
>
1
u
hi


1

(
k
)
,
-
1
≤
u
hi


1

(
k
)
≤
1
-
1
,
u
hi


1

(
k
)
<
-
1


x
hi


2

(
k
)
=
{
1
,
u
hi


2

(
k
)
>
1
u
hi


2

(
k
)
+
u
hi


2

(
k
-
1
)
,
-
1
≤
u
hi


2

(
k
)
≤
1
-
1
,
u
hi


2

(
k
)
<
-
1


x
hi


3

(
k
)
=
{
1
,
u
hi


3

(
k
)
>
1
u
hi


3

(
k
)
-
u
hi


3

(
k
-
1
)
,
-
1
≤
u
hi


3

(
k
)
≤
1
-
1
,
u
hi


3

(
k
)
<
-
1
and wherein xhi1 is the ratio parameter of one of the gain variables, xhi2 is the integral parameter of one of the gain variables, and xhi3 is the differential parameter of one of the gain variables, for i: 1˜n.

### Claim 18 (depends on 14)

The heat dissipation control system as recited in claim 14, wherein the first fan has a plurality of weighted values corresponding to each part of the server, and wherein each of the control signals is multiplied by each of the weighted values generated by the artificial neural network computing unit for generating a plurality of driving signals.

### Claim 19 (depends on 18)

The heat dissipation control system as recited in claim 18, wherein the first fan receives the driving signals at the same time and is driven by the driving signal corresponding to a highest rotational speed value.

### Claim 20 (depends on 14)

The heat dissipation control system as recited in claim 14, wherein the fan unit further comprises a second fan, the second fan having a plurality of weighted values corresponding to each part of the server, wherein each of the control signals is multiplied by each of the weighted values generated by each of the artificial neural network computing units for generating a plurality of driving signals.